30. April 2025 · Transkript

ÜBER DIESE EPISODE

Transforming Data Science Strategies: From Plans to Behavioral Commitments

While navigating the intricate world of data science, I’ve encountered countless misguided attempts at formulating strategies. The realization struck me that many organizations often mistake detailed plans for effective strategies. I remember a particular workshop I facilitated where a financial services company presented their 18-month plan, which was essentially obsolete within months due to shifting market conditions. This experience served as a turning point in understanding how a genuine data strategy transcends mere activities and instead focuses on establishing behavioral commitments that truly differentiate organizations.Understanding Plans vs. StrategiesDefining Plans and StrategiesLet’s start by clarifying what we mean by plans and strategies. A plan typically includes a list of tasks, timelines, and deliverables. It’s like a roadmap, guiding us step by step. In contrast, a strategy is broader. It involves a commitment to a specific pattern of behavior intended to achieve long-term goals. As Gary Pisano aptly puts it, “A strategy is nothing more than a commitment to a pattern of behavior intended to help win a competition.” This distinction is crucial for any organization wanting to thrive.Common Misconceptions in OrganizationsMany organizations fall into the trap of thinking that having a detailed plan equates to having a solid strategy. This leads to confusion and sometimes frustration. After all, plans can become obsolete quickly, especially in fast-paced environments. Have you ever witnessed a team cling to a rigid plan, only to watch it fail when market conditions change?* Misconception: Plans are effective substitutes for strategy.* Reality: Plans without a guiding behavioral framework often lead to subpar outcomes.The Impact of Market Changes on Rigid PlanningHere’s a thought to ponder: how often do market conditions shift unexpectedly? If your organization relies solely on a fixed plan, you might find yourself at a disadvantage. For instance, I saw a financial services company with an 18-month project plan. This plan quickly became outdated as market dynamics shifted. The lack of flexibility crippled their ability to adapt.In contrast, teams that adopt a more fluid approach can pivot when necessary. They can respond to changes in consumer behavior, regulations, or competitor actions. This adaptability is a core component of a true strategy.Behavioral Commitments vs. Task ListingsLet’s talk about behavioral commitments. These are the underlying principles guiding a team’s actions. They go beyond merely completing tasks. I’ve worked with data science teams that excelled when they focused on how they wanted to behave rather than just what they needed to do. A healthcare analytics team I encountered had an extensive tactical plan but was often unsure about their guiding principles. They struggled to defend their approach, leading to inefficiency.In contrast, successful teams prioritize their commitments. They decide on their guiding behaviors first, and then plan tactically around them. It’s about creating a culture that supports innovation and risk-taking.Case Study: The Healthcare Analytics TeamThe illustrative case of the healthcare analytics team highlights this phenomenon well. They created a detailed tactical plan but faced challenges due to a lack of coherent behavioral principles. They found it tough to navigate the complex landscape of healthcare data without a strong strategic foundation. In essence, their plan was rigid, while a strategy could have allowed for more flexibility and a better alignment with evolving priorities.Reflections on the Evolution of Strategic ThoughtAs I reflect on my experiences, I see how strategic thought has evolved. There’s a growing recognition that true strategies require adaptability and coherence. I often encourage teams to focus on three essential requirements for successful strategies:* Consistency: This means decisions should support the same competitive advantage over time.* Coherence: All commitments should align to avoid conflicting priorities.* Alignment: The strategy must connect with broader organizational goals.When teams embrace these principles, they are better equipped to deal with uncertainties. They become agile, able to respond to changes without losing sight of their goals.In conclusion, understanding the difference between plans and strategies can profoundly impact an organization’s effectiveness. By moving beyond rigid planning and embracing a strategy rooted in behavioral commitments, we can position our teams to thrive in an ever-changing landscape.The Core Ingredients for Successful StrategiesCrafting a successful strategy is like baking a cake. You need the right ingredients to achieve a delightful outcome. In my journey through the world of data science and analytics, I’ve identified three essential requirements for any successful strategy: consistency, coherence, and alignment.1. Consistency in Decision-Making FrameworksConsistency is crucial. It ensures that every decision reinforces the same competitive advantage over time. Think about it: if a team continually shifts its focus, how can it expect to build a solid foundation? Imagine a ship with no steady course; it will drift aimlessly.In my experience, I have seen many teams struggle with this. For instance, a retail analytics team I worked with prioritized customer-facing improvements instead of jumping between various short-term projects. This decision-making framework allowed them to adapt swiftly during the pandemic when shopping patterns changed. Consistency in their approach led to measurable business impact.2. Coherence Among Behavioral CommitmentsCoherence is another key component. It demands that all behavioral commitments support each other. Without coherence, teams can face the dreaded conflicting priorities. Imagine two gears trying to turn in opposite directions; they will only grind against each other, leading to inefficiency and frustration.* Example: A healthcare analytics team created an extensive tactical plan but struggled because their guiding principles were unclear. They faced internal conflicts that stifled progress.* Insight: I learned that prioritizing clear behavioral commitments can minimize these conflicts.When commitment is coherent, every action taken aligns with the team’s objectives. This creates synergy, allowing everyone to work towards common goals without distraction.3. Alignment with Organizational GoalsLastly, alignment is vital. A strategy must connect with the broader organizational goals. If a team’s actions don’t correlate with the organization’s objectives, the results can be undesirable.Consider this: a retail analytics team once focused heavily on sophisticated customer value modeling. They soon realized this focus misaligned with their organization’s strategy based on supply chain efficiency. This led to wasted resources and confusion. When teams are aligned, they contribute to the overall success of the organization.Understanding the Consequences of Conflicting PrioritiesConflicting priorities can cripple a team's efforts. In the pursuit of excellence, teams often take on too much, leading to chaos. This is like juggling too many balls at once; eventually, some will fall. I’ve seen this firsthand in a pharmaceutical research team with dual commitments to high data quality and rapid development. Their conflicting priorities caused internal conflict, making it hard to achieve either goal effectively.As leaders, we must recognize these challenges. Are we clear on our priorities? Are we supporting our teams with the right frameworks to achieve their goals? This reflection is essential for fostering a productive environment.Personal Insights from Strategy Formulation ExperiencesThroughout my experiences in strategy formulation, I’ve learned the importance of evaluation. I often ask myself: “Are we focusing on the right behaviors?” An effective strategy should articulate how a team will behave differently compared to competitors. It’s not just about what we do; it’s about how we do it.I’ve seen organizations benefit from rethinking their approach. By emphasizing behavioral patterns, we can establish a solid foundation for success. For small teams, succinct strategic commitments can drive efficiency. In one instance, a four-person team at a retailer created a one-page document outlining three simple behavioral commitments. This clarity significantly improved their focus and performance.In summary, understanding the core ingredients for successful strategies—consistency, coherence, and alignment—can drive sustained competitive advantage. Reflecting on these elements can help teams navigate the complexities of their environments and achieve lasting success. Let’s continue to evaluate and refine our strategies to ensure they are robust and effective in meeting organizational objectives.Lessons from R&D: Innovative Approaches to Strategic ThinkingWhen we think about R&D strategies, it’s easy to overlook their value, especially in data science. But these strategies can provide us with essential frameworks for navigating uncertainty and fostering innovation. Let's explore how we can learn from R&D to improve our strategic thinking in data science.1. How R&D Strategies Can Inform Data ScienceR&D strategies often focus on exploration and experimentation. This is crucial for data science, where the landscape changes rapidly. Just think about it: how often do new tools or methods emerge that can shift your entire approach? R&D teaches us that embracing uncertainty can lead to significant breakthroughs.In my experience, teams that adopt a mindset similar to R&D tend to be more adaptable. They commit to ongoing learning, allowing them to pivot when necessary. For instance, a data analytics team in a retail environment saw massive improvements when they started treating each project as a learning opportunity rather than a

Become a supporter of this podcast: https://www.spreaker.com/podcast/datascience-show-podcast--6817783/support.

I share practical AI leadership notes on LinkedIn — the kind you can forward internally or reuse in executive discussions.
Follow Mirko on LinkedIn if you want decision-ready frameworks, not hype.

So geht’s

Bildung

Technologie

Englisch

Vereinigte Staaten

Verwenden SieAPI, um diese Episode inJSONabzurufen

IN DIESER FOLGE

# gary pisano

Diese Tags bearbeiten

TRANSKRIPT 🔗

1
00:00:00,100 --> 00:00:02,303
MIRKO PETERS: What if I told you
that most data science teams
2
00:00:02,441 --> 00:00:05,965
don't have a Strategy at all?
They have plans, roadmaps, and
3
00:00:06,027 --> 00:00:08,891
backlogs. But these aren't
strategies. They're just lists
4
00:00:08,892 --> 00:00:10,750
of activities without a guiding
principle.
5
00:00:11,352 --> 00:00:14,094
In this video, I'm going to
share the exact framework I use
6
00:00:14,095 --> 00:00:17,203
to build successful data science
strategies for teams ranging
7
00:00:17,234 --> 00:00:20,047
from three people at a startup
to 50 plus at a Fortune 500
8
00:00:20,062 --> 00:00:24,031
company. You'll learn how to
create a Strategy that focuses
9
00:00:24,266 --> 00:00:28,031
your team's efforts and drives
measurable business impact.
10
00:00:29,156 --> 00:00:33,639
The one-sentence definition that
changes everything. What if most
11
00:00:33,659 --> 00:00:36,202
data science leaders are getting
Strategy completely wrong?
12
00:00:37,241 --> 00:00:40,491
I've consulted with dozens of
teams across industries, and
13
00:00:40,569 --> 00:00:43,671
almost every time I ask to see
their data science Strategy,
14
00:00:44,194 --> 00:00:47,741
they proudly present me with a
list. Sometimes it's a list of
15
00:00:47,835 --> 00:00:51,725
projects they plan to complete.
Other times, it's technologies
16
00:00:51,741 --> 00:00:53,022
they intend to implement.
17
00:00:53,241 --> 00:00:56,116
Occasionally, it's hiring plans
or organizational structures.
18
00:00:56,319 --> 00:00:59,074
But here's the problem. None of
these things constitute a
19
00:00:59,114 --> 00:01:02,556
Strategy. This misconception
isn't unique to data science.
20
00:01:03,216 --> 00:01:06,302
Organizations everywhere confuse
plans with Strategy.
21
00:01:07,326 --> 00:01:09,724
The difference might seem
semantic, but it creates
22
00:01:09,802 --> 00:01:12,865
profound consequences for data
teams trying to deliver value.
23
00:01:13,607 --> 00:01:17,873
When we treat Strategy as merely
a roadmap of activities, we fail
24
00:01:17,888 --> 00:01:20,607
to address the fundamental
question that should guide our
25
00:01:20,669 --> 00:01:20,935
work.
26
00:01:21,841 --> 00:01:25,196
How will we behave to create
competitive advantage? Consider
27
00:01:25,216 --> 00:01:27,679
a financial services company I
worked with last year.
28
00:01:28,739 --> 00:01:33,343
Their data science Strategy was
a detailed 18-month plan listing
29
00:01:33,382 --> 00:01:36,007
predictive models they would
build, a data lake
30
00:01:36,046 --> 00:01:39,812
implementation timeline, and
hiring targets for specific
31
00:01:40,132 --> 00:01:41,077
technical roles.
32
00:01:42,280 --> 00:01:46,140
Six months in, market conditions
shifted dramatically and
33
00:01:46,202 --> 00:01:49,452
customer needs changed. Their
beautifully crafted plan became
34
00:01:49,546 --> 00:01:50,843
irrelevant almost overnight.
35
00:01:51,796 --> 00:01:54,578
Without guiding principles for
behavior, the team floundered,
36
00:01:54,719 --> 00:01:57,301
chasing new priorities while
abandoning half-completed
37
00:01:57,321 --> 00:02:01,043
initiatives. In another case, a
healthcare analytics team spent
38
00:02:01,285 --> 00:02:04,848
months creating what they called
a comprehensive data Strategy.
39
00:02:05,512 --> 00:02:08,957
The document outlined specific
tools, methodologies, and
40
00:02:08,988 --> 00:02:11,973
projects, all tactical
components without a unifying
41
00:02:12,020 --> 00:02:14,660
strategic direction. When
leadership changed, the new
42
00:02:14,692 --> 00:02:17,535
executive questioned these
choices. Without clear
43
00:02:17,598 --> 00:02:21,020
behavioral commitments tied to
competitive advantage. The team
44
00:02:21,080 --> 00:02:22,702
couldn't defend their approach.
The result?
45
00:02:23,262 --> 00:02:26,364
Resources were redirected, team
morale collapsed, and months of
46
00:02:26,407 --> 00:02:29,286
work went to waste. These
examples reflect what happens
47
00:02:29,372 --> 00:02:33,130
when we mistake activity for
Strategy. The plans themselves
48
00:02:33,255 --> 00:02:36,692
weren't necessarily bad. They
were just insufficient. They
49
00:02:36,762 --> 00:02:39,512
answered what without addressing
how or why.
50
00:02:40,778 --> 00:02:44,262
They focused on tasks rather
than behavioral patterns that
51
00:02:44,325 --> 00:02:47,044
would differentiate the team and
create value regardless of
52
00:02:47,059 --> 00:02:50,021
changing circumstances. This
brings us to the definition that
53
00:02:50,082 --> 00:02:54,166
transformed my own understanding
of Strategy. Gary Pisano, in his
54
00:02:54,248 --> 00:02:57,990
work on R&D Strategy, offers
this powerful insight. A
55
00:02:58,006 --> 00:03:00,896
Strategy is nothing more than a
commitment to a pattern of
56
00:03:00,912 --> 00:03:03,131
behavior intended to help win a
competition.
57
00:03:04,232 --> 00:03:08,084
This elegant formulation cuts
through the confusion and
58
00:03:08,178 --> 00:03:11,303
provides clarity about what
Strategy actually is. Let's
59
00:03:11,381 --> 00:03:13,849
break this down for data science
teams specifically.
60
00:03:14,849 --> 00:03:18,332
First, Strategy is a commitment
to a pattern of behavior. It's
61
00:03:18,352 --> 00:03:21,432
not about specific projects or
technologies. Those are tactical
62
00:03:21,512 --> 00:03:24,653
choices that should flow from
your Strategy. Instead, it's
63
00:03:24,692 --> 00:03:29,317
about how your team consistently
acts and makes decisions. Will
64
00:03:29,332 --> 00:03:32,598
you prioritize speed over
perfection? Will you focus on
65
00:03:32,614 --> 00:03:35,793
specialized expertise or
versatile capabilities? Will you
66
00:03:35,809 --> 00:03:40,543
develop proprietary methods or
leverage open source solutions?
67
00:03:41,496 --> 00:03:44,590
These behavioral patterns
consistently applied define your
68
00:03:44,637 --> 00:03:48,151
Strategy. Second, this behavior
is intended to help win a
69
00:03:48,171 --> 00:03:51,132
competition. Data science
doesn't exist in a vacuum.
70
00:03:51,694 --> 00:03:55,995
Your team competes for resources
within your organization, for
71
00:03:56,058 --> 00:03:59,722
talent in the marketplace, and
ultimately to create value that
72
00:03:59,784 --> 00:04:03,519
helps your company outperform
competitors. A true Strategy
73
00:04:03,878 --> 00:04:07,565
acknowledges this competitive
context and makes explicit
74
00:04:07,675 --> 00:04:09,956
choices about where and how to
compete.
75
00:04:11,128 --> 00:04:14,031
For data science leaders, this
definition is transformative
76
00:04:14,091 --> 00:04:17,373
because it shifts focus from
project deliverables to the
77
00:04:17,416 --> 00:04:21,138
decision-making frameworks that
guide all activities. When
78
00:04:21,216 --> 00:04:24,279
market conditions change or new
technologies emerge, a project
79
00:04:24,341 --> 00:04:27,599
list quickly becomes obsolete.
But behavioral commitments can
80
00:04:27,646 --> 00:04:31,021
remain consistent even as
tactical execution evolves.
81
00:04:31,271 --> 00:04:34,349
Take a retail analytics team I
advised that completely
82
00:04:34,427 --> 00:04:37,490
reimagined their approach using
this definition. Instead of
83
00:04:37,537 --> 00:04:40,624
listing dashboard projects and
machine learning models, They
84
00:04:40,664 --> 00:04:44,048
committed to behavioral
patterns. They would prioritize
85
00:04:44,108 --> 00:04:47,311
customer-facing improvements
over internal efficiencies,
86
00:04:47,831 --> 00:04:52,436
favor rapid experimentation over
lengthy perfectionism, and
87
00:04:52,514 --> 00:04:56,100
emphasize business-intelligible
algorithms over black-box
88
00:04:56,303 --> 00:04:56,882
solutions.
89
00:04:57,821 --> 00:05:00,962
These commitments didn't specify
what they would build, but they
90
00:05:01,322 --> 00:05:04,203
created clear guidelines for how
they would operate in any
91
00:05:04,261 --> 00:05:07,242
circumstance. When retail
shopping patterns drastically
92
00:05:07,304 --> 00:05:10,421
changed during the pandemic,
teams operating from project
93
00:05:10,445 --> 00:05:11,640
plans scrambled to adjust.
94
00:05:12,460 --> 00:05:15,249
But this team's behavioral
Strategy provided immediate
95
00:05:15,265 --> 00:05:17,921
guidance on how to respond.
Their commitment to
96
00:05:17,968 --> 00:05:21,109
customer-facing improvements
meant quickly shifting to new
97
00:05:21,171 --> 00:05:25,203
purchasing patterns. Their rapid
experimentation approach allowed
98
00:05:25,234 --> 00:05:28,130
them to test and learn in
uncertain conditions.
99
00:05:28,310 --> 00:05:31,074
Their focus on intelligible
algorithms meant business
100
00:05:31,132 --> 00:05:33,714
partners could understand and
trust their evolving
101
00:05:33,757 --> 00:05:36,976
recommendations. This strategic
approach changes everything
102
00:05:37,039 --> 00:05:40,726
about how data science teams
operate. Instead of asking, what
103
00:05:40,804 --> 00:05:44,070
projects should we work on next
quarter? It prompts more
104
00:05:44,086 --> 00:05:45,070
fundamental questions.
105
00:05:46,070 --> 00:05:49,070
How should we behave to create
maximum competitive advantage?
106
00:05:49,492 --> 00:05:52,320
What patterns of decision-making
will consistently lead to
107
00:05:52,367 --> 00:05:56,137
success regardless of specific
projects? The transformation
108
00:05:56,177 --> 00:05:58,257
extends to how teams evaluate
success.
109
00:05:58,697 --> 00:06:01,396
Rather than measuring progress
against a project timeline,
110
00:06:01,478 --> 00:06:05,463
which rewards activity but not
necessarily impact, they assess
111
00:06:05,556 --> 00:06:08,517
whether their behavioral
patterns are creating the
112
00:06:08,580 --> 00:06:12,095
intended competitive advantage.
Are they making decisions faster
113
00:06:12,127 --> 00:06:15,095
than competitors? Are they
connecting with customers more
114
00:06:15,127 --> 00:06:18,986
effectively? Are they generating
insights that others miss? For
115
00:06:19,049 --> 00:06:22,377
team members, this definition
provides clarity and autonomy.
116
00:06:23,265 --> 00:06:25,825
When everyone understands the
behavioral patterns that define
117
00:06:25,826 --> 00:06:28,567
the Strategy, they can make
consistent decisions without
118
00:06:28,607 --> 00:06:31,708
constant supervision. They know
how to prioritize when competing
119
00:06:31,747 --> 00:06:34,849
demands arise and can adapt to
changing circumstances while
120
00:06:34,888 --> 00:06:38,450
remaining strategically aligned.
This shift in thinking from
121
00:06:38,489 --> 00:06:41,513
projects to behaviors, from
activities to competitive
122
00:06:41,528 --> 00:06:43,122
advantage isn't merely
theoretical.
123
00:06:43,185 --> 00:06:45,872
It's the difference between data
science teams that consistently
124
00:06:45,903 --> 00:06:48,294
deliver value and those that
struggle to justify their
125
00:06:48,325 --> 00:06:51,390
existence when conditions
change. The power of this
126
00:06:51,430 --> 00:06:54,691
definition becomes especially
evident during organizational
127
00:06:54,714 --> 00:06:57,714
transitions. When leadership
changes or budgets tighten,
128
00:06:58,097 --> 00:07:01,058
teams without a true Strategy
find their work suddenly
129
00:07:01,121 --> 00:07:02,417
questioned or redirected.
130
00:07:03,199 --> 00:07:06,917
Teams with clear behavioral
commitments tied to competitive
131
00:07:06,980 --> 00:07:10,792
advantage can articulate their
value regardless of who's in
132
00:07:10,855 --> 00:07:13,636
charge. This doesn't mean you
should abandon planning
133
00:07:13,699 --> 00:07:16,980
altogether. Tactical plans
remain important for execution,
134
00:07:17,558 --> 00:07:21,375
but they should derive from...
Your strategic behavioral
135
00:07:21,455 --> 00:07:24,558
commitments rather than
constitute the Strategy itself.
136
00:07:24,757 --> 00:07:27,519
The projects you pursue should
be expressions of your strategic
137
00:07:27,679 --> 00:07:30,840
behaviors, not the Strategy
itself. So what is the one
138
00:07:30,925 --> 00:07:34,425
sentence definition of Strategy
that changes everything? It's
139
00:07:34,449 --> 00:07:37,918
that a Strategy is a commitment
to a pattern of behavior
140
00:07:37,996 --> 00:07:40,777
intended to win a competition,
not just a plan or roadmap.
141
00:07:41,386 --> 00:07:45,621
This definition liberates data
science teams from the tyranny
142
00:07:45,777 --> 00:07:49,269
of project lists that quickly
become outdated and focuses
143
00:07:49,329 --> 00:07:52,072
attention on the consistent
behaviors that create lasting
144
00:07:52,191 --> 00:07:56,418
competitive advantage. Plans
versus Strategy, the critical
145
00:07:56,496 --> 00:08:00,418
difference. If Strategy is about
behavioral commitments rather
146
00:08:00,457 --> 00:08:04,910
than project lists, why do 85%
of data science teams still
147
00:08:04,972 --> 00:08:05,785
confuse the two?
148
00:08:06,410 --> 00:08:08,644
The distinction isn't just
academic, it's the difference
149
00:08:08,676 --> 00:08:11,051
between teams that create
lasting value and those that
150
00:08:11,113 --> 00:08:14,332
constantly restart from zero.
When data science leaders
151
00:08:14,441 --> 00:08:17,379
proudly slide documents across
the table labeled our data
152
00:08:17,441 --> 00:08:20,762
Strategy, I often find myself
looking at elaborate
153
00:08:20,783 --> 00:08:23,523
spreadsheets listing
technologies to implement,
154
00:08:23,766 --> 00:08:27,930
dashboards to build, or models
to deploy over the next 12 to 18
155
00:08:27,992 --> 00:08:28,227
months.
156
00:08:29,211 --> 00:08:32,516
These documents typically follow
a predictable pattern. They
157
00:08:32,555 --> 00:08:35,016
begin with current state
assessment, move to gap
158
00:08:35,078 --> 00:08:38,266
analysis, then outline a series
of projects with timelines,
159
00:08:38,641 --> 00:08:41,922
resource requirements, and
sometimes even ROI calculations.
160
00:08:42,859 --> 00:08:45,734
At first glance, these documents
appear comprehensive. They
161
00:08:45,750 --> 00:08:49,243
contain Gantt charts. Showing
implementation phases for data
162
00:08:49,302 --> 00:08:52,507
lakes, machine learning
platforms, or analytics tools,
163
00:08:52,706 --> 00:08:55,749
they outline hiring plans for
specific roles. Data engineers,
164
00:08:55,953 --> 00:08:59,734
ML engineers, business analysts,
they specify which algorithms
165
00:08:59,890 --> 00:09:03,273
will be deployed and which
business units will receive
166
00:09:03,374 --> 00:09:04,999
analytics capabilities first.
167
00:09:05,624 --> 00:09:08,234
They're often visually
impressive, with professional
168
00:09:08,281 --> 00:09:11,281
graphics and executive summaries
designed for leadership
169
00:09:11,343 --> 00:09:14,640
consumption. The problem isn't
that these documents lack detail
170
00:09:14,687 --> 00:09:17,583
or thought. The problem is that
they answer an entirely
171
00:09:17,622 --> 00:09:20,206
different question than what
Strategy should address. They
172
00:09:20,245 --> 00:09:23,487
explain what the team plans to
do, but not how they'll behave
173
00:09:23,530 --> 00:09:25,890
differently to outperform
competitors or create
174
00:09:25,913 --> 00:09:27,054
sustainable advantage.
175
00:09:27,929 --> 00:09:29,991
Let me illustrate this
distinction with a concrete
176
00:09:30,015 --> 00:09:30,452
example.
177
00:09:31,319 --> 00:09:33,976
Imagine two retail companies
facing the same challenge,
178
00:09:34,741 --> 00:09:37,726
improving customer retention
through data science. The first
179
00:09:37,757 --> 00:09:41,523
company develops what they call
a Strategy that lists projects.
180
00:09:42,382 --> 00:09:44,194
Implement a customer data
platform in Q1.
181
00:09:45,537 --> 00:09:49,785
Build churn prediction models in
Q2, develop recommendation
182
00:09:49,843 --> 00:09:53,691
engines in Q3, and deploy
personalized marketing campaigns
183
00:09:53,769 --> 00:09:54,074
in Q4.
184
00:09:55,371 --> 00:09:57,902
The second company takes a
different approach. Their
185
00:09:57,941 --> 00:10:02,512
Strategy states, We will win by
rapidly testing small targeted
186
00:10:02,573 --> 00:10:05,335
interventions based on simple
customer behavior patterns
187
00:10:05,674 --> 00:10:07,479
rather than building
comprehensive models.
188
00:10:08,081 --> 00:10:11,182
We commit to running at least
five small experiments per
189
00:10:11,284 --> 00:10:15,901
month, making decisions within
48 hours of getting results, and
190
00:10:15,979 --> 00:10:18,729
scaling successful approaches
immediately, even with
191
00:10:18,854 --> 00:10:22,854
incomplete information. When
market conditions change, say a
192
00:10:22,885 --> 00:10:26,213
major competitor launches an
innovative loyalty program, the
193
00:10:26,229 --> 00:10:29,399
responses reveal the critical
difference. The first company's
194
00:10:29,459 --> 00:10:30,959
detailed plan becomes obsolete.
195
00:10:31,641 --> 00:10:35,045
They must restart their planning
cycle, re-evaluate priorities,
196
00:10:35,506 --> 00:10:38,986
and develop a new project
sequence. Meanwhile, the second
197
00:10:39,010 --> 00:10:41,947
company simply applies their
established behavioral pattern
198
00:10:42,010 --> 00:10:45,330
to the new situation, quickly
testing counter-moves and
199
00:10:45,580 --> 00:10:48,799
scaling what works. This
distinction becomes even more
200
00:10:48,815 --> 00:10:50,611
critical as data science teams
grow.
201
00:10:51,877 --> 00:10:54,518
A startup with three data
scientists might survive with
202
00:10:54,549 --> 00:10:57,633
just a plan because Informal
communication and rapid pivoting
203
00:10:57,672 --> 00:11:01,156
happen naturally. But as teams
grow to 10, 30 or 100 people
204
00:11:01,176 --> 00:11:04,137
across multiple business units,
the lack of consistent
205
00:11:04,176 --> 00:11:06,543
behavioral commitments leads to
fragmentation.
206
00:11:07,043 --> 00:11:09,504
Without clear strategic
behaviors, different team
207
00:11:09,543 --> 00:11:12,168
members make conflicting
decisions, pursue contradictory
208
00:11:12,199 --> 00:11:16,590
approaches and optimize for
incompatible outcomes. In
209
00:11:16,637 --> 00:11:19,340
enterprises, I've witnessed
teams with impressive technical
210
00:11:19,387 --> 00:11:22,637
talent who produced minimal
business impact because they
211
00:11:22,699 --> 00:11:24,980
lacked strategic behavioral
commitments.
212
00:11:25,458 --> 00:11:28,699
Their Strategy documents
outlined ambitious roadmaps for
213
00:11:28,738 --> 00:11:33,896
advanced AI capabilities, but
said nothing about how decisions
214
00:11:33,919 --> 00:11:37,982
would be made, which problems
would take priority, or what
215
00:11:38,177 --> 00:11:39,677
trade-offs would guide their
work.
216
00:11:40,380 --> 00:11:42,786
The result was a collection of
technically impressive but
217
00:11:42,818 --> 00:11:46,146
disconnected initiatives that
failed to build on each other or
218
00:11:46,161 --> 00:11:49,208
create lasting competitive
advantage. What we're witnessing
219
00:11:49,224 --> 00:11:52,866
is the fundamental nature of
plans versus strategies. Plans
220
00:11:52,946 --> 00:11:56,391
specify tasks, timelines, and
deliverables. They answer what
221
00:11:56,609 --> 00:11:57,652
and when questions.
222
00:11:58,051 --> 00:12:00,355
They are necessary operational
documents, but they don't
223
00:12:00,375 --> 00:12:02,676
provide guidance for how to
respond to changing
224
00:12:02,738 --> 00:12:05,480
circumstances or how to make
difficult trade-offs. They
225
00:12:05,519 --> 00:12:08,504
become outdated the moment
market conditions shift.
226
00:12:09,223 --> 00:12:13,082
Strategies, by contrast, specify
behavioral patterns and decision
227
00:12:13,144 --> 00:12:15,223
criteria. They answer how
questions.
228
00:12:15,644 --> 00:12:19,238
How will we make decisions? How
will we allocate resources? How
229
00:12:19,254 --> 00:12:21,207
will we approach problems
differently than our
230
00:12:21,269 --> 00:12:24,083
competitors? These behavioral
commitments provide consistent
231
00:12:24,124 --> 00:12:27,329
guidance, even as specific
projects evolve. Another
232
00:12:27,368 --> 00:12:29,950
critical difference emerges in
how success is measured.
233
00:12:30,907 --> 00:12:34,376
Plans are evaluated based on
completion metrics. Did we
234
00:12:34,516 --> 00:12:37,540
finish the projects on time and
within budget? Strategies are
235
00:12:37,555 --> 00:12:41,305
evaluated based on competitive
outcomes. Did our behavioral
236
00:12:41,368 --> 00:12:43,696
patterns create advantage
compared to alternative
237
00:12:43,727 --> 00:12:46,680
approaches? This distinction
forces teams to connect their
238
00:12:46,727 --> 00:12:49,212
work directly to competitive
position rather than just
239
00:12:49,274 --> 00:12:50,930
checking boxes on a project
list.
240
00:12:51,490 --> 00:12:54,291
The confusion between plans and
strategies creates particular
241
00:12:54,330 --> 00:12:56,852
problems for data science
because of the field's inherent
242
00:12:56,930 --> 00:13:00,531
uncertainty. In established
functions like finance or
243
00:13:00,532 --> 00:13:04,133
operations, detailed plans often
work because the environment is
244
00:13:04,172 --> 00:13:07,633
relatively stable and processes
are well understood.
245
00:13:08,727 --> 00:13:12,039
But data science involves
exploration, experimentation and
246
00:13:12,055 --> 00:13:15,930
discovery. Detailed project
plans for such work invariably
247
00:13:15,977 --> 00:13:18,818
become fiction shortly after
they are created. This reality
248
00:13:18,858 --> 00:13:22,081
doesn't mean data science teams
should abandon planning. Rather,
249
00:13:22,183 --> 00:13:25,124
it means their strategies must
establish behavioral patterns
250
00:13:25,526 --> 00:13:27,710
specifically designed to
navigate uncertainty
251
00:13:27,765 --> 00:13:28,249
effectively.
252
00:13:28,765 --> 00:13:31,647
These might include commitments
to specific experimentation
253
00:13:31,694 --> 00:13:35,030
cadences, explicit resource
allocation between exploration
254
00:13:35,077 --> 00:13:38,155
and exploitation, or decision
processes for rapidly
255
00:13:38,233 --> 00:13:41,702
terminating unproductive
avenues. One multinational
256
00:13:41,733 --> 00:13:44,530
consumer goods company I worked
with transformed their approach
257
00:13:44,546 --> 00:13:47,016
when they shifted from a
detailed AI implementation
258
00:13:47,056 --> 00:13:50,258
roadmap to a Strategy focused on
rapid learning cycles.
259
00:13:50,961 --> 00:13:54,359
Their new Strategy committed to
specific behaviors. All
260
00:13:54,406 --> 00:13:57,242
initiatives would start with
minimum viable data products
261
00:13:57,289 --> 00:14:01,328
delivered in under four weeks.
Business stakeholders would be
262
00:14:01,375 --> 00:14:04,609
embedded in technical teams
rather than just receiving
263
00:14:04,687 --> 00:14:05,156
updates.
264
00:14:06,047 --> 00:14:10,500
And technical complexity would
be deliberately limited until
265
00:14:10,781 --> 00:14:13,687
business value was validated.
These behavioral commitments
266
00:14:13,734 --> 00:14:17,259
remained constant. Even as
specific projects evolved
267
00:14:17,419 --> 00:14:20,925
dramatically based on market
feedback. The root of this
268
00:14:21,003 --> 00:14:25,425
confusion often lies in how
requests for Strategy are framed
269
00:14:25,488 --> 00:14:26,425
within organizations.
270
00:14:26,550 --> 00:14:29,831
When an executive asks a data
science leader to develop our
271
00:14:29,894 --> 00:14:33,456
data Strategy, both parties
frequently default to thinking
272
00:14:33,488 --> 00:14:36,331
about roadmaps and
implementation plans because
273
00:14:36,378 --> 00:14:38,660
these are concrete and familiar
formats.
274
00:14:39,503 --> 00:14:43,441
Breaking this pattern requires
explicitly reframing the request
275
00:14:43,972 --> 00:14:46,872
around behavioral commitments
and competitive approaches
276
00:14:47,453 --> 00:14:51,177
rather than project sequences.
As teams grow in size and
277
00:14:51,197 --> 00:14:54,122
complexity, the distinction
between plans and strategies
278
00:14:54,162 --> 00:14:55,263
becomes even more crucial.
279
00:14:55,841 --> 00:14:58,966
Small teams might succeed
through the intuitive behaviors
280
00:14:58,982 --> 00:15:01,490
of a few talented individuals,
but larger teams require
281
00:15:01,568 --> 00:15:04,474
explicit behavioral commitments
to maintain coherence. Without
282
00:15:04,537 --> 00:15:07,755
them, different team members
optimize for incompatible
283
00:15:07,849 --> 00:15:12,824
outcomes, technical decisions,
lack consistency, and the
284
00:15:12,865 --> 00:15:16,129
cumulative effect of the team's
work diminishes rather than
285
00:15:16,187 --> 00:15:16,750
compounds.
286
00:15:18,750 --> 00:15:22,031
The three requirements of
effective Strategy. Beyond mere
287
00:15:22,054 --> 00:15:26,656
behavioral patterns lies a more
fundamental challenge, crafting
288
00:15:26,657 --> 00:15:30,906
a Strategy that actually works
when implemented. What makes the
289
00:15:30,937 --> 00:15:33,312
difference between data science
strategies that drive
290
00:15:33,359 --> 00:15:35,812
organizational transformation
versus those that become
291
00:15:35,906 --> 00:15:39,966
irrelevant within months. Three
specific requirements separate
292
00:15:40,005 --> 00:15:41,446
success from failure.
293
00:15:41,868 --> 00:15:44,829
In all my years consulting with
data science teams, I've
294
00:15:44,891 --> 00:15:47,876
witnessed a recurring pattern. A
brilliant team develops what
295
00:15:47,891 --> 00:15:50,594
appears to be a compelling
Strategy. They've correctly
296
00:15:50,938 --> 00:15:54,180
identified that Strategy is
about behavioral commitments
297
00:15:54,321 --> 00:15:57,415
rather than project lists. Their
document beautifully articulates
298
00:15:57,462 --> 00:15:59,540
how they'll behave to create
competitive advantage.
299
00:16:00,071 --> 00:16:03,774
Everyone nods enthusiastically
at the kick-off meeting. Yet six
300
00:16:03,852 --> 00:16:07,768
months later, the Strategy has
been abandoned. Or Worse is
301
00:16:07,808 --> 00:16:10,310
being followed religiously
despite producing minimal
302
00:16:10,349 --> 00:16:15,154
results. Gary Pisano's research
on R&D Strategy goes beyond
303
00:16:15,279 --> 00:16:17,755
simply defining Strategy as
behavioral commitments.
304
00:16:18,240 --> 00:16:21,044
He identifies three critical
requirements that separate
305
00:16:21,122 --> 00:16:23,966
theoretical strategies from
those that actually transform
306
00:16:24,029 --> 00:16:28,513
organizations. Consistency,
coherence, and alignment. When
307
00:16:28,810 --> 00:16:32,232
data science strategies fail,
despite correct intentions, it's
308
00:16:32,294 --> 00:16:35,357
typically because they've missed
one or more of these fundamental
309
00:16:35,416 --> 00:16:38,658
requirements. Let's examine each
requirement and understand why
310
00:16:38,720 --> 00:16:41,283
they matter specifically for
data science teams.
311
00:16:41,806 --> 00:16:44,728
Consistency refers to making
decisions that contribute to the
312
00:16:44,767 --> 00:16:48,025
same competitive advantage over
time. This doesn't mean rigidly
313
00:16:48,087 --> 00:16:51,306
sticking to the same tactics
regardless of circumstances, but
314
00:16:51,431 --> 00:16:54,931
rather ensuring that the
underlying behavioral patterns
315
00:16:55,290 --> 00:16:58,994
reinforce the same strategic
goals even as implementation
316
00:16:59,056 --> 00:17:00,056
details evolve.
317
00:17:00,959 --> 00:17:04,219
Consider a financial technology
startup, I advised, whose
318
00:17:04,420 --> 00:17:07,398
Strategy emphasized rapid
experimentation to discover
319
00:17:07,461 --> 00:17:11,379
customer needs. Their behavioral
commitment was to conduct at
320
00:17:11,465 --> 00:17:13,324
least two customer experiments
weekly.
321
00:17:13,824 --> 00:17:16,363
Initially, this worked
brilliantly. But when they
322
00:17:16,402 --> 00:17:19,098
received a large investment, the
leadership suddenly shifted
323
00:17:19,160 --> 00:17:21,707
focus to building
enterprise-grade infrastructure
324
00:17:21,723 --> 00:17:24,551
that would take months to
implement with no customer
325
00:17:24,629 --> 00:17:25,051
feedback.
326
00:17:26,066 --> 00:17:28,379
While they claimed their
Strategy remained unchanged,
327
00:17:28,547 --> 00:17:31,190
their actual decisions no longer
reflected their stated
328
00:17:31,229 --> 00:17:34,533
behavioral commitments. Within
three months, they had built
329
00:17:34,693 --> 00:17:37,732
sophisticated systems that
customers didn't want, forcing
330
00:17:37,756 --> 00:17:40,279
them to start over with
substantial sunk costs. The
331
00:17:40,295 --> 00:17:40,561
lesson?
332
00:17:42,076 --> 00:17:44,498
Consistency doesn't mean
inflexibility, but it does
333
00:17:44,545 --> 00:17:48,451
require discipline to maintain
core behavioral commitments,
334
00:17:48,607 --> 00:17:50,561
even when tempting alternatives
arise.
335
00:17:51,686 --> 00:17:54,717
For data science teams,
consistency is particularly
336
00:17:54,779 --> 00:17:57,170
challenging because
opportunities constantly emerge
337
00:17:57,217 --> 00:18:00,956
to apply. New technologies or
methodologies that may distract
338
00:18:01,296 --> 00:18:04,735
from strategic focus. The second
requirement, coherence, means
339
00:18:04,775 --> 00:18:07,696
ensuring all parts of your
Strategy support rather than
340
00:18:07,775 --> 00:18:08,657
contradict each other.
341
00:18:09,618 --> 00:18:13,157
Many data science strategies
include behavioral commitments
342
00:18:13,181 --> 00:18:15,720
that sound reasonable
individually but create friction
343
00:18:15,751 --> 00:18:19,501
when combined. A pharmaceutical
research team I worked with
344
00:18:19,892 --> 00:18:23,657
committed both to maintaining
gold standard data quality and
345
00:18:23,860 --> 00:18:27,092
rapidly prototyping analytics
models. In practice, these
346
00:18:27,112 --> 00:18:31,097
commitments created constant
conflict as the rigorous data
347
00:18:31,156 --> 00:18:34,339
preparation processes made rapid
prototyping impossible.
348
00:18:35,019 --> 00:18:39,425
Team members split into camps
with data engineers prioritizing
349
00:18:39,464 --> 00:18:42,488
quality and data scientists
prioritizing speed. The
350
00:18:42,489 --> 00:18:45,566
resulting tension paralyzed
decision making. Creating
351
00:18:45,597 --> 00:18:49,175
coherence means critically
examining how your behavioral
352
00:18:49,238 --> 00:18:50,675
commitments interact with each
other.
353
00:18:50,910 --> 00:18:54,300
Do they create synergies or do
they pull your team in different
354
00:18:54,316 --> 00:18:58,062
directions? This requires
honestly assessing trade-offs
355
00:18:58,501 --> 00:19:02,107
rather than pretending
contradictions don't exist. The
356
00:19:02,126 --> 00:19:05,869
pharmaceutical team eventually
revised their Strategy to focus
357
00:19:06,033 --> 00:19:10,673
on fit-for-purpose data quality,
maintaining rigorous standards
358
00:19:10,674 --> 00:19:14,642
for production systems while
accepting approximations for
359
00:19:14,736 --> 00:19:16,236
early-stage experimentation.
360
00:19:17,626 --> 00:19:21,892
This single adjustment created
coherence that allowed both
361
00:19:22,064 --> 00:19:25,567
rapid innovation and reliable
production systems The third
362
00:19:25,607 --> 00:19:29,167
requirement, alignment, connects
your Strategy to broader
363
00:19:29,268 --> 00:19:31,468
organizational goals and
competitive context.
364
00:19:32,249 --> 00:19:35,565
Even the most consistent and
coherent data science Strategy
365
00:19:35,972 --> 00:19:39,167
fails if it doesn't help the
organization win in its specific
366
00:19:39,253 --> 00:19:42,464
competitive environment. A
retail analytics team I worked
367
00:19:42,495 --> 00:19:45,573
with had developed a beautifully
coherent Strategy focused on
368
00:19:45,932 --> 00:19:49,526
sophisticated customer lifetime
value modeling. Their behavioral
369
00:19:49,573 --> 00:19:52,167
commitments all supported this
focus consistently.
370
00:19:53,131 --> 00:19:56,178
The problem? Their organization
competed primarily on supply
371
00:19:56,258 --> 00:19:59,547
chain efficiency, not customer
intimacy. While their Strategy
372
00:19:59,606 --> 00:20:02,925
was internally sound, It lacked
alignment with organizational
373
00:20:02,986 --> 00:20:06,269
priorities. Creating alignment
requires understanding how your
374
00:20:06,328 --> 00:20:09,011
team's distinctive capabilities
contribute to your
375
00:20:09,070 --> 00:20:10,875
organization's competitive
position.
376
00:20:11,476 --> 00:20:14,359
This means regularly reassessing
whether your behavioral
377
00:20:14,414 --> 00:20:17,508
commitments still serve the
organization's evolving needs.
378
00:20:18,023 --> 00:20:21,179
The retail analytics team
ultimately revised their
379
00:20:21,242 --> 00:20:23,992
Strategy to focus on inventory
optimization and demand
380
00:20:24,039 --> 00:20:27,023
forecasting, creating immediate
value recognition from
381
00:20:27,101 --> 00:20:27,523
leadership.
382
00:20:28,516 --> 00:20:31,560
These three requirements become
exponentially more difficult to
383
00:20:31,579 --> 00:20:34,622
maintain as data science teams
scale. With a team of three,
384
00:20:35,384 --> 00:20:37,849
consistency often emerges
naturally through daily
385
00:20:37,927 --> 00:20:40,825
interactions. With 30 team
members spread across different
386
00:20:40,888 --> 00:20:44,411
projects and locations,
maintaining consistency requires
387
00:20:45,099 --> 00:20:47,552
deliberate communication and
governance structures.
388
00:20:48,474 --> 00:20:52,224
A global insurance company I
advised struggled with this
389
00:20:52,536 --> 00:20:55,880
exact challenge. Their data
science Strategy appeared sound
390
00:20:55,881 --> 00:20:59,502
on paper. But regional teams
interpreted and implemented
391
00:20:59,624 --> 00:21:02,206
behavioral commitments
differently based on local
392
00:21:02,245 --> 00:21:05,405
conditions. The result was
fragmentation rather than
393
00:21:05,491 --> 00:21:06,311
collective impact.
394
00:21:07,186 --> 00:21:10,592
Similarly, coherence becomes
more challenging as teams grow
395
00:21:10,655 --> 00:21:13,749
because different subgroups
naturally optimize for different
396
00:21:13,811 --> 00:21:16,358
outcomes based on their specific
responsibilities.
397
00:21:17,592 --> 00:21:21,186
Alignment becomes more complex
as teams must interface with
398
00:21:21,389 --> 00:21:24,780
more stakeholders holding
divergent priorities. The
399
00:21:24,800 --> 00:21:27,163
structured governance that helps
larger teams maintain
400
00:21:27,221 --> 00:21:30,167
consistency can sometimes hinder
the flexibility needed for
401
00:21:30,225 --> 00:21:32,487
alignment with changing
organizational needs.
402
00:21:33,409 --> 00:21:37,276
When data science strategies
fail, despite being focused on
403
00:21:37,315 --> 00:21:40,518
behavioral commitments rather
than project lists, the root
404
00:21:40,596 --> 00:21:43,495
cause typically traces back to
missing one of these three
405
00:21:43,542 --> 00:21:44,135
requirements.
406
00:21:45,073 --> 00:21:48,057
A technology company I worked
with committed to democratizing
407
00:21:48,089 --> 00:21:50,339
data access by building
self-service tools for all
408
00:21:50,385 --> 00:21:53,260
departments. This behavioral
commitment was clear and
409
00:21:53,323 --> 00:21:56,266
consistently followed. But
lacked coherence with their
410
00:21:56,346 --> 00:21:59,567
other commitment to protecting
sensitive customer information.
411
00:22:00,310 --> 00:22:04,251
The result was constant security
exceptions, duplicated efforts
412
00:22:04,290 --> 00:22:07,399
to build department-specific
solutions with varying controls,
413
00:22:07,993 --> 00:22:11,899
and frustrated stakeholders.
Another common failure pattern
414
00:22:12,087 --> 00:22:14,884
is when teams maintain
consistency and coherence but
415
00:22:14,978 --> 00:22:18,446
lose alignment over time. An
energy sector analytics team
416
00:22:18,868 --> 00:22:22,149
followed their Strategy of
building advanced predictive
417
00:22:22,212 --> 00:22:24,469
maintenance models. With
impressive discipline.
418
00:22:24,910 --> 00:22:27,672
Their behavioral commitments
were consistent and coherent.
419
00:22:27,871 --> 00:22:30,797
However, when market pressures
shifted their company's focus to
420
00:22:30,855 --> 00:22:33,801
renewable energy initiatives,
they continued optimizing
421
00:22:33,840 --> 00:22:36,777
traditional operations while the
organization pivoted around
422
00:22:36,824 --> 00:22:40,887
them. Despite executing their
Strategy flawlessly, they became
423
00:22:40,949 --> 00:22:43,090
increasingly irrelevant to
business outcomes.
424
00:22:43,605 --> 00:22:45,996
Testing your data science
Strategy against these three
425
00:22:46,027 --> 00:22:49,105
requirements provides a powerful
way to identify potential
426
00:22:49,168 --> 00:22:53,552
failure points before they
manifest. Ask yourself. Will
427
00:22:53,612 --> 00:22:56,594
these behavioral commitments
still apply when conditions
428
00:22:56,655 --> 00:23:00,460
change? Consistency. Do these
commitments support or
429
00:23:00,499 --> 00:23:02,378
contradict each other?
Coherence.
430
00:23:03,284 --> 00:23:05,807
And do these commitments help
the organization win in its
431
00:23:05,846 --> 00:23:07,565
competitive context? Alignment.
432
00:23:08,448 --> 00:23:12,089
Why data science teams struggle
with Strategy? Testing your data
433
00:23:12,151 --> 00:23:15,339
science Strategy against these
requirements reveals an
434
00:23:15,511 --> 00:23:18,417
uncomfortable truth. Data
science teams operate in
435
00:23:18,479 --> 00:23:21,354
organizational territory where
consistency see.
436
00:23:21,456 --> 00:23:24,579
Coherence and alignment are
uniquely difficult to maintain.
437
00:23:25,020 --> 00:23:28,544
What if the very nature of data
science work itself, not poor
438
00:23:28,602 --> 00:23:32,227
leadership or unclear vision, is
what makes strategic success so
439
00:23:32,345 --> 00:23:36,290
elusive? Data science occupies a
peculiar position in most
440
00:23:36,352 --> 00:23:37,048
organizations.
441
00:23:38,157 --> 00:23:42,235
Unlike marketing, sales, or even
traditional IT, data teams sit
442
00:23:42,579 --> 00:23:45,532
at a complex intersection.
They're highly technical
443
00:23:45,595 --> 00:23:47,782
specialists who must
simultaneously translate
444
00:23:47,876 --> 00:23:51,805
abstract concepts into business
value. This hybrid identity
445
00:23:51,885 --> 00:23:55,709
creates an organizational limbo
where data teams are expected to
446
00:23:55,710 --> 00:23:59,635
be both innovative researchers
and reliable service providers,
447
00:24:00,213 --> 00:24:03,018
often to multiple stakeholders
with competing priorities.
448
00:24:03,994 --> 00:24:07,198
This positioning creates
strategic quicksand that few
449
00:24:07,307 --> 00:24:10,182
leaders are prepared to
navigate. Consider how data
450
00:24:10,229 --> 00:24:13,979
science must bridge multiple
worlds. They need deep technical
451
00:24:14,026 --> 00:24:16,963
expertise to implement
solutions, business domain
452
00:24:17,026 --> 00:24:22,151
knowledge to identify valuable
problems, and political savvy to
453
00:24:22,190 --> 00:24:23,991
navigate organizational
priorities.
454
00:24:24,893 --> 00:24:27,534
Each of these domains operates
on different timescales and
455
00:24:27,557 --> 00:24:29,534
success metrics, making
consistent behavioral
456
00:24:29,581 --> 00:24:32,721
commitments extraordinarily
difficult. The technological
457
00:24:32,784 --> 00:24:36,088
foundation of data science
brings its own strategic
458
00:24:36,151 --> 00:24:36,917
complications.
459
00:24:37,120 --> 00:24:40,432
While marketing strategies might
remain stable for years, data
460
00:24:40,495 --> 00:24:43,292
science teams face a landscape
where foundational tools and
461
00:24:43,307 --> 00:24:46,432
techniques evolve quarterly. A
commitment to specific
462
00:24:46,495 --> 00:24:49,580
technologies, whether database
architectures, programming
463
00:24:49,640 --> 00:24:52,842
frameworks or model
architectures risks becoming
464
00:24:53,003 --> 00:24:56,487
obsolete before the Strategy can
even be fully implemented.
465
00:24:56,866 --> 00:24:59,854
What other department faces a
situation where their primary
466
00:24:59,932 --> 00:25:03,010
tools might be rendered outdated
by a single research paper or
467
00:25:03,112 --> 00:25:06,095
open source release? Take the
experience of a mid-size
468
00:25:06,135 --> 00:25:09,520
insurance company that committed
to a Strategy centered around
469
00:25:09,559 --> 00:25:11,141
traditional machine learning
models.
470
00:25:11,641 --> 00:25:14,621
Six months into implementation,
advances in transformer-based
471
00:25:14,707 --> 00:25:17,223
architectures made their
approach significantly less
472
00:25:17,270 --> 00:25:20,676
competitive. Their commitment to
a specific technical approach,
473
00:25:21,082 --> 00:25:24,395
rather than to the behavioral
pattern of continuously
474
00:25:24,473 --> 00:25:28,535
evaluating emerging methods,
left them scrambling to catch up
475
00:25:28,816 --> 00:25:30,441
while competitors surged ahead.
476
00:25:31,457 --> 00:25:34,575
Beyond the technical flux, Data
science teams face unique
477
00:25:34,615 --> 00:25:37,900
challenges in maintaining
coherence across their strategic
478
00:25:37,939 --> 00:25:40,380
commitments. Their
cross-functional nature means
479
00:25:40,482 --> 00:25:42,786
they often serve as the
connective tissue between
480
00:25:42,825 --> 00:25:45,427
departments with fundamentally
different perspectives and
481
00:25:45,443 --> 00:25:46,005
incentives.
482
00:25:46,990 --> 00:25:49,630
Product teams want quick
experiments to validate
483
00:25:49,677 --> 00:25:52,974
features, finance needs,
reliable forecasting models,
484
00:25:53,693 --> 00:25:57,661
operations, demands, efficient
optimization, algorithms, and
485
00:25:57,755 --> 00:26:01,378
executives request comprehensive
dashboards. All from the same
486
00:26:01,439 --> 00:26:04,062
team working with the same
finite resources.
487
00:26:05,222 --> 00:26:07,644
A healthcare analytics startup
illustrated this challenge
488
00:26:07,706 --> 00:26:11,066
perfectly when they attempted to
simultaneously commit to rapid
489
00:26:11,128 --> 00:26:14,831
experimentation for clinical
products while also promising
490
00:26:15,034 --> 00:26:18,042
enterprise-grade reliability for
their patient management
491
00:26:18,074 --> 00:26:18,449
systems.
492
00:26:19,417 --> 00:26:22,152
These behavioral commitments
directly contradicted each other
493
00:26:22,199 --> 00:26:25,324
in practice, creating impossible
situations where team members
494
00:26:25,386 --> 00:26:28,730
couldn't satisfy both strategic
directions simultaneously.
495
00:26:29,805 --> 00:26:32,128
When pushed to prioritize,
leadership would often give
496
00:26:32,167 --> 00:26:35,771
conflicting guidance, depending
on which stakeholder had most
497
00:26:35,829 --> 00:26:39,017
recently raised concerns. The
alignment challenge for data
498
00:26:39,072 --> 00:26:42,337
science stems from its
relatively novel position in
499
00:26:42,517 --> 00:26:43,822
organizational hierarchies.
500
00:26:44,556 --> 00:26:47,556
While established departments
like finance or sales have
501
00:26:47,618 --> 00:26:51,165
clearly defined relationships to
business outcomes, data
502
00:26:51,228 --> 00:26:53,931
science's contribution can be
more difficult to quantify and
503
00:26:53,947 --> 00:26:57,926
communicate. This creates an
alignment gap where Data teams
504
00:26:58,026 --> 00:27:00,549
might technically deliver on
their behavioral commitments
505
00:27:01,009 --> 00:27:03,990
while still failing to connect
those behaviors to competitive
506
00:27:04,033 --> 00:27:04,513
advantage.
507
00:27:05,357 --> 00:27:08,677
A telecommunications provider
exemplified this struggle when
508
00:27:08,756 --> 00:27:11,701
their data science team
perfectly executed on their
509
00:27:11,841 --> 00:27:15,123
strategic commitment to develop
sophisticated customer
510
00:27:15,185 --> 00:27:16,170
segmentation models.
511
00:27:17,201 --> 00:27:20,248
The models were technically
impressive, statistically sound,
512
00:27:20,591 --> 00:27:23,123
and delivered on schedule.
However, because the team hadn't
513
00:27:23,170 --> 00:27:25,670
aligned their Strategy with the
organization's primary
514
00:27:25,732 --> 00:27:29,319
competitive challenge, reducing
network infrastructure costs.
515
00:27:30,120 --> 00:27:33,665
The sophisticated segmentation
work generated minimal business
516
00:27:33,723 --> 00:27:35,665
value despite being a technical
success.
517
00:27:36,587 --> 00:27:40,368
For smaller data science teams,
these strategic challenges often
518
00:27:40,454 --> 00:27:44,040
manifest as whiplash-inducing
priority shifts. A three-person
519
00:27:44,071 --> 00:27:47,430
data team at a logistics startup
found themselves ricocheting
520
00:27:47,477 --> 00:27:50,352
between commitments to build
predictive maintenance
521
00:27:50,384 --> 00:27:54,821
capabilities and enhance
customer-facing analytics.
522
00:27:55,505 --> 00:27:59,390
Because different executives had
different visions of how data
523
00:27:59,429 --> 00:28:00,769
science should create value.
524
00:28:02,191 --> 00:28:05,394
Without a consistent strategic
direction that the organization
525
00:28:05,457 --> 00:28:08,800
had collectively endorsed, the
team's behavior became reactive
526
00:28:08,980 --> 00:28:12,394
rather than strategic. At the
enterprise scale, these
527
00:28:12,457 --> 00:28:15,378
challenges can create strategic
fragmentation.
528
00:28:15,832 --> 00:28:19,988
A global manufacturing company
with data scientists embedded
529
00:28:20,050 --> 00:28:23,394
across multiple business units
discovered that each team had
530
00:28:23,425 --> 00:28:25,479
developed different and
sometimes contradictory
531
00:28:25,539 --> 00:28:29,162
behavioral commitments. Some
prioritized algorithmic
532
00:28:29,244 --> 00:28:34,268
sophistication, others focused
on stakeholder partnership, and
533
00:28:34,330 --> 00:28:36,534
still others emphasized
production engineering
534
00:28:36,588 --> 00:28:37,073
excellence.
535
00:28:38,174 --> 00:28:40,721
When these teams needed to
collaborate on cross-functional
536
00:28:40,752 --> 00:28:44,330
initiatives, these different
behavioral patterns created
537
00:28:44,565 --> 00:28:48,065
significant friction and
inefficiency. The experimental
538
00:28:48,143 --> 00:28:51,784
nature of data science work
amplifies these strategic
539
00:28:51,846 --> 00:28:56,510
difficulties. Unlike departments
with predictable workflows and
540
00:28:56,651 --> 00:29:00,491
outcomes, data teams engage in
work with inherently uncertain
541
00:29:00,573 --> 00:29:01,034
results.
542
00:29:01,534 --> 00:29:05,690
Models may not achieve expected
accuracy, data quality issues
543
00:29:05,870 --> 00:29:10,127
might emerge midway through
projects, or analysis might
544
00:29:10,174 --> 00:29:14,237
reveal that initial assumptions
were incorrect. This uncertainty
545
00:29:14,299 --> 00:29:16,877
makes it challenging to
establish consistent behavioral
546
00:29:16,940 --> 00:29:20,721
patterns when the work itself
requires adaptation and course
547
00:29:20,737 --> 00:29:21,143
correction.
548
00:29:22,449 --> 00:29:25,932
A digital native retail company
committed to a behavioral
549
00:29:26,014 --> 00:29:30,158
pattern of delivering model
updates weekly, but found this
550
00:29:30,377 --> 00:29:33,236
commitment impossible to
maintain because some models
551
00:29:33,299 --> 00:29:36,502
required extensive retraining
and validation before
552
00:29:36,526 --> 00:29:37,104
deployment.
553
00:29:37,721 --> 00:29:40,526
Their technical reality
conflicted with their strategic
554
00:29:40,588 --> 00:29:43,791
commitment, creating tension
between quality and consistency,
555
00:29:44,432 --> 00:29:47,557
only when they refined their
Strategy to commit to regular,
556
00:29:47,697 --> 00:29:49,369
risk-appropriate update cycles.
557
00:29:49,745 --> 00:29:51,788
Could they establish a
sustainable pattern that
558
00:29:51,849 --> 00:29:54,052
accommodated the inherent
uncertainty in their work?
559
00:29:54,673 --> 00:29:58,478
Additionally, data science teams
face the challenge of balancing
560
00:29:58,501 --> 00:30:01,829
exploration, discovering new
insights and approaches, with
561
00:30:01,884 --> 00:30:05,063
exploitation. Delivering
reliable value from established
562
00:30:05,123 --> 00:30:05,502
methods.
563
00:30:06,924 --> 00:30:10,129
This dual mandate creates
strategic tension that few other
564
00:30:10,149 --> 00:30:13,293
departments face so acutely.
Marketing might test new
565
00:30:13,372 --> 00:30:17,793
channels, but data science must
often simultaneously explore
566
00:30:17,817 --> 00:30:21,239
entirely new methodologies while
delivering consistent results
567
00:30:21,254 --> 00:30:22,145
from existing ones.
568
00:30:23,098 --> 00:30:25,879
A financial services analytics
team tried to resolve this
569
00:30:25,942 --> 00:30:28,270
tension by assigning different
team members to either
570
00:30:28,426 --> 00:30:31,618
innovation or production work
only to discover that this
571
00:30:31,677 --> 00:30:34,582
created knowledge silos and
inconsistent outputs.
572
00:30:35,843 --> 00:30:39,504
Their revised Strategy committed
instead to a behavior of
573
00:30:39,949 --> 00:30:45,269
time-boxed exploration within
production cycles, allowing all
574
00:30:45,730 --> 00:30:49,363
team members to participate in
both modes while maintaining
575
00:30:49,394 --> 00:30:51,004
coherent delivery patterns.
576
00:30:51,863 --> 00:30:54,457
These strategic challenges
aren't signs of failure. They're
577
00:30:54,488 --> 00:30:57,222
inherent to the position data
science occupies in modern
578
00:30:57,269 --> 00:31:01,081
organizations. The teams that
succeed strategically Recognize
579
00:31:01,121 --> 00:31:04,224
these unique difficulties and
craft strategies that explicitly
580
00:31:04,265 --> 00:31:06,706
account for them rather than
struggling against them.
581
00:31:07,488 --> 00:31:11,831
Learning from R&D Strategic
Lessons for Data Science The
582
00:31:11,855 --> 00:31:14,894
strategic difficulties facing
data science aren't unique in
583
00:31:14,895 --> 00:31:17,613
the corporate landscape. They're
remarkably similar to challenges
584
00:31:17,659 --> 00:31:20,847
that R&D departments have been
solving since the 1950s.
585
00:31:21,784 --> 00:31:25,144
Yet most data leaders overlook
this valuable source of
586
00:31:25,191 --> 00:31:28,550
strategic wisdom, attempting to
build frameworks from scratch
587
00:31:28,988 --> 00:31:32,835
rather than adapt. Proven
approaches. When we examine the
588
00:31:32,976 --> 00:31:35,376
function of research and
development departments across
589
00:31:35,458 --> 00:31:38,243
industries, the parallels with
data science become striking.
590
00:31:38,923 --> 00:31:42,845
Both operate at the intersection
of discovery and practical
591
00:31:42,907 --> 00:31:47,712
application. Both face uncertain
outcomes and timelines. Both
592
00:31:47,775 --> 00:31:49,993
must balance deep technical
expertise with business
593
00:31:50,071 --> 00:31:52,603
relevance. And perhaps most
importantly, both must deliver
594
00:31:52,650 --> 00:31:55,400
innovation while maintaining
enough predictability to satisfy
595
00:31:55,556 --> 00:31:56,556
organizational demands.
596
00:31:57,196 --> 00:32:00,787
This similarity isn't merely
superficial. The strategic
597
00:32:00,867 --> 00:32:04,713
frameworks that have guided R&D
departments through decades of
598
00:32:04,771 --> 00:32:08,717
technological revolution offer
ready-made solutions to many of
599
00:32:08,718 --> 00:32:12,115
the unique challenges data
science teams face today.
600
00:32:12,756 --> 00:32:15,842
Corporate R&D departments
pioneered approaches to managing
601
00:32:15,904 --> 00:32:18,779
innovation portfolios long
before data science entered our
602
00:32:18,826 --> 00:32:22,810
vocabulary. Consider how mature
R&D organizations approach
603
00:32:22,967 --> 00:32:27,013
portfolio management. They don't
view projects as independent
604
00:32:27,076 --> 00:32:30,290
initiatives, but as a
deliberately balanced collection
605
00:32:30,768 --> 00:32:34,050
designed to distribute risk
while maximizing potential
606
00:32:34,089 --> 00:32:34,589
returns.
607
00:32:35,991 --> 00:32:39,593
A typical R&D portfolio might
allocate approximately 70% of
608
00:32:39,608 --> 00:32:42,632
resources toward incremental
improvements to existing
609
00:32:42,694 --> 00:32:47,132
products, 20% toward adjacent
innovations that extend into new
610
00:32:47,210 --> 00:32:51,132
markets or technologies, and 10%
toward transformational
611
00:32:51,272 --> 00:32:54,116
moonshots that could
fundamentally reshape the
612
00:32:54,132 --> 00:32:54,476
business.
613
00:32:56,109 --> 00:32:59,112
One pharmaceutical data science
organization that adopted this
614
00:32:59,273 --> 00:33:01,476
R&D-inspired approach
transformed their strategic
615
00:33:01,534 --> 00:33:02,276
approach completely.
616
00:33:02,874 --> 00:33:06,159
Rather than evaluating each
potential project in isolation,
617
00:33:06,722 --> 00:33:11,722
they classified initiatives as
incremental, improving existing
618
00:33:11,909 --> 00:33:16,816
analytical models, adjacent,
applying proven techniques to
619
00:33:16,862 --> 00:33:21,347
new business domains, or
transformational, exploring
620
00:33:21,409 --> 00:33:25,586
cutting-edge AI methodologies.
With uncertain but potentially
621
00:33:25,687 --> 00:33:26,648
massive returns.
622
00:33:27,951 --> 00:33:31,111
This portfolio view allowed them
to maintain a healthy innovation
623
00:33:31,193 --> 00:33:33,958
pipeline while still delivering
predictable short-term value,
624
00:33:34,318 --> 00:33:37,021
precisely the balance many data
science teams struggle to
625
00:33:37,060 --> 00:33:40,818
achieve. Another core R&D
strategic principle involves
626
00:33:40,896 --> 00:33:42,568
deliberately structuring time
horizons.
627
00:33:42,958 --> 00:33:45,833
Unlike many corporate functions
that operate primarily on
628
00:33:45,943 --> 00:33:50,177
quarterly cycles, sophisticated
R&D organizations simultaneously
629
00:33:50,224 --> 00:33:53,521
manage initiatives across
multiple time horizons, from
630
00:33:53,725 --> 00:33:56,787
immediate tactical improvements
to decade-long fundamental
631
00:33:56,848 --> 00:34:00,430
research. This multi-horizon
approach helps organizations
632
00:34:00,531 --> 00:34:03,457
avoid the common trap of
sacrificing long-term innovation
633
00:34:03,535 --> 00:34:04,879
for short-term deliverables.
634
00:34:05,535 --> 00:34:08,777
A financial services data
science team incorporated this
635
00:34:08,817 --> 00:34:11,504
principle by establishing three
parallel work streams with
636
00:34:11,535 --> 00:34:15,332
different time horizons. One
team focused on immediate
637
00:34:15,410 --> 00:34:17,567
analytical needs with two-week
delivery cycles.
638
00:34:18,004 --> 00:34:21,238
A second team tackled
medium-term innovations with
639
00:34:21,332 --> 00:34:24,951
quarterly objectives. A third
small team explored emerging
640
00:34:24,990 --> 00:34:27,914
technologies and methodologies
with annual goals and
641
00:34:27,933 --> 00:34:30,394
lightweight accountability
structures. Rather than
642
00:34:30,476 --> 00:34:33,922
competing for resources, these
teams operated as a coherent
643
00:34:33,976 --> 00:34:37,641
ecosystem, each reinforcing the
other's work while delivering
644
00:34:37,680 --> 00:34:39,484
value across different
timescales.
645
00:34:40,109 --> 00:34:42,609
R&D organizations have also
mastered the art of risk
646
00:34:42,703 --> 00:34:45,187
balancing through their
strategic approaches. They
647
00:34:45,266 --> 00:34:49,187
recognize that innovation
inherently involves failure, but
648
00:34:49,250 --> 00:34:52,000
they make this palatable to the
broader organization by
649
00:34:52,062 --> 00:34:55,551
structuring. How and when those
failures occur.
650
00:34:56,853 --> 00:35:00,102
Early-stage ideas face rigorous
but appropriate testing designed
651
00:35:00,121 --> 00:35:03,430
to quickly eliminate non-viable
concepts before significant
652
00:35:03,469 --> 00:35:06,987
investment occurs. As concepts
prove themselves, they receive
653
00:35:07,108 --> 00:35:09,889
progressively more resources and
increasingly stringent
654
00:35:09,928 --> 00:35:12,893
evaluation. A technology
company's data science division
655
00:35:12,932 --> 00:35:15,877
applied this staged funding
approach to their predictive
656
00:35:15,917 --> 00:35:16,854
modeling initiatives.
657
00:35:17,854 --> 00:35:21,479
Initial concepts received
minimal resources but faced
658
00:35:21,713 --> 00:35:24,792
weekly viability checks. Models
that demonstrated promise
659
00:35:24,838 --> 00:35:27,792
graduated to more substantial
funding with monthly evaluation
660
00:35:27,854 --> 00:35:30,448
cycles. Only proven approaches
that cleared these hurdles
661
00:35:30,495 --> 00:35:32,713
received full production
implementation resources.
662
00:35:32,948 --> 00:35:36,111
This staged approach... Allowed
the team to explore dozens of
663
00:35:36,171 --> 00:35:39,755
potential innovations while
focusing substantial resources
664
00:35:39,994 --> 00:35:42,841
only on approaches with
demonstrated potential, a
665
00:35:42,857 --> 00:35:46,201
dramatic improvement over their
previous all-or-nothing funding
666
00:35:46,240 --> 00:35:46,497
model.
667
00:35:47,404 --> 00:35:50,138
Perhaps the most valuable
strategic framework that data
668
00:35:50,201 --> 00:35:53,607
science can borrow from R&D
involves how these departments
669
00:35:53,966 --> 00:35:56,591
handle the fundamental tension
between exploration and
670
00:35:56,638 --> 00:35:59,888
exploitation. R&D leaders
understand that organizations
671
00:36:00,138 --> 00:36:02,242
need both. Breakthrough
innovations and continuous
672
00:36:02,283 --> 00:36:03,763
improvements to existing
offerings.
673
00:36:04,183 --> 00:36:07,929
Rather than treating this as a
binary choice, sophisticated R&D
674
00:36:08,046 --> 00:36:11,050
departments create structural
solutions that accommodate both
675
00:36:11,128 --> 00:36:14,714
needs. A healthcare analytics
organization that adopted this
676
00:36:14,777 --> 00:36:18,746
principle created distinct teams
with different behavioral
677
00:36:18,793 --> 00:36:19,277
commitments.
678
00:36:19,902 --> 00:36:23,527
Their exploration team operated
with greater autonomy, longer
679
00:36:23,558 --> 00:36:27,230
time horizons, and acceptance of
higher failure rates. Their
680
00:36:27,339 --> 00:36:30,785
exploitation team focused on
refining and scaling. Proven
681
00:36:30,805 --> 00:36:33,128
approaches with more structured
processes and delivery
682
00:36:33,187 --> 00:36:33,888
expectations.
683
00:36:34,468 --> 00:36:37,792
Critically, they established
deliberate transfer mechanisms
684
00:36:37,831 --> 00:36:40,835
between these teams, ensuring
successful explorations
685
00:36:41,335 --> 00:36:44,898
eventually became reliable
exploitations. R&D departments
686
00:36:44,921 --> 00:36:47,663
have also pioneered approaches
to measuring outcomes under
687
00:36:47,742 --> 00:36:51,538
uncertainty, another challenge
that plagues data science teams.
688
00:36:52,445 --> 00:36:55,507
Rather than attempting to
predict precise ROI for
689
00:36:55,617 --> 00:36:59,185
innovative initiatives, mature
R&D organizations establish
690
00:36:59,285 --> 00:37:02,528
learning objectives and progress
metrics that acknowledge the
691
00:37:02,589 --> 00:37:04,852
inherent uncertainty of
discovery work.
692
00:37:06,051 --> 00:37:09,313
A manufacturing company's data
science team shifted from
693
00:37:09,438 --> 00:37:12,477
attempting to quantify exact
financial returns for
694
00:37:12,618 --> 00:37:15,415
early-stage machine learning
initiatives to measuring
695
00:37:15,696 --> 00:37:19,071
learning velocity and knowledge
acquisition. For projects in the
696
00:37:19,072 --> 00:37:22,915
discovery phase, success metrics
included hypothesis validation
697
00:37:22,993 --> 00:37:25,930
rates, dataset quality
improvements, and reduction in
698
00:37:25,993 --> 00:37:28,055
predictive errors, not direct...
699
00:37:28,137 --> 00:37:31,358
Business outcomes. As
initiatives matured, measurement
700
00:37:31,438 --> 00:37:34,139
shifted toward operational
metrics and ultimately financial
701
00:37:34,198 --> 00:37:38,358
impact. This staged measurement
approach aligned perfectly with
702
00:37:38,381 --> 00:37:42,256
the inherent uncertainty of data
innovation. The strategic
703
00:37:42,342 --> 00:37:45,467
sophistication of R&D extends to
how these departments handle
704
00:37:45,499 --> 00:37:46,014
their talent.
705
00:37:46,124 --> 00:37:49,249
They recognize that innovation
requires different skills and
706
00:37:49,264 --> 00:37:51,827
working styles than
optimization, and they structure
707
00:37:51,858 --> 00:37:55,795
roles accordingly. Some R&D
staff excel at exploratory
708
00:37:55,858 --> 00:37:59,685
research while others specialize
in translating discoveries into
709
00:37:59,763 --> 00:38:03,427
practical applications. A retail
data science team applied this
710
00:38:03,485 --> 00:38:06,173
principle by creating
differentiated roles that
711
00:38:06,212 --> 00:38:07,970
acknowledge these distinct skill
sets.
712
00:38:08,931 --> 00:38:11,978
Some data scientists focused on
experimental applications and
713
00:38:12,032 --> 00:38:15,040
novel analytical approaches,
while others specialized in
714
00:38:15,134 --> 00:38:18,478
hardening promising techniques
into reliable production-grade
715
00:38:18,509 --> 00:38:18,978
solutions.
716
00:38:19,931 --> 00:38:23,071
By recognizing these as
complementary rather than
717
00:38:23,149 --> 00:38:26,345
competing approaches, they
created career paths. That
718
00:38:26,465 --> 00:38:30,507
valued both innovation and
implementation expertise. What
719
00:38:30,566 --> 00:38:34,164
makes these R&D strategic
approaches so valuable to data
720
00:38:34,226 --> 00:38:37,304
science is how directly they
address the unique challenges we
721
00:38:37,406 --> 00:38:38,265
identified earlier.
722
00:38:39,367 --> 00:38:42,343
The portfolio approach provides
a framework for maintaining
723
00:38:42,390 --> 00:38:45,562
consistency across diverse
initiatives. Multi-horizon
724
00:38:45,625 --> 00:38:48,515
planning offers a path to
coherence when balancing short-
725
00:38:48,547 --> 00:38:51,450
and long-term objectives. And
the structured approach to
726
00:38:51,630 --> 00:38:54,651
exploration and exploitation
helps maintain alignment between
727
00:38:54,733 --> 00:38:57,034
innovative impulses and
organizational needs.
728
00:38:58,636 --> 00:39:01,917
Best of all, these strategic
approaches come with decades of
729
00:39:01,964 --> 00:39:04,425
refinement and practical
validation across industries.
730
00:39:04,886 --> 00:39:07,886
Data science leaders don't need
to invent strategic frameworks
731
00:39:07,932 --> 00:39:10,964
from scratch. They can adapt
proven approaches from a
732
00:39:10,995 --> 00:39:13,807
discipline that has faced
similar challenges for
733
00:39:13,854 --> 00:39:14,511
generations.
734
00:39:15,479 --> 00:39:19,493
Applying Strategy to AI and ML
teams. Adapting proven
735
00:39:19,573 --> 00:39:22,514
frameworks is essential, but
what happens when your team
736
00:39:22,574 --> 00:39:25,375
ventures into the uncharted
territory of artificial
737
00:39:25,414 --> 00:39:30,219
intelligence? Despite borrowing
strategic wisdom from R&D, I
738
00:39:30,234 --> 00:39:34,437
discovered, through painful
experience, that AI and ML teams
739
00:39:34,898 --> 00:39:38,070
face unique challenges that
require fundamentally different
740
00:39:38,164 --> 00:39:40,430
behavioral commitments than
other data disciplines.
741
00:39:41,070 --> 00:39:44,476
My first leadership role over an
AI team taught me this lesson
742
00:39:44,492 --> 00:39:47,914
the hard way. We had implemented
a data science Strategy that
743
00:39:48,039 --> 00:39:50,744
worked brilliantly for our
analytics projects, clear
744
00:39:50,785 --> 00:39:53,471
timelines, defined deliverables,
and structured evaluation
745
00:39:53,510 --> 00:39:56,818
metrics. But when we applied the
same approach to our machine
746
00:39:56,857 --> 00:39:59,678
learning initiatives, we hit
unexpected roadblocks that no
747
00:39:59,679 --> 00:39:59,982
one had ever seen.
748
00:40:00,120 --> 00:40:03,544
Amount of planning could solve.
The core issue became evident.
749
00:40:04,263 --> 00:40:07,548
AI and machine learning work
operates under fundamentally
750
00:40:07,587 --> 00:40:10,552
different conditions than
traditional analytics. The
751
00:40:10,607 --> 00:40:13,912
uncertainty isn't just a factor,
it's the dominant condition.
752
00:40:14,294 --> 00:40:17,576
While analytics teams can often
predict how long it will take to
753
00:40:17,638 --> 00:40:21,576
extract insights from existing
data, AI teams are navigating a
754
00:40:21,607 --> 00:40:24,982
landscape where the very
feasibility of a solution might
755
00:40:25,029 --> 00:40:29,013
remain unknown until significant
resources have been invested.
756
00:40:29,760 --> 00:40:33,603
This unpredictability stems from
three characteristics unique to
757
00:40:33,664 --> 00:40:37,746
AI ML work. First, the search
space for solutions is
758
00:40:38,285 --> 00:40:41,191
exponentially larger, with
countless architectural
759
00:40:41,230 --> 00:40:44,957
decisions, hyperparameter
configurations, and data
760
00:40:45,035 --> 00:40:47,738
transformations that might
unlock performance
761
00:40:47,754 --> 00:40:48,238
breakthroughs.
762
00:40:49,332 --> 00:40:52,660
Second, the relationship between
input and output is often
763
00:40:52,847 --> 00:40:56,644
non-linear. Small changes can
produce dramatic improvements or
764
00:40:56,691 --> 00:40:57,972
failures with little warning.
765
00:40:59,352 --> 00:41:02,593
Third, the infrastructure
requirements evolve rapidly,
766
00:41:03,073 --> 00:41:06,073
sometimes necessitating complete
rebuilds of systems that worked
767
00:41:06,155 --> 00:41:09,675
perfectly just months earlier. A
Strategy designed for business
768
00:41:09,714 --> 00:41:13,097
intelligence or even traditional
statistical modeling simply
769
00:41:13,136 --> 00:41:15,401
doesn't accommodate these
realities. It's like trying to
770
00:41:15,448 --> 00:41:18,433
use a roadmap for ocean
navigation. The tools and
771
00:41:18,434 --> 00:41:20,839
assumptions fundamentally don't
match the territory.
772
00:41:22,214 --> 00:41:24,433
Consider what happened at a
computer vision startup I
773
00:41:24,464 --> 00:41:28,215
consulted for. Their initial
Strategy emphasized consistent
774
00:41:28,274 --> 00:41:31,337
weekly progress demonstrations
to investors. This worked
775
00:41:31,419 --> 00:41:33,997
perfectly for their dashboard
and analytics features, but
776
00:41:34,059 --> 00:41:37,465
created immense pressure on
their object detection team, who
777
00:41:37,504 --> 00:41:40,989
experienced weeks of seemingly
no progress, followed by sudden
778
00:41:41,067 --> 00:41:41,606
breakthroughs.
779
00:41:42,145 --> 00:41:45,129
The weekly demonstration cadence
pushed the team toward
780
00:41:45,254 --> 00:41:48,317
incremental improvements to show
something, rather than the
781
00:41:48,333 --> 00:41:51,692
exploratory work needed for
significant advances. The
782
00:41:51,739 --> 00:41:53,661
solution was a complete
strategic reframing.
783
00:41:54,484 --> 00:41:57,645
Rather than committing to weekly
visual progress, they shifted to
784
00:41:57,646 --> 00:42:00,606
a behavioral pattern of
maintaining comprehensive
785
00:42:00,645 --> 00:42:03,786
experiment logs and weekly
research reviews focused on
786
00:42:03,864 --> 00:42:06,966
learning rather than results.
This single strategic shift
787
00:42:07,286 --> 00:42:10,325
transformed their team's
performance, ultimately leading
788
00:42:10,380 --> 00:42:14,114
to a technical breakthrough that
became their core competitive
789
00:42:14,208 --> 00:42:14,755
advantage.
790
00:42:15,552 --> 00:42:18,708
This experience highlights why
AI and ML teams need unique
791
00:42:18,723 --> 00:42:22,597
behavioral commitments. While an
analytics team might thrive with
792
00:42:22,619 --> 00:42:26,203
a Strategy centered on rapid
iteration with stakeholders, ML
793
00:42:26,261 --> 00:42:29,664
teams often need extended
periods of technical exploration
794
00:42:30,008 --> 00:42:32,265
before stakeholder involvement
becomes valuable.
795
00:42:32,789 --> 00:42:35,687
The strategic behaviors that
create competitive advantage
796
00:42:36,187 --> 00:42:40,172
differ significantly. Successful
AI and ML strategies typically
797
00:42:40,219 --> 00:42:43,594
include specific behavioral
commitments in four key areas.
798
00:42:44,140 --> 00:42:45,844
First, experimentation cadence.
799
00:42:46,234 --> 00:42:49,905
How teams structure their
learning loops. One enterprise
800
00:42:49,985 --> 00:42:52,527
AI team I worked with committed
to running at least 20
801
00:42:52,647 --> 00:42:55,709
architecture experiments weekly
while documenting results in a
802
00:42:55,732 --> 00:42:58,772
standardized format. This
seemingly simple behavior
803
00:42:59,193 --> 00:43:02,615
created an invaluable knowledge
repository that accelerated
804
00:43:02,678 --> 00:43:04,959
their progress significantly
compared to competitors.
805
00:43:05,615 --> 00:43:09,178
Second, knowledge distribution
patterns, how insights flow
806
00:43:09,240 --> 00:43:12,272
between team members. A natural
language processing startup
807
00:43:12,553 --> 00:43:15,428
adopted a behavioral commitment
to pair programming for all
808
00:43:15,475 --> 00:43:18,615
model development work, despite
the apparent inefficiency.
809
00:43:19,564 --> 00:43:23,449
This strategic behavior resulted
in dramatically reduced
810
00:43:23,609 --> 00:43:27,191
knowledge silos and more robust
architecture decisions that
811
00:43:27,312 --> 00:43:28,375
competitors couldn't match.
812
00:43:29,297 --> 00:43:34,679
Third, failure tolerance
frameworks. How teams respond
813
00:43:35,062 --> 00:43:38,281
when approaches don't work. One
autonomous systems team
814
00:43:38,750 --> 00:43:41,672
committed to celebrating
valuable failures with formal
815
00:43:41,718 --> 00:43:44,875
recognition for experiments that
disproved promising hypotheses,
816
00:43:44,968 --> 00:43:46,297
saving future resources.
817
00:43:47,265 --> 00:43:49,904
This behavioral commitment
transformed how they approached
818
00:43:49,964 --> 00:43:53,005
risk and accelerated their
overall progress forth.
819
00:43:53,544 --> 00:43:57,122
Deployment bridges. How research
insights transition to
820
00:43:57,185 --> 00:43:58,044
production systems.
821
00:43:58,364 --> 00:44:02,403
A healthcare AI team implemented
a model shadowing period where
822
00:44:02,591 --> 00:44:06,528
new algorithms ran parallel to
existing systems for extended
823
00:44:06,575 --> 00:44:09,888
evaluation before taking
control. This strategic behavior
824
00:44:09,950 --> 00:44:13,091
built trust with clinical
partners and enabled more
825
00:44:13,122 --> 00:44:16,638
aggressive innovation than
competitors who faced resistance
826
00:44:16,685 --> 00:44:17,528
to implementation.
827
00:44:18,444 --> 00:44:21,387
These strategic behaviors
contrast sharply with those that
828
00:44:21,468 --> 00:44:24,809
work for analytics teams. Where
analytics teams might commit to
829
00:44:24,950 --> 00:44:28,934
business value in every sprint,
ML teams often need behavioral
830
00:44:28,997 --> 00:44:31,841
commitments like exploring at
least three novel approaches
831
00:44:31,895 --> 00:44:32,997
before optimization.
832
00:44:33,942 --> 00:44:37,380
Analytics teams can often commit
to fixed delivery dates, while
833
00:44:37,598 --> 00:44:41,114
ML teams need flexible
development cycles with rigorous
834
00:44:41,161 --> 00:44:41,942
evaluation gates.
835
00:44:42,801 --> 00:44:45,833
The contrast becomes even more
apparent in how these teams
836
00:44:46,255 --> 00:44:48,731
approach. Stakeholder
relationships. A financial
837
00:44:48,771 --> 00:44:52,096
services AI team, I advised,
struggled until they shifted
838
00:44:52,139 --> 00:44:55,201
from the analytics-inspired
behavior of demonstrating
839
00:44:55,280 --> 00:44:58,928
incremental progress to
stakeholders bi-weekly to
840
00:44:59,416 --> 00:45:03,120
educating stakeholders on
research processes and
841
00:45:03,239 --> 00:45:05,801
establishing quarterly
evaluation points.
842
00:45:07,407 --> 00:45:10,969
This strategic shift created the
space they needed for meaningful
843
00:45:11,008 --> 00:45:13,727
exploration while maintaining
organizational trust.
844
00:45:14,891 --> 00:45:17,297
Another critical difference
appears in technical debt
845
00:45:17,360 --> 00:45:17,813
management.
846
00:45:18,688 --> 00:45:22,219
Traditional analytics teams can
often defer maintenance work in
847
00:45:22,282 --> 00:45:26,094
favor of new features. For ML
teams, algorithm maintenance is
848
00:45:26,126 --> 00:45:29,560
existential. Models degrade as
the world changes. A retail
849
00:45:29,600 --> 00:45:33,264
recommendation engine team
committed to allocating 30% of
850
00:45:33,284 --> 00:45:36,608
resources to continuous
evaluation and retraining, which
851
00:45:36,666 --> 00:45:38,471
competitors viewed as
inefficient.
852
00:45:39,549 --> 00:45:42,416
This strategic behavior
ultimately proved decisive when
853
00:45:42,495 --> 00:45:45,776
market patterns shifted during
the pandemic, and their models
854
00:45:45,854 --> 00:45:49,276
adapted while competitors
failed. The distinction extends
855
00:45:49,277 --> 00:45:52,166
to talent development, too. A
computer vision team at a
856
00:45:52,198 --> 00:45:55,198
manufacturing company struggled
until they revised their
857
00:45:55,245 --> 00:45:57,541
Strategy to include a behavioral
commitment.
858
00:45:57,956 --> 00:46:01,759
Of dedicating 20% of senior
researchers' time to internal
859
00:46:01,779 --> 00:46:04,922
knowledge transfer. This
apparently costly behavior
860
00:46:05,023 --> 00:46:07,484
became their competitive
advantage as they could rapidly
861
00:46:07,547 --> 00:46:11,172
train new team members while
competitors fought over a
862
00:46:11,312 --> 00:46:14,945
limited pool of experienced
specialists. The research
863
00:46:15,007 --> 00:46:18,070
production tension requires
particularly careful strategic
864
00:46:18,132 --> 00:46:18,632
navigation.
865
00:46:19,132 --> 00:46:21,554
One autonomous vehicle company
committed to maintaining
866
00:46:21,664 --> 00:46:24,617
parallel research and production
implementations of key systems
867
00:46:24,956 --> 00:46:27,616
With explicit processes for
migrating innovations between
868
00:46:27,657 --> 00:46:31,098
them, this behavioral pattern
allowed them to simultaneously
869
00:46:31,176 --> 00:46:33,837
pursue cutting-edge research
while maintaining reliable
870
00:46:33,915 --> 00:46:36,962
production systems, a balance
many competitors failed to
871
00:46:36,977 --> 00:46:40,758
achieve. Even the approach to
failure requires different
872
00:46:40,938 --> 00:46:41,954
strategic behaviors.
873
00:46:42,860 --> 00:46:46,782
An NLP startup adopted a bounty
system for identifying failure
874
00:46:46,797 --> 00:46:50,297
modes in production models with
dedicated time allocations for
875
00:46:50,360 --> 00:46:54,422
addressing discoveries. This
strategic behavior created
876
00:46:55,008 --> 00:46:59,554
Resilient systems that
outperformed more academically
877
00:46:59,831 --> 00:47:03,718
advanced competitors, whose
models failed in unexpected
878
00:47:03,820 --> 00:47:04,835
real-world conditions.
879
00:47:05,398 --> 00:47:08,539
These examples demonstrate why
applying general data science
880
00:47:08,578 --> 00:47:12,749
strategies to AI and ML teams
frequently fails. The unique
881
00:47:12,781 --> 00:47:14,765
characteristics of machine
learning work, its
882
00:47:14,812 --> 00:47:17,937
exploration-heavy nature,
non-linear progress patterns,
883
00:47:17,968 --> 00:47:20,749
and complex deployment
challenges, require tailored
884
00:47:20,953 --> 00:47:24,093
strategic approaches focused on
specific behavioral commitments.
885
00:47:24,840 --> 00:47:26,922
The common Strategy mistake
leaders make.
886
00:47:27,502 --> 00:47:31,448
Behavioral patterns that drive
AI success highlight the
887
00:47:31,467 --> 00:47:33,967
critical difference between
planning and Strategy, a
888
00:47:34,006 --> 00:47:37,014
distinction most leaders miss
entirely. What if the reason
889
00:47:37,389 --> 00:47:41,139
most data science strategies
fail isn't about execution at
890
00:47:41,201 --> 00:47:44,858
all but stems from a fundamental
misunderstanding that happens
891
00:47:44,920 --> 00:47:46,920
the moment an executive makes
the request?
892
00:47:47,936 --> 00:47:51,545
Picture this scenario. A VP
calls you into their office and
893
00:47:51,608 --> 00:47:55,531
says, We need a data science
Strategy for our organization.
894
00:47:55,771 --> 00:47:57,853
Can you have something on my
desk by the end of the month?
895
00:47:59,775 --> 00:48:03,639
You nod confidently, but as you
walk away, a wave of uncertainty
896
00:48:03,717 --> 00:48:04,021
hits.
897
00:48:05,264 --> 00:48:08,342
What exactly are they asking
for? With limited guidance, most
898
00:48:08,381 --> 00:48:12,006
leaders default to creating what
they believe executives want to
899
00:48:12,146 --> 00:48:16,131
see. Detailed timelines,
technology selections, and
900
00:48:16,334 --> 00:48:17,287
project roadmaps.
901
00:48:18,334 --> 00:48:21,537
This scenario plays out
thousands of times across
902
00:48:22,136 --> 00:48:25,876
organizations worldwide, and it
triggers the most common and
903
00:48:25,896 --> 00:48:29,476
damaging mistake in data science
leadership, misinterpreting the
904
00:48:29,519 --> 00:48:32,917
request for Strategy as a
request for a comprehensive plan
905
00:48:32,918 --> 00:48:33,699
of activities.
906
00:48:34,660 --> 00:48:38,183
The distinction might seem
subtle, but the consequences are
907
00:48:38,199 --> 00:48:38,621
profound.
908
00:48:39,980 --> 00:48:43,855
When executives request a data
science Strategy, they typically
909
00:48:43,886 --> 00:48:47,183
expect a document that bears
little resemblance to what would
910
00:48:47,199 --> 00:48:51,204
actually help the team succeed.
They anticipate receiving an
911
00:48:51,224 --> 00:48:55,266
impressive slide deck filled
with timelines for implementing
912
00:48:55,309 --> 00:48:59,712
specific technologies, forecasts
of resource requirements, and
913
00:48:59,751 --> 00:49:01,657
promises of specific
deliverables.
914
00:49:02,259 --> 00:49:05,555
What they should receive instead
is a clear articulation of how
915
00:49:05,579 --> 00:49:08,259
the team will behave differently
from competitors to create
916
00:49:08,274 --> 00:49:11,337
sustainable advantage. I
witnessed this mistake firsthand
917
00:49:11,477 --> 00:49:14,180
at a major telecommunications
company where the newly
918
00:49:14,227 --> 00:49:18,009
appointed chief data officer
spent three months developing
919
00:49:18,055 --> 00:49:21,548
what he called a comprehensive
data science Strategy.
920
00:49:22,368 --> 00:49:25,769
The impressive 60-page document
detailed every project the team
921
00:49:25,851 --> 00:49:29,030
would undertake over the next
two years, the technologies they
922
00:49:29,046 --> 00:49:31,632
would implement, and the
projected ROI for each
923
00:49:31,687 --> 00:49:35,210
initiative. Leadership approved
it enthusiastically, and the
924
00:49:35,226 --> 00:49:37,929
team began execution
immediately. Six months later,
925
00:49:37,944 --> 00:49:39,757
the Strategy was effectively
abandoned.
926
00:49:40,616 --> 00:49:43,335
Market conditions had shifted,
executive priorities had
927
00:49:43,398 --> 00:49:46,413
changed, and several key
technologies mentioned in the
928
00:49:46,414 --> 00:49:50,918
document had been superseded. By
newer alternatives. The document
929
00:49:50,959 --> 00:49:54,043
provided no guidance on how to
adapt because it had focused
930
00:49:54,121 --> 00:49:57,406
entirely on what the team would
do, rather than how they would
931
00:49:57,469 --> 00:49:59,766
make decisions and behave to
create advantage.
932
00:50:00,513 --> 00:50:03,617
This contrasts sharply with a
mid-size insurance company whose
933
00:50:03,656 --> 00:50:06,180
data science leader took a
different approach. When asked
934
00:50:06,219 --> 00:50:10,086
for a Strategy, she delivered a
concise five-page document that
935
00:50:10,125 --> 00:50:12,562
barely mentioned specific
projects or technologies.
936
00:50:12,726 --> 00:50:16,562
Instead, it articulated clear
behavioral commitments, how they
937
00:50:16,594 --> 00:50:19,578
would evaluate investment
opportunities, their approach to
938
00:50:19,609 --> 00:50:22,875
balancing exploratory versus
exploitative work, and their
939
00:50:22,922 --> 00:50:25,953
decision criteria for pursuing
or abandoning initiatives.
940
00:50:27,219 --> 00:50:31,246
When market conditions shifted,
this team adapted their project
941
00:50:31,326 --> 00:50:34,531
portfolio while maintaining
consistent decision-making
942
00:50:34,589 --> 00:50:37,273
patterns that continued to
create advantage.
943
00:50:38,652 --> 00:50:41,574
The mistake stems from a
fundamental misconception of
944
00:50:41,613 --> 00:50:45,199
what constitutes Strategy. Most
data science leaders, especially
945
00:50:45,222 --> 00:50:48,621
those without formal Strategy
training, equate Strategy with
946
00:50:48,699 --> 00:50:51,964
detailed planning. They believe
that the more comprehensive and
947
00:50:52,011 --> 00:50:54,746
specific their plan, the more
strategic it appears.
948
00:50:55,074 --> 00:50:58,167
This leads to documents that
impress initially but provide
949
00:50:58,277 --> 00:51:01,138
minimal actual strategic
guidance. A strategic approach
950
00:51:01,179 --> 00:51:04,283
would instead focus on questions
like, what types of problems
951
00:51:04,302 --> 00:51:06,908
will we prioritize? How will we
decide between competing
952
00:51:06,924 --> 00:51:10,072
opportunities? What criteria
will guide our investment
953
00:51:10,088 --> 00:51:10,611
decisions?
954
00:51:11,666 --> 00:51:15,088
How frequently will we evaluate
results and pivot? How will we
955
00:51:15,119 --> 00:51:17,931
balance immediate needs against
long-term capability building?
956
00:51:19,010 --> 00:51:21,853
At an e-commerce company I
consulted with, the data science
957
00:51:21,854 --> 00:51:24,463
director initially presented a
Strategy that specified
958
00:51:24,525 --> 00:51:26,478
implementing a customer
segmentation model in Q1.
959
00:51:27,221 --> 00:51:30,264
Developing a recommendation
engine in Q2, and launching
960
00:51:30,424 --> 00:51:33,946
personalized marketing
automation in Q3. When asked how
961
00:51:33,989 --> 00:51:36,594
these activities would create
advantage relative to
962
00:51:36,610 --> 00:51:39,594
competitors who were pursuing
similar projects, he struggled
963
00:51:39,595 --> 00:51:40,032
to answer.
964
00:51:40,477 --> 00:51:44,078
The revised Strategy shifted
focus to behavioral commitments,
965
00:51:44,625 --> 00:51:47,407
running twice as many customer
experiments as competitors,
966
00:51:48,078 --> 00:51:52,078
making go-no-go decisions within
one week of results, and
967
00:51:52,125 --> 00:51:56,063
maintaining a consistent 30% of
resources dedicated to platform
968
00:51:56,157 --> 00:51:58,783
improvements, rather than
immediate applications.
969
00:51:59,364 --> 00:52:02,108
This Strategy proved resilient
through multiple business cycles
970
00:52:02,127 --> 00:52:04,889
and leadership changes because
it focused on how the team would
971
00:52:04,928 --> 00:52:09,475
behave rather than specific
deliverables. Two years later,
972
00:52:09,975 --> 00:52:12,373
not a single project from the
original plan remained
973
00:52:12,436 --> 00:52:15,905
unchanged, but the behavioral
patterns continued to generate
974
00:52:15,967 --> 00:52:16,998
competitive advantage.
975
00:52:17,155 --> 00:52:21,108
The mistake becomes increasingly
damaging as organizations grow
976
00:52:21,202 --> 00:52:24,418
and contexts evolve. A small
team might temporarily succeed
977
00:52:24,498 --> 00:52:27,320
with a project-focused approach
due to tight communication and
978
00:52:27,382 --> 00:52:30,804
shared implicit understanding.
But as teams expand, the absence
979
00:52:30,805 --> 00:52:33,206
of clear behavioral commitments
leads to inconsistent
980
00:52:33,230 --> 00:52:36,230
decision-making, resource
allocation conflicts, and
981
00:52:36,308 --> 00:52:37,074
strategic drift.
982
00:52:38,034 --> 00:52:41,878
A global financial services firm
experienced this firsthand as
983
00:52:41,909 --> 00:52:44,784
their data science function grew
from 15 to over 100
984
00:52:44,800 --> 00:52:48,378
professionals in 18 months. The
original Strategy document
985
00:52:48,456 --> 00:52:51,066
focused on specific analytical
models and technology
986
00:52:51,128 --> 00:52:51,847
implementations.
987
00:52:52,213 --> 00:52:56,117
Provided no guidance on how to
prioritize competing requests or
988
00:52:56,156 --> 00:52:59,340
balance innovation against
reliability. By the time they
989
00:52:59,379 --> 00:53:02,360
realized the issue, they had
dozens of disconnected
990
00:53:02,445 --> 00:53:05,789
initiatives with no coherent
pattern of behavior, resulting
991
00:53:05,805 --> 00:53:08,430
in organizational frustration
and wasted investment.
992
00:53:09,508 --> 00:53:11,758
Reframing the request for a data
science Strategy can
993
00:53:11,789 --> 00:53:14,680
fundamentally change outcomes.
When an executive asks for a
994
00:53:14,681 --> 00:53:17,664
Strategy, skilled leaders
clarify whether they're seeking
995
00:53:17,665 --> 00:53:20,799
a plan of activities or a
definition of how the team will
996
00:53:20,839 --> 00:53:24,581
behave to create advantage. This
simple clarification often
997
00:53:24,643 --> 00:53:27,725
surprises executives who haven't
considered the distinction,
998
00:53:28,069 --> 00:53:30,264
opening the door for a more
meaningful strategic
999
00:53:30,327 --> 00:53:30,866
conversation.
1000
00:53:31,811 --> 00:53:34,155
A healthcare analytics leader
handled this effectively when
1001
00:53:34,210 --> 00:53:37,218
asked for a Strategy by the CFO.
Instead of immediately producing
1002
00:53:37,219 --> 00:53:39,796
a document, she scheduled a
discussion to understand what
1003
00:53:39,843 --> 00:53:42,483
competitive challenges the
organization faced and how data
1004
00:53:42,514 --> 00:53:44,358
science could contribute to
addressing them.
1005
00:53:45,281 --> 00:53:48,164
This conversation revealed that
the CFO was primarily concerned
1006
00:53:48,203 --> 00:53:50,967
with improving operational
efficiency rather than
1007
00:53:51,026 --> 00:53:54,229
developing cutting-edge
analytical capabilities. Armed
1008
00:53:54,291 --> 00:53:58,252
with this insight, she developed
a Strategy focused on behavioral
1009
00:53:58,330 --> 00:54:00,994
commitments around rapid
implementation of proven
1010
00:54:01,057 --> 00:54:04,174
analytical techniques rather
than exploration of novel
1011
00:54:04,221 --> 00:54:04,737
approaches.
1012
00:54:05,643 --> 00:54:08,580
A pattern of behavior
specifically designed to address
1013
00:54:08,627 --> 00:54:12,705
the organization's actual
competitive needs. This approach
1014
00:54:13,034 --> 00:54:17,544
shifts the strategic focus from
what we will build to how we
1015
00:54:17,624 --> 00:54:18,585
will behave to win.
1016
00:54:19,167 --> 00:54:21,788
It transforms the conversation
from project selection to
1017
00:54:21,847 --> 00:54:25,948
capability building. Rather than
committing to specific outputs
1018
00:54:26,034 --> 00:54:30,112
that may become irrelevant, it
commits to decision patterns and
1019
00:54:30,113 --> 00:54:33,362
behaviors that remain valuable
regardless of how specific
1020
00:54:33,425 --> 00:54:34,253
projects evolve.
1021
00:54:35,768 --> 00:54:38,987
The most effective data science
strategies I've seen have been
1022
00:54:39,503 --> 00:54:42,222
remarkably brief in terms of
specific project commitments,
1023
00:54:42,268 --> 00:54:44,866
but crystal clear about
behavioral patterns. One
1024
00:54:44,886 --> 00:54:47,592
particularly successful example
came from a manufacturing
1025
00:54:47,633 --> 00:54:51,223
company whose entire data
science Strategy could be
1026
00:54:51,844 --> 00:54:54,090
summarized in three behavioral
commitments.
1027
00:54:54,732 --> 00:54:57,814
They would validate all models
with frontline operators before
1028
00:54:57,836 --> 00:55:01,158
deployment, maintain a
consistent four-week cycle from
1029
00:55:01,260 --> 00:55:05,725
concept to initial
implementation, and dedicate 25%
1030
00:55:05,803 --> 00:55:08,904
of capacity to improving
existing solutions rather than
1031
00:55:08,982 --> 00:55:13,123
creating new ones. These simple
principles guided hundreds of
1032
00:55:13,154 --> 00:55:16,326
decisions across dozens of
projects, creating a consistent
1033
00:55:16,420 --> 00:55:19,076
pattern that delivered
measurable competitive
1034
00:55:19,092 --> 00:55:19,592
advantage.
1035
00:55:21,279 --> 00:55:24,869
Strategic thinking for small
data science teams. Those simple
1036
00:55:24,929 --> 00:55:28,329
behavioral patterns created
measurable advantage for large
1037
00:55:28,368 --> 00:55:32,633
organizations. But what about
the small data science team of
1038
00:55:32,711 --> 00:55:35,915
just three people who told me
they were too small to need a
1039
00:55:35,930 --> 00:55:36,368
Strategy?
1040
00:55:37,532 --> 00:55:40,946
They later discovered something
surprising. Strategic thinking
1041
00:55:41,274 --> 00:55:45,977
becomes more critical as team
size decreases, not less. Small
1042
00:55:46,008 --> 00:55:48,930
data science teams face
constraints that make strategic
1043
00:55:48,993 --> 00:55:51,790
thinking seem like a luxury.
When you're a team of two to
1044
00:55:51,821 --> 00:55:54,987
five people. You're constantly
juggling immediate deliverables
1045
00:55:55,026 --> 00:55:56,207
while wearing multiple hats.
1046
00:55:56,709 --> 00:55:58,990
One day you're building models,
the next you're presenting to
1047
00:55:59,010 --> 00:56:02,112
executives, and the day after,
you're troubleshooting
1048
00:56:02,197 --> 00:56:05,213
infrastructure issues. The
pressure to deliver quick wins
1049
00:56:05,260 --> 00:56:08,604
is relentless, with stakeholders
expecting results despite your
1050
00:56:08,605 --> 00:56:12,260
limited resources. The typical
small team operates in perpetual
1051
00:56:12,338 --> 00:56:15,010
reactive mode. A marketing
request comes in Monday,
1052
00:56:15,088 --> 00:56:16,822
followed by a finance priority
on Tuesday.
1053
00:56:17,229 --> 00:56:19,494
Then the CEO sends an urgent
request Wednesday that
1054
00:56:19,526 --> 00:56:23,169
supersedes everything else. With
no bandwidth buffer, Each new
1055
00:56:23,209 --> 00:56:25,992
request creates a ripple effect
that disrupts previous
1056
00:56:26,031 --> 00:56:28,756
commitments. This constant
context switching severely
1057
00:56:28,814 --> 00:56:32,181
limits productivity and makes
focused work nearly impossible.
1058
00:56:32,681 --> 00:56:35,142
When I suggested strategic
thinking to a four-person
1059
00:56:35,221 --> 00:56:38,400
analytics team at a mid-sized
retailer, their leader laughed.
1060
00:56:38,822 --> 00:56:41,650
We barely have time to finish
our dashboards, much less
1061
00:56:41,713 --> 00:56:44,806
develop a formal Strategy
document. His perspective made
1062
00:56:44,822 --> 00:56:48,056
perfect sense. Formal Strategy
processes can be time-consuming.
1063
00:56:48,416 --> 00:56:50,791
But this fundamentally
misunderstands what Strategy
1064
00:56:50,916 --> 00:56:54,141
really entails for small teams.
The paradox is that these very
1065
00:56:54,182 --> 00:56:57,985
constraints make strategic
thinking not a luxury, but an
1066
00:56:58,024 --> 00:56:58,868
absolute necessity.
1067
00:56:59,387 --> 00:57:03,954
With severely limited resources,
small teams simply cannot afford
1068
00:57:04,235 --> 00:57:07,774
to waste time on initiatives
that don't create meaningful
1069
00:57:07,837 --> 00:57:12,055
competitive advantage. They lack
the buffer that larger teams
1070
00:57:12,102 --> 00:57:15,587
have to absorb misdirected
efforts or failed experiments.
1071
00:57:15,868 --> 00:57:19,040
Every hour spent on the wrong
priority is disproportionately
1072
00:57:19,133 --> 00:57:19,493
costly.
1073
00:57:20,784 --> 00:57:23,928
A strategic approach for small
teams doesn't require elaborate
1074
00:57:23,947 --> 00:57:26,729
documentation or formal
processes. Instead, it needs
1075
00:57:26,811 --> 00:57:29,674
only a clear understanding of
how the team will behave to
1076
00:57:29,713 --> 00:57:32,838
create competitive advantage
documented in the simplest
1077
00:57:32,940 --> 00:57:36,502
possible format. The retail team
eventually created a single-page
1078
00:57:36,580 --> 00:57:39,822
document with three behavioral
commitments that transformed
1079
00:57:39,838 --> 00:57:40,619
their effectiveness.
1080
00:57:41,166 --> 00:57:43,994
This lightweight approach
emphasized clear decision
1081
00:57:44,041 --> 00:57:46,447
criteria rather than
comprehensive planning. The
1082
00:57:46,494 --> 00:57:50,333
retail team committed to three
behaviors. They would only build
1083
00:57:50,454 --> 00:57:54,235
analytics that directly impacted
customer retention metrics. They
1084
00:57:54,278 --> 00:57:57,477
would deliver initial results
within two weeks of any project
1085
00:57:57,540 --> 00:58:00,524
initiation, and they would
allocate Friday afternoons to
1086
00:58:00,563 --> 00:58:02,243
improving their data
infrastructure.
1087
00:58:03,024 --> 00:58:06,134
These simple guardrails allowed
them to quickly evaluate
1088
00:58:06,243 --> 00:58:09,274
incoming requests without
lengthy meetings or analysis
1089
00:58:09,305 --> 00:58:13,493
paralysis. The startup FinTech
team I worked with implemented
1090
00:58:13,494 --> 00:58:17,261
what they called strategic
triage. A five-minute evaluation
1091
00:58:17,302 --> 00:58:19,404
of any new request against three
questions.
1092
00:58:20,384 --> 00:58:23,829
Does this directly support our
competitive differentiation in
1093
00:58:23,927 --> 00:58:25,111
automated loan processing?
1094
00:58:26,290 --> 00:58:29,056
Can we complete a valuable
version within our two-week work
1095
00:58:29,134 --> 00:58:32,431
cycle? Will the solution scale
beyond the immediate use case?
1096
00:58:33,150 --> 00:58:35,884
Any request failing these tests
was either declined or
1097
00:58:35,947 --> 00:58:39,603
renegotiated. Small teams can
implement strategic behavioral
1098
00:58:39,665 --> 00:58:42,822
patterns without bureaucracy by
focusing on clear decision
1099
00:58:42,884 --> 00:58:45,976
criteria, shared
knowledge-building habits, and
1100
00:58:46,036 --> 00:58:47,177
simplified communication.
1101
00:58:47,617 --> 00:58:51,119
The three-person data science
team at a property management
1102
00:58:51,158 --> 00:58:54,682
company created a simple
collaboration ritual, a
1103
00:58:55,260 --> 00:58:58,721
15-minute daily stand-up focused
not on status updates but on
1104
00:58:58,760 --> 00:59:02,002
decision alignment. Each team
member would share one key
1105
00:59:02,033 --> 00:59:05,424
decision they were facing and
get rapid input from colleagues.
1106
00:59:06,080 --> 00:59:08,846
This ensured consistent
strategic behavior across their
1107
00:59:08,877 --> 00:59:09,127
work.
1108
00:59:10,174 --> 00:59:13,127
Another effective pattern comes
from a five-person healthcare
1109
00:59:13,158 --> 00:59:17,390
analytics team who established
Strategy anchors, physical
1110
00:59:17,431 --> 00:59:20,552
reminders of their core
behavioral commitments. They
1111
00:59:20,634 --> 00:59:23,173
posted three questions
prominently in their workspace.
1112
00:59:23,916 --> 00:59:27,220
Are we prioritizing clinical
impact over technical elegance?
1113
00:59:28,400 --> 00:59:31,080
Are we creating solutions nurses
can use without training?
1114
00:59:32,142 --> 00:59:35,173
Are we reducing, not adding, to
clinician workload?
1115
00:59:36,111 --> 00:59:39,408
These visible reminders helped
maintain consistent behavior
1116
00:59:39,814 --> 00:59:43,314
without formal processes. What's
particularly powerful about
1117
00:59:43,361 --> 00:59:47,206
these approaches is that they
enable small teams to punch far
1118
00:59:47,267 --> 00:59:48,269
above their weight class.
1119
00:59:49,171 --> 00:59:52,159
The three-person insurance
analytics team I consulted with
1120
00:59:52,619 --> 00:59:55,361
competed directly against
competitors with 30-plus data
1121
00:59:55,402 --> 00:59:59,103
scientists. By focusing all
their efforts on a specific
1122
00:59:59,346 --> 01:00:02,869
customer segment underserved by
larger players and committing to
1123
01:00:03,650 --> 01:00:06,455
weekly customer interviews, they
created specialized solutions
1124
01:00:06,494 --> 01:00:07,650
with remarkable market fit.
1125
01:00:08,494 --> 01:00:11,150
Their behavioral commitment to
deep customer understanding
1126
01:00:11,244 --> 01:00:14,541
created leverage that larger
teams spread across many
1127
01:00:14,619 --> 01:00:18,010
initiatives couldn't match.
Similarly, a two-person data
1128
01:00:18,056 --> 01:00:21,213
science team at a regional
manufacturing company created
1129
01:00:21,371 --> 01:00:24,635
disproportionate impact by
focusing exclusively on
1130
01:00:24,713 --> 01:00:25,873
production line efficiency.
1131
01:00:26,956 --> 01:00:29,756
Rather than attempting
comprehensive coverage of all
1132
01:00:29,819 --> 01:00:33,225
potential analytics use cases,
they identified the single
1133
01:00:33,280 --> 01:00:35,983
highest leverage point in the
business and directed all their
1134
01:00:36,045 --> 01:00:40,217
work toward it. Their strategic
commitment to depth over breadth
1135
01:00:40,530 --> 01:00:43,936
allowed them to build solutions
with profound business impact
1136
01:00:43,937 --> 01:00:45,311
despite their tiny team size.
1137
01:00:45,851 --> 01:00:47,994
Small teams should focus their
strategic thinking on
1138
01:00:48,033 --> 01:00:50,576
identifying these leverage
points rather than attempting
1139
01:00:50,634 --> 01:00:54,400
comprehensive coverage. The
medical device startup I advised
1140
01:00:54,783 --> 01:00:58,166
succeeded by focusing
exclusively on enhancing their
1141
01:00:58,283 --> 01:01:01,767
sales team's effectiveness,
ignoring potentially valuable
1142
01:01:01,830 --> 01:01:04,783
but secondary applications in
operations and finance.
1143
01:01:05,752 --> 01:01:09,220
Their strategic behavior pattern
included a weekly leverage test
1144
01:01:09,408 --> 01:01:12,111
that evaluated whether any
current work was drifting away
1145
01:01:12,158 --> 01:01:15,732
from their core focus area. One
of the most Challenging aspects
1146
01:01:15,752 --> 01:01:19,415
for small teams is maintaining
strategic consistency despite
1147
01:01:19,458 --> 01:01:20,739
frequent priority shifts.
1148
01:01:21,700 --> 01:01:25,122
The four-person analytics team
at a software company addressed
1149
01:01:25,137 --> 01:01:28,684
this by creating a Strategy
contract with leadership. This
1150
01:01:28,723 --> 01:01:31,387
simple document outlined their
behavioral commitments and
1151
01:01:31,434 --> 01:01:35,043
clarified that any priority
shift requiring deviation from
1152
01:01:35,106 --> 01:01:37,872
these commitments would trigger
a formal review.
1153
01:01:38,653 --> 01:01:42,215
This moved them from passive
recipients of changing demands
1154
01:01:42,262 --> 01:01:45,392
to active partners in strategic
decisions. Another effective
1155
01:01:45,412 --> 01:01:48,395
technique comes from a
three-person municipal
1156
01:01:48,475 --> 01:01:51,237
government data team who
established priority buffers,
1157
01:01:52,096 --> 01:01:55,839
deliberately keeping 30% of
their capacity unallocated to
1158
01:01:55,885 --> 01:01:58,542
absorb inevitable urgent
requests without derailing
1159
01:01:58,628 --> 01:01:59,487
strategic work.
1160
01:02:00,729 --> 01:02:04,104
This buffer wasn't idle time,
but rather protected capacity
1161
01:02:04,464 --> 01:02:07,354
that allowed them to maintain
their core behavioral
1162
01:02:07,401 --> 01:02:10,944
commitments even when urgent
requests arrived. The two-person
1163
01:02:11,004 --> 01:02:14,066
team at an education technology
startup maintained consistency
1164
01:02:14,187 --> 01:02:16,549
through what they called
strategic minimums,
1165
01:02:17,068 --> 01:02:19,791
non-negotiable time allocations
for their highest leverage
1166
01:02:19,853 --> 01:02:20,416
activities.
1167
01:02:21,135 --> 01:02:24,838
They committed to spending at
least 25% of their time directly
1168
01:02:24,900 --> 01:02:27,674
observing how teachers used
their analytics products,
1169
01:02:27,877 --> 01:02:31,908
regardless of other demands.
This minimum viable strategic
1170
01:02:31,970 --> 01:02:35,674
behavior ensured they maintained
their competitive advantage in
1171
01:02:35,892 --> 01:02:40,043
user-centered design, despite
constant pressure to build. New
1172
01:02:40,183 --> 01:02:40,724
features.
1173
01:02:41,745 --> 01:02:45,108
Behaviors that win, building
your Strategy. Strategic
1174
01:02:45,190 --> 01:02:48,510
minimums establish a foundation,
but they don't tell you which
1175
01:02:48,573 --> 01:02:50,495
behaviors will actually win in
your market.
1176
01:02:51,596 --> 01:02:54,557
The most successful data science
teams I've worked with don't
1177
01:02:54,596 --> 01:02:55,815
just protect their time.
1178
01:02:56,792 --> 01:02:59,917
They deliberately engineer
behavioral patterns across four
1179
01:02:59,964 --> 01:03:02,839
domains that their competitors
consistently overlook.
1180
01:03:04,155 --> 01:03:06,716
These patterns aren't universal
templates, they're precision
1181
01:03:06,776 --> 01:03:09,778
instruments calibrated to your
specific competitive landscape.
1182
01:03:10,235 --> 01:03:13,497
Understanding which specific
behaviors to prioritize requires
1183
01:03:13,498 --> 01:03:17,161
a clear-eyed assessment of your
competitive context. What makes
1184
01:03:17,162 --> 01:03:21,559
a behavior strategic isn't that
it's good practice, it's that it
1185
01:03:21,638 --> 01:03:25,325
creates disproportionate
advantage in your particular
1186
01:03:25,356 --> 01:03:25,794
situation.
1187
01:03:26,341 --> 01:03:29,778
For example, a behavior like
meeting with stakeholders weekly
1188
01:03:30,278 --> 01:03:33,964
might be necessary hygiene in
one context, but a genuine
1189
01:03:34,044 --> 01:03:38,046
competitive advantage in another
organization where competitors
1190
01:03:38,448 --> 01:03:41,890
maintain distant relationships
with business partners.
1191
01:03:42,851 --> 01:03:45,312
The most effective data science
strategies I've encountered
1192
01:03:45,476 --> 01:03:48,718
organize behaviors into four
fundamental categories that
1193
01:03:48,741 --> 01:03:52,077
together form a comprehensive
approach. The first category
1194
01:03:52,187 --> 01:03:55,827
focuses on knowledge
acquisition, how your team
1195
01:03:56,140 --> 01:03:59,937
learns, shares information, and
builds institutional memory. The
1196
01:03:59,968 --> 01:04:03,087
second centers on decision
making. The processes, criteria,
1197
01:04:03,488 --> 01:04:05,569
and cadence of choices your team
makes.
1198
01:04:05,967 --> 01:04:10,651
The third addresses stakeholder
interaction, how you engage with
1199
01:04:10,768 --> 01:04:13,948
partners, users, and leadership.
The fourth encompasses your
1200
01:04:14,026 --> 01:04:17,791
technical approach, the methods,
tools, and standards that define
1201
01:04:18,104 --> 01:04:21,432
your work. Let's examine
specific winning behaviors
1202
01:04:21,526 --> 01:04:24,807
within each category. For
knowledge acquisition, a
1203
01:04:24,838 --> 01:04:28,291
consumer packaged goods company
I worked with committed to
1204
01:04:28,307 --> 01:04:31,120
spending 10% of their time
living with the data.
1205
01:04:32,123 --> 01:04:35,586
Directly examining raw data
without intermediary dashboards.
1206
01:04:36,488 --> 01:04:39,430
This behavior uncovered subtle
patterns their competitors
1207
01:04:39,691 --> 01:04:43,215
missed. A financial services
firm instituted mandatory
1208
01:04:43,332 --> 01:04:47,059
cross-pollination sessions where
data scientists regularly
1209
01:04:47,137 --> 01:04:51,121
rotated between product teams,
preventing knowledge silos that
1210
01:04:51,199 --> 01:04:52,699
plagued their industry peers.
1211
01:04:53,847 --> 01:04:56,810
For decision-making behaviors, a
retail analytics team
1212
01:04:56,929 --> 01:05:00,093
established a kill criteria
framework where projects were
1213
01:05:00,112 --> 01:05:03,394
evaluated against predetermined
failure conditions every two
1214
01:05:03,417 --> 01:05:03,714
weeks.
1215
01:05:04,714 --> 01:05:07,722
This prevented the resource
drain of prolonged unsuccessful
1216
01:05:07,784 --> 01:05:10,456
initiatives that their
competitors experienced. A
1217
01:05:10,550 --> 01:05:13,972
healthcare analytics
organization implemented what
1218
01:05:13,973 --> 01:05:17,191
they called parallel paths,
always maintaining at least two
1219
01:05:17,206 --> 01:05:20,175
competing approaches to
significant problems until clear
1220
01:05:20,237 --> 01:05:23,552
superiority emerged. This
prevented premature commitment
1221
01:05:23,553 --> 01:05:24,833
to suboptimal solutions.
1222
01:05:25,534 --> 01:05:29,118
In stakeholder interaction, a
manufacturing analytics team
1223
01:05:29,200 --> 01:05:33,059
practiced embedded rotation,
where data scientists spent one
1224
01:05:33,098 --> 01:05:36,067
week per quarter working
directly alongside factory
1225
01:05:36,083 --> 01:05:39,833
personnel. This created
intuitive understanding of
1226
01:05:39,895 --> 01:05:43,255
operational realities that
competitors with centralized
1227
01:05:43,270 --> 01:05:43,958
teams lacked.
1228
01:05:44,708 --> 01:05:48,317
A software company implemented
premortem workshops with
1229
01:05:48,395 --> 01:05:51,700
stakeholders before starting
projects. Imagining potential
1230
01:05:51,760 --> 01:05:55,265
failures, and building shared
risk ownership that competitive
1231
01:05:55,284 --> 01:05:58,749
teams struggle to establish. For
technical approach, a
1232
01:05:58,765 --> 01:06:01,507
telecommunications data science
team committed to maintaining
1233
01:06:01,546 --> 01:06:05,874
what they called model twins,
simpler, interpretable models
1234
01:06:06,077 --> 01:06:07,491
alongside more complex ones.
1235
01:06:08,476 --> 01:06:10,913
This balance between
explainability and performance
1236
01:06:11,241 --> 01:06:14,429
created trust with leadership
that competitors couldn't match.
1237
01:06:14,616 --> 01:06:18,288
A government agency analytics
team standardized on specific
1238
01:06:18,382 --> 01:06:21,551
modeling approaches within
domains, rather than chasing the
1239
01:06:21,572 --> 01:06:25,116
latest algorithms, creating
consistency that outperformed
1240
01:06:25,155 --> 01:06:28,479
competitors using more advanced
but less reliable techniques.
1241
01:06:29,358 --> 01:06:32,444
The specific competitive
advantage created varies
1242
01:06:32,522 --> 01:06:35,803
dramatically by context. A
behavior like conducting weekly
1243
01:06:35,905 --> 01:06:38,991
user research might create
massive advantage in a
1244
01:06:39,006 --> 01:06:42,069
consumer-facing digital product
company where competitors
1245
01:06:42,147 --> 01:06:45,975
prioritize technical innovation
over user feedback. That same
1246
01:06:46,022 --> 01:06:48,241
behavior might create minimal
advantage.
1247
01:06:48,431 --> 01:06:51,692
In industrial analytics where
domain expertise and system
1248
01:06:51,751 --> 01:06:55,811
reliability matter more than
user experience. To assess which
1249
01:06:55,913 --> 01:06:58,292
behaviors will create
competitive advantage in your
1250
01:06:58,331 --> 01:07:01,354
specific situation, start by
identifying your organization's
1251
01:07:01,456 --> 01:07:04,034
strategic priorities and
competitive landscape.
1252
01:07:04,956 --> 01:07:08,518
If your organization competes on
innovation speed, behaviors,
1253
01:07:09,050 --> 01:07:11,800
emphasizing rapid
experimentation, and lightweight
1254
01:07:11,893 --> 01:07:15,081
implementation will create
advantage. If it competes on
1255
01:07:15,143 --> 01:07:18,428
precision and reliability
Behaviors emphasizing thorough
1256
01:07:18,449 --> 01:07:21,090
validation and systematic
documentation will matter more.
1257
01:07:21,672 --> 01:07:24,414
Next, honestly evaluate your
competitors' behavioral
1258
01:07:24,473 --> 01:07:27,258
patterns. The greatest advantage
comes from strategic
1259
01:07:27,321 --> 01:07:30,938
differentiation, deliberately
choosing behaviors that contrast
1260
01:07:31,000 --> 01:07:32,102
with competitive norms.
1261
01:07:32,641 --> 01:07:35,399
If your competitors maintain
strict separation between data
1262
01:07:35,461 --> 01:07:38,883
scientists and business users,
behavior like co-development
1263
01:07:38,930 --> 01:07:42,611
with stakeholders might create
disproportionate advantage. If
1264
01:07:42,691 --> 01:07:45,895
they over-index on technical
sophistication at the expense of
1265
01:07:45,914 --> 01:07:49,176
implementation speed, a behavior
like deployment within two weeks
1266
01:07:49,177 --> 01:07:51,063
of approval might create more
value.
1267
01:07:51,938 --> 01:07:55,242
Document these behaviors in a
simple, actionable Strategy
1268
01:07:55,321 --> 01:07:58,453
statement that team members can
readily understand and apply.
1269
01:07:59,125 --> 01:08:02,328
Effective Strategy statements
avoid vague aspirations like
1270
01:08:02,516 --> 01:08:05,516
become data-driven in favor of
specific behavioral commitments.
1271
01:08:06,172 --> 01:08:09,141
A transportation analytics team
I worked with documented their
1272
01:08:09,235 --> 01:08:12,920
Strategy in a single page with
statements like We will maintain
1273
01:08:13,100 --> 01:08:15,963
no more than three active
projects per data scientist.
1274
01:08:17,025 --> 01:08:20,049
We will deploy models only after
they've been shadowed in
1275
01:08:20,068 --> 01:08:21,189
production for two weeks.
1276
01:08:22,088 --> 01:08:25,533
And we will dedicate every
Friday afternoon to improving
1277
01:08:25,611 --> 01:08:27,572
our collective knowledge through
teaching sessions.
1278
01:08:28,431 --> 01:08:31,213
The strongest Strategy
statements include not just what
1279
01:08:31,244 --> 01:08:34,822
you will do, but what you won't
do. A retail analytics team
1280
01:08:35,541 --> 01:08:40,650
explicitly committed to refusing
requests for one-off analysis
1281
01:08:40,713 --> 01:08:43,496
without... Clear connection to
customer retention. This
1282
01:08:43,576 --> 01:08:46,394
clarified priorities and
prevented resource dispersion
1283
01:08:46,894 --> 01:08:48,136
that competitors experienced.
1284
01:08:48,597 --> 01:08:51,215
Test your selected behaviors
against the three requirements
1285
01:08:51,216 --> 01:08:54,355
we discussed earlier. For
consistency, ask whether these
1286
01:08:54,394 --> 01:08:57,816
behaviors can be maintained
through market changes and
1287
01:08:57,925 --> 01:09:01,425
leadership transitions. For
coherence, verify that the
1288
01:09:01,426 --> 01:09:04,504
behaviors complement rather than
contradict each other. For
1289
01:09:04,535 --> 01:09:08,563
alignment, confirm that these
behaviors directly support your
1290
01:09:08,623 --> 01:09:10,804
organization's broader
competitive Strategy.
1291
01:09:11,185 --> 01:09:14,187
A technology company initially
committed to both rapid
1292
01:09:14,269 --> 01:09:18,269
prototyping and comprehensive
testing before discovering these
1293
01:09:18,316 --> 01:09:21,793
behaviors created internal
conflict. They refined their
1294
01:09:21,816 --> 01:09:25,472
approach to focus on targeted
testing of critical components,
1295
01:09:25,722 --> 01:09:27,847
which maintained quality while
enabling speed.
1296
01:09:28,644 --> 01:09:32,050
A healthcare analytics team
committed to behaviors optimized
1297
01:09:32,082 --> 01:09:34,832
for innovative discovery only to
realize their organization
1298
01:09:34,910 --> 01:09:37,695
competed primarily on
operational reliability. By
1299
01:09:37,795 --> 01:09:41,438
realigning their behaviors with
this reality, they created
1300
01:09:41,479 --> 01:09:44,964
significantly more value. The
most powerful test comes from
1301
01:09:45,260 --> 01:09:49,050
asking, if our competitors
adopted these exact same
1302
01:09:49,065 --> 01:09:51,206
behaviors, would we still have
an advantage?
1303
01:09:52,675 --> 01:09:55,138
If the answer is no, your
behaviors aren't strategically
1304
01:09:55,199 --> 01:09:58,441
differentiated enough. If yes,
you've identified genuinely
1305
01:09:58,499 --> 01:09:59,601
strategic commitments.
1306
01:10:00,562 --> 01:10:04,667
The patterns that consistently
lead to success aren't universal
1307
01:10:04,706 --> 01:10:07,870
best practices. Their behaviors
specifically match to your
1308
01:10:07,910 --> 01:10:11,253
competitive context. The most
successful data science teams
1309
01:10:11,347 --> 01:10:14,222
commit to patterns in knowledge
acquisition, decision-making,
1310
01:10:14,722 --> 01:10:17,863
stakeholder interaction, and
technical approach that create
1311
01:10:17,894 --> 01:10:19,816
advantage in their specific
situation.
1312
01:10:20,731 --> 01:10:23,875
Each organization must develop
its own unique combination based
1313
01:10:23,894 --> 01:10:26,619
on honest assessment of
organizational priorities and
1314
01:10:26,658 --> 01:10:29,978
competitive landscape. Testing
your Strategy, plan or true
1315
01:10:30,017 --> 01:10:30,400
Strategy.
1316
01:10:31,260 --> 01:10:33,963
Think your data science team has
a solid Strategy? Most
1317
01:10:34,002 --> 01:10:37,471
organizations confidently say
yes until they apply this simple
1318
01:10:37,502 --> 01:10:39,892
test that reveals they've
actually created nothing more
1319
01:10:39,924 --> 01:10:43,689
than a sophisticated plan. What
separates true strategies from
1320
01:10:43,752 --> 01:10:46,783
well-disguised task lists is a
specific element that's missing
1321
01:10:46,830 --> 01:10:48,611
in over 80% of Strategy
documents.
1322
01:10:49,191 --> 01:10:51,913
After years of reviewing data
science Strategy documents
1323
01:10:51,954 --> 01:10:55,298
across dozens of organizations,
I've developed a straightforward
1324
01:10:55,337 --> 01:10:57,958
test that can instantly
determine whether you've created
1325
01:10:57,959 --> 01:11:02,201
a genuine Strategy or merely an
elaborate plan disguised as one.
1326
01:11:03,161 --> 01:11:05,904
This test has proven remarkably
effective regardless of team
1327
01:11:05,982 --> 01:11:09,107
size or industry context. The
challenge with evaluating your
1328
01:11:09,185 --> 01:11:10,638
own Strategy is objectivity.
1329
01:11:11,513 --> 01:11:15,029
When we invest weeks or months
creating what we believe is a
1330
01:11:15,076 --> 01:11:18,513
strategic document, we develop
blind spots, we become attached.
1331
01:11:18,703 --> 01:11:22,648
To our creation and interpret
ambiguity in favorable ways.
1332
01:11:22,886 --> 01:11:26,190
This is why external Strategy
consultants exist. They provide
1333
01:11:26,210 --> 01:11:29,530
the objective perspective we
often lack, but you don't need
1334
01:11:29,554 --> 01:11:32,194
an expensive consultant to
evaluate your Strategy.
1335
01:11:32,913 --> 01:11:36,194
Instead, start with this
fundamental question. Does your
1336
01:11:36,241 --> 01:11:38,944
Strategy document clearly
articulate how your team will
1337
01:11:38,976 --> 01:11:41,788
behave differently from
competitors or alternative
1338
01:11:41,804 --> 01:11:46,194
approaches to create a specific
advantage? This seemingly simple
1339
01:11:46,257 --> 01:11:48,444
question cuts to the heart of
strategic thinking.
1340
01:11:49,319 --> 01:11:53,684
If your document primarily lists
what you'll do rather than how
1341
01:11:53,782 --> 01:11:56,926
you'll behave to win, it's
likely not a Strategy at all.
1342
01:11:57,168 --> 01:12:00,250
The test forces you to confront
whether you've made genuine
1343
01:12:00,289 --> 01:12:02,852
behavioral commitments that
create differentiation and
1344
01:12:02,875 --> 01:12:06,172
advantage. Let me walk you
through how this test works in
1345
01:12:06,188 --> 01:12:06,594
practice.
1346
01:12:07,500 --> 01:12:10,875
Recently, I evaluated a Strategy
document from a mid-size
1347
01:12:10,922 --> 01:12:13,985
insurance company's data science
team. Their document initially
1348
01:12:14,016 --> 01:12:16,877
appeared impressive. It
contained detailed sections on
1349
01:12:16,897 --> 01:12:19,739
technologies, methodologies, and
even specific projects scheduled
1350
01:12:19,759 --> 01:12:23,044
for implementation. But when I
applied the test question, the
1351
01:12:23,045 --> 01:12:24,067
document failed completely.
1352
01:12:24,364 --> 01:12:27,325
It contained no clear statements
about behavioral patterns that
1353
01:12:27,364 --> 01:12:30,028
would differentiate their
approach from competitors or
1354
01:12:30,091 --> 01:12:33,419
create specific advantages. When
I pointed this out, the team
1355
01:12:33,919 --> 01:12:37,513
initially pushed back, arguing
that their technology selections
1356
01:12:37,575 --> 01:12:40,865
represented strategic choices.
But upon deeper examination,
1357
01:12:40,884 --> 01:12:43,406
they recognized that their
choices were industry-standard
1358
01:12:43,427 --> 01:12:45,871
approaches that created no
competitive differentiation.
1359
01:12:46,468 --> 01:12:49,797
Their document was essentially
an implementation plan, not a
1360
01:12:49,812 --> 01:12:53,156
Strategy. Beyond the fundamental
test question, I recommend
1361
01:12:53,195 --> 01:12:55,883
applying these additional
diagnostic questions to
1362
01:12:55,992 --> 01:12:59,836
thoroughly evaluate your
Strategy. First, can your
1363
01:12:59,961 --> 01:13:03,429
behavioral commitments remain
stable if market conditions
1364
01:13:03,554 --> 01:13:04,523
change dramatically?
1365
01:13:05,463 --> 01:13:08,486
True strategic behaviors should
remain valid even when the
1366
01:13:08,685 --> 01:13:11,830
external environment shifts. If
your Strategy would need to be
1367
01:13:11,869 --> 01:13:14,810
completely rewritten following a
market disruption, it's likely
1368
01:13:14,849 --> 01:13:19,420
too tactical. Second, do your
commitments create tension by
1369
01:13:19,560 --> 01:13:22,803
forcing difficult trade-offs?
Effective strategies require
1370
01:13:22,896 --> 01:13:25,881
saying no to legitimate
opportunities.
1371
01:13:27,006 --> 01:13:30,287
If your Strategy allows you to
pursue virtually any reasonable
1372
01:13:30,334 --> 01:13:33,224
data science initiative, it
lacks the necessary focus to
1373
01:13:33,271 --> 01:13:36,865
create advantage. Third. Would
your competitors be
1374
01:13:37,185 --> 01:13:40,128
uncomfortable or unwilling to
adopt the same behavioral
1375
01:13:40,168 --> 01:13:43,652
patterns? If your stated
approach would be equally
1376
01:13:43,855 --> 01:13:47,574
appealing to competitors, it
likely lacks the differentiation
1377
01:13:47,636 --> 01:13:49,355
necessary for competitive
advantage.
1378
01:13:50,316 --> 01:13:53,261
Fourth, can team members easily
apply your Strategy to make
1379
01:13:53,324 --> 01:13:55,761
daily decisions without
constantly seeking approval?
1380
01:13:57,168 --> 01:14:00,793
Effective strategies provide
clear decision criteria that can
1381
01:14:00,794 --> 01:14:03,793
be applied consistently across
the organization. Let's examine
1382
01:14:04,083 --> 01:14:06,765
how these tests play out with
real examples.
1383
01:14:07,648 --> 01:14:10,948
A healthcare analytics team's
Strategy document stated they
1384
01:14:11,011 --> 01:14:14,370
would leverage cutting-edge
machine learning techniques to
1385
01:14:14,409 --> 01:14:15,675
improve patient outcomes.
1386
01:14:16,714 --> 01:14:19,980
This failed the test because it
described a goal rather than
1387
01:14:20,034 --> 01:14:23,519
specific behavioral patterns.
When pressed, the team couldn't
1388
01:14:23,581 --> 01:14:26,300
articulate how their approach
differed from competitors or
1389
01:14:26,331 --> 01:14:27,113
created advantage.
1390
01:14:27,191 --> 01:14:31,097
By contrast, another healthcare
organization's Strategy passed
1391
01:14:31,206 --> 01:14:34,204
the test with flying colors.
Their document committed to
1392
01:14:34,285 --> 01:14:37,346
conducting at least five
clinical workflow observations
1393
01:14:37,389 --> 01:14:40,491
per month before developing any
patient-facing analytics.
1394
01:14:41,530 --> 01:14:44,873
This behavioral pattern created
advantage by ensuring their
1395
01:14:44,952 --> 01:14:47,639
solutions addressed actual
clinical needs rather than
1396
01:14:47,694 --> 01:14:49,866
assumed ones, something their
competitors weren't
1397
01:14:49,913 --> 01:14:53,538
systematically doing. Even more
tellingly, when I tested this
1398
01:14:53,616 --> 01:14:56,288
second healthcare team's
Strategy against market
1399
01:14:56,319 --> 01:14:57,709
disruptions, it remained valid.
1400
01:14:58,288 --> 01:15:01,069
When the pandemic dramatically
altered healthcare delivery
1401
01:15:01,147 --> 01:15:04,123
models, their commitment to
workflow observation actually
1402
01:15:04,162 --> 01:15:07,506
became more valuable, not less.
They simply shifted from
1403
01:15:07,588 --> 01:15:10,186
in-person observations to
virtual ones while maintaining
1404
01:15:10,209 --> 01:15:14,076
the core behavioral commitment.
When organizations discover
1405
01:15:14,131 --> 01:15:17,678
their Strategy fails these
tests, they typically exhibit
1406
01:15:17,756 --> 01:15:18,881
one of three common gaps.
1407
01:15:19,319 --> 01:15:22,959
The first is excessive focus on
technical approaches rather than
1408
01:15:23,100 --> 01:15:26,366
competitive behaviors. The
second is a lack of clear
1409
01:15:26,506 --> 01:15:29,616
trade-offs that force
prioritization. The third is
1410
01:15:30,022 --> 01:15:33,996
insufficient specificity. About
behavioral patterns, using vague
1411
01:15:34,035 --> 01:15:37,417
statements like emphasize
quality, without defining what
1412
01:15:37,480 --> 01:15:39,640
specific behaviors constitute
that emphasis.
1413
01:15:40,359 --> 01:15:43,703
Addressing these gaps requires
iterative refinement. Start by
1414
01:15:43,788 --> 01:15:46,328
identifying the competitive
advantages your team needs to
1415
01:15:46,351 --> 01:15:50,406
create, then work backward to
define the specific behavioral
1416
01:15:50,468 --> 01:15:52,890
patterns that would generate
those advantages.
1417
01:15:53,390 --> 01:15:56,421
Test each behavioral commitment
against the diagnostic
1418
01:15:56,453 --> 01:15:59,265
questions, refining until you
have a set of commitments that
1419
01:15:59,328 --> 01:16:02,378
clearly pass the test. A
technology company I worked with
1420
01:16:02,719 --> 01:16:05,421
initially had a data science
Strategy focused entirely on
1421
01:16:05,460 --> 01:16:08,226
implementing specific machine
learning techniques across
1422
01:16:08,304 --> 01:16:09,507
various business functions.
1423
01:16:10,109 --> 01:16:13,367
When they applied the test, they
realized this approach failed to
1424
01:16:13,390 --> 01:16:16,288
differentiate them from
competitors who had access to
1425
01:16:16,289 --> 01:16:18,976
the same techniques. Through
several iterations, they
1426
01:16:19,054 --> 01:16:21,882
transformed their Strategy to
focus on behavioral commitments
1427
01:16:21,945 --> 01:16:24,804
around rapid experimentations.
Specifically, they committed to
1428
01:16:25,382 --> 01:16:26,304
running at least three.
1429
01:16:26,725 --> 01:16:29,465
Customer-facing experiments
weekly and making go-no-go
1430
01:16:29,565 --> 01:16:33,485
decisions within 24 hours of
receiving results. This approach
1431
01:16:33,606 --> 01:16:37,184
created advantage through speed
and learning velocity that their
1432
01:16:37,247 --> 01:16:40,669
more bureaucratic competitors
couldn't match. The revision
1433
01:16:40,708 --> 01:16:43,489
process typically requires three
to five iterations before a
1434
01:16:43,505 --> 01:16:46,036
Strategy successfully passes all
test questions.
1435
01:16:46,505 --> 01:16:49,270
Don't be discouraged if your
first attempt fails. This is
1436
01:16:49,317 --> 01:16:51,911
normal and part of the
development process. Each
1437
01:16:52,051 --> 01:16:54,567
iteration should bring you
closer to a true Strategy?
1438
01:16:55,237 --> 01:16:57,599
Focused on behavioral patterns
that create competitive
1439
01:16:57,638 --> 01:17:00,157
advantage. Once you've
successfully tested your
1440
01:17:00,200 --> 01:17:02,817
Strategy, document it concisely.
1441
01:17:04,102 --> 01:17:06,798
The most effective data science
strategies I've seen are rarely
1442
01:17:06,876 --> 01:17:10,001
longer than three to five pages.
They focus on clearly
1443
01:17:10,102 --> 01:17:12,845
articulating behavioral
commitments, decision criteria,
1444
01:17:13,001 --> 01:17:15,407
and the competitive advantages
these behaviors create.
1445
01:17:16,298 --> 01:17:18,923
Lengthy technical appendices or
detailed project plans may
1446
01:17:19,001 --> 01:17:21,079
complement a Strategy, but
should never substitute for
1447
01:17:21,142 --> 01:17:24,070
clear behavioral commitments.
Remember that Strategy testing
1448
01:17:24,129 --> 01:17:27,529
isn't a one-time exercise.
Revisit your test questions
1449
01:17:27,592 --> 01:17:30,928
quarterly to ensure your
Strategy continues to create the
1450
01:17:31,029 --> 01:17:34,295
intended advantages and remains
relevant as competitive
1451
01:17:34,334 --> 01:17:35,248
landscapes evolve.
1452
01:17:35,670 --> 01:17:38,850
This doesn't mean rewriting your
Strategy frequently. True
1453
01:17:38,912 --> 01:17:42,100
strategic behaviors should
remain stable over time, but
1454
01:17:42,147 --> 01:17:45,397
rather confirming that your
behavioral commitments continue
1455
01:17:45,412 --> 01:17:47,615
to create differentiation and
advantage.
1456
01:17:48,694 --> 01:17:51,818
Aligning your Strategy for
success. With your Strategy now
1457
01:17:51,859 --> 01:17:55,181
battle-tested and refined, you
face perhaps the most deceptive
1458
01:17:55,220 --> 01:17:58,404
challenge in implementation,
organizational alignment.
1459
01:17:59,603 --> 01:18:02,626
Most data science leaders
discover too late that their
1460
01:18:02,806 --> 01:18:06,611
technically sound strategies are
colliding with invisible
1461
01:18:06,673 --> 01:18:09,697
organizational barriers,
creating friction that can
1462
01:18:09,790 --> 01:18:12,478
quietly undermine months of
strategic work.
1463
01:18:13,400 --> 01:18:17,587
This alignment challenge often
emerges as a disconnect between
1464
01:18:17,588 --> 01:18:18,918
what you're doing and what
you're not doing. Team has
1465
01:18:18,978 --> 01:18:21,321
committed to and what the
organization is prepared to
1466
01:18:21,340 --> 01:18:24,403
support. Consider what happened
at a major automotive
1467
01:18:24,422 --> 01:18:27,110
manufacturer, where the data
science team developed a
1468
01:18:27,125 --> 01:18:30,211
brilliant Strategy focused on
predictive maintenance.
1469
01:18:30,610 --> 01:18:33,555
Their behavioral commitments
included running continuous
1470
01:18:33,633 --> 01:18:36,297
monitoring experiments and
deploying models within six-week
1471
01:18:36,360 --> 01:18:40,578
cycles. Despite the Strategy
passing all technical tests, it
1472
01:18:40,641 --> 01:18:43,547
faltered within months as
manufacturing unit managers
1473
01:18:44,000 --> 01:18:48,510
resisted integrating new systems
into their operations. And IT
1474
01:18:48,650 --> 01:18:52,114
governance committees imposed
nine-month security review
1475
01:18:52,192 --> 01:18:52,634
cycles.
1476
01:18:53,536 --> 01:18:56,439
The fundamental issue wasn't the
Strategy itself, but its
1477
01:18:56,517 --> 01:18:59,821
integration with the broader
organization. Research from the
1478
01:19:00,079 --> 01:19:04,845
MIT Sloan Management Review
indicates that 78% of data
1479
01:19:04,985 --> 01:19:08,017
initiatives fail not because of
technology or talent
1480
01:19:08,048 --> 01:19:10,939
limitations, but because of
organizational and cultural
1481
01:19:10,970 --> 01:19:11,407
factors.
1482
01:19:12,048 --> 01:19:15,331
This statistic reveals that the
critical importance of alignment
1483
01:19:15,411 --> 01:19:17,955
as the final but essential
component of Strategy
1484
01:19:18,033 --> 01:19:21,217
implementation. Successful
alignment requires managing
1485
01:19:21,256 --> 01:19:23,783
three crucial organizational
interfaces. The first is
1486
01:19:23,861 --> 01:19:27,220
executive leadership, where
alignment means ensuring your
1487
01:19:27,259 --> 01:19:30,642
behavioral commitments directly
address the competitive
1488
01:19:30,720 --> 01:19:32,127
challenges they perceive.
1489
01:19:33,330 --> 01:19:36,736
A pharmaceutical data science
leader discovered this when her
1490
01:19:36,814 --> 01:19:40,439
team's focus on accelerating
clinical trial analysis seemed
1491
01:19:40,517 --> 01:19:44,035
disconnected from leadership's.
Primary concern about declining
1492
01:19:44,255 --> 01:19:45,177
R&D productivity.
1493
01:19:45,897 --> 01:19:49,098
By reframing their Strategy to
explicitly show how faster
1494
01:19:49,122 --> 01:19:52,161
analysis would identify
unpromising compounds earlier,
1495
01:19:52,747 --> 01:19:56,911
saving millions in development
costs, she secured enthusiastic
1496
01:19:56,989 --> 01:19:59,997
executive support. Executive
alignment techniques that.
1497
01:20:00,048 --> 01:20:04,133
Work include regular Strategy
dialogues rather than one-time
1498
01:20:04,153 --> 01:20:07,856
presentations. A financial
services data science leader
1499
01:20:08,333 --> 01:20:11,778
established quarterly review
sessions with C-Suite executives
1500
01:20:11,856 --> 01:20:15,481
where he would present not just
outcomes but how specific
1501
01:20:15,520 --> 01:20:18,590
behavioral commitments were
creating competitive advantage.
1502
01:20:19,544 --> 01:20:21,872
These sessions evolved from
skeptical questioning to
1503
01:20:21,903 --> 01:20:24,934
collaborative Strategy
refinement as executives began
1504
01:20:24,965 --> 01:20:28,262
to see themselves as co-creators
rather than just approvers.
1505
01:20:28,948 --> 01:20:31,932
Another effective approach is
creating a translator document
1506
01:20:32,531 --> 01:20:35,312
that maps your strategic
behavioral commitments to
1507
01:20:35,375 --> 01:20:38,117
executive priorities. A
healthcare analytics team
1508
01:20:38,195 --> 01:20:41,258
created a one-page visualization
showing how each of their
1509
01:20:41,320 --> 01:20:44,906
strategic behaviors directly
supported the CEO's three stated
1510
01:20:44,969 --> 01:20:45,547
priorities.
1511
01:20:46,812 --> 01:20:50,016
This visual aid became a
powerful alignment tool that
1512
01:20:50,062 --> 01:20:52,656
executives actually kept on
their desks for reference. The
1513
01:20:52,672 --> 01:20:55,219
second crucial interface is with
partner teams, the groups whose
1514
01:20:55,250 --> 01:20:58,422
cooperation you need to succeed.
These typically include IT,
1515
01:20:58,764 --> 01:21:01,585
product teams, business units,
and sometimes compliance or
1516
01:21:01,624 --> 01:21:02,444
legal departments.
1517
01:21:02,985 --> 01:21:05,767
Alignment here requires
understanding their incentives,
1518
01:21:05,825 --> 01:21:08,509
constraints, and working styles.
A retail data science team
1519
01:21:08,563 --> 01:21:11,087
discovered that their commitment
to rapid experimentation was
1520
01:21:11,126 --> 01:21:13,985
constantly frustrated by the IT
department's quarterly release
1521
01:21:14,024 --> 01:21:14,384
schedule.
1522
01:21:15,048 --> 01:21:18,251
Rather than battling against
this constraint, they created a
1523
01:21:18,454 --> 01:21:21,907
lightweight experimentation
environment in collaboration
1524
01:21:21,985 --> 01:21:26,626
with IT, allowing them to run
customer-facing tests without
1525
01:21:26,673 --> 01:21:30,121
disrupting production systems.
This collaborative solution
1526
01:21:30,200 --> 01:21:32,880
preserved their strategic
behavior while respecting IT's
1527
01:21:32,942 --> 01:21:34,860
legitimate concerns about system
stability.
1528
01:21:35,282 --> 01:21:38,758
For partner team alignment,
joint working sessions are far
1529
01:21:38,805 --> 01:21:41,782
more effective than formal
handoffs. A manufacturing
1530
01:21:41,844 --> 01:21:44,587
analytics team instituted
monthly Strategy alignment
1531
01:21:44,665 --> 01:21:48,188
workshops where representatives
from operations, engineering,
1532
01:21:48,204 --> 01:21:52,266
and IT would work through
practical implications of the
1533
01:21:52,282 --> 01:21:54,266
data science team's behavioral
commitments.
1534
01:21:55,266 --> 01:21:58,568
These sessions revealed
potential conflicts early and
1535
01:21:58,668 --> 01:22:02,471
created shared ownership of
solutions. Creating explicit
1536
01:22:02,534 --> 01:22:06,299
interface agreements has also
proven highly effective. An
1537
01:22:06,300 --> 01:22:09,541
insurance data science team
developed partnership charters
1538
01:22:10,018 --> 01:22:12,659
with each major internal
collaborator, documenting
1539
01:22:12,706 --> 01:22:15,924
expectations, communication
protocols, and escalation
1540
01:22:15,971 --> 01:22:16,487
procedures.
1541
01:22:16,924 --> 01:22:20,237
These agreements reduced
friction and helped all parties
1542
01:22:20,409 --> 01:22:23,627
understand how their success
metrics interrelated. The third
1543
01:22:23,690 --> 01:22:26,681
critical interface is within
your own team. Ensuring
1544
01:22:26,761 --> 01:22:29,144
individual contributors
understand and embrace the
1545
01:22:29,163 --> 01:22:30,023
strategic behaviors.
1546
01:22:30,702 --> 01:22:35,007
A study by Gallup found that
only 27% of employees strongly
1547
01:22:35,070 --> 01:22:38,109
agree that they believe in their
company's values. For data
1548
01:22:38,156 --> 01:22:41,999
science teams, this disconnect
often appears as misalignment
1549
01:22:42,038 --> 01:22:44,577
between stated strategic
behaviors and daily work
1550
01:22:44,624 --> 01:22:48,140
priorities. A telecommunications
data science leader addressed
1551
01:22:48,156 --> 01:22:51,562
this by redesigning performance
metrics to directly reflect the
1552
01:22:51,577 --> 01:22:52,827
team's strategic behaviors.
1553
01:22:53,280 --> 01:22:56,864
If the Strategy emphasized rapid
experimentation, individual
1554
01:22:56,903 --> 01:23:00,567
performance reviews highlighted
experiment velocity and learning
1555
01:23:00,625 --> 01:23:05,594
quality rather than just model
accuracy. This alignment created
1556
01:23:05,735 --> 01:23:09,118
powerful reinforcement of
strategic behaviors throughout
1557
01:23:09,133 --> 01:23:09,899
the organization.
1558
01:23:10,352 --> 01:23:13,321
Team alignment also requires
making strategic behaviors
1559
01:23:13,383 --> 01:23:16,258
visible in daily work. A
technology company's data
1560
01:23:16,305 --> 01:23:19,024
science team integrated their
behavioral commitments into
1561
01:23:19,102 --> 01:23:22,305
project kickoff templates,
stand-ups, and retrospectives.
1562
01:23:23,020 --> 01:23:26,741
They started each meeting with
the question, how is today's
1563
01:23:26,860 --> 01:23:28,920
work advancing our strategic
behaviors?
1564
01:23:29,940 --> 01:23:33,104
This constant reinforcement
transformed abstract commitments
1565
01:23:33,143 --> 01:23:35,784
into practical daily choices.
While establishing these
1566
01:23:35,799 --> 01:23:38,401
alignments, you'll inevitably
encounter resistance and
1567
01:23:38,424 --> 01:23:41,409
blockers. The key is identifying
them before they derail your
1568
01:23:41,440 --> 01:23:41,877
Strategy.
1569
01:23:43,190 --> 01:23:45,768
Common alignment blockers
include conflicting incentive
1570
01:23:45,784 --> 01:23:49,502
structures, legacy governance
processes, and competing
1571
01:23:49,518 --> 01:23:50,893
priorities among stakeholders.
1572
01:23:52,024 --> 01:23:56,769
To identify potential blockers,
conduct a premortem exercise
1573
01:23:56,770 --> 01:23:59,851
where you imagine it's one year
later and your Strategy has
1574
01:24:00,011 --> 01:24:03,995
failed due to alignment issues.
What happened? A biotechnology
1575
01:24:04,058 --> 01:24:07,597
data science team used this
technique to identify that their
1576
01:24:08,198 --> 01:24:11,839
commitment to rapid model
deployment would conflict with
1577
01:24:11,840 --> 01:24:14,089
the company's rigid validation
protocols.
1578
01:24:14,901 --> 01:24:18,214
By proactively addressing this
with quality assurance teams,
1579
01:24:18,917 --> 01:24:21,816
they developed a tiered
validation approach. That
1580
01:24:21,916 --> 01:24:25,018
maintained necessary safeguards
while enabling their strategic
1581
01:24:25,079 --> 01:24:28,542
behaviors. Once identified,
addressing blockers requires a
1582
01:24:28,561 --> 01:24:31,163
combination of education,
adaptation, and negotiation.
1583
01:24:32,405 --> 01:24:35,710
Education means helping others
understand the competitive
1584
01:24:35,749 --> 01:24:37,608
rationale behind your strategic
behaviors.
1585
01:24:38,678 --> 01:24:42,038
Adaptation means modifying
implementation details while
1586
01:24:42,053 --> 01:24:45,132
preserving core behavioral
commitments. Negotiation means
1587
01:24:45,241 --> 01:24:47,866
finding compromise solutions
that meet multiple needs.
1588
01:24:48,534 --> 01:24:51,657
Perhaps most challenging is
maintaining alignment as both
1589
01:24:51,676 --> 01:24:53,377
your Strategy and the
organization evolve.
1590
01:24:53,877 --> 01:24:56,838
A consumer goods company's data
science team instituted
1591
01:24:56,924 --> 01:25:00,042
quarterly alignment health
checks where they systematically
1592
01:25:00,088 --> 01:25:02,565
assess the current state of
alignment across all three
1593
01:25:02,604 --> 01:25:05,815
interfaces. These structured
reviews helped them spot
1594
01:25:05,924 --> 01:25:08,096
emerging misalignments before
they became problematic.
1595
01:25:08,237 --> 01:25:11,049
Creating mechanisms for ongoing
dialogue is essential.
1596
01:25:11,877 --> 01:25:16,002
A healthcare analytics team
established a Strategy advisory
1597
01:25:16,096 --> 01:25:19,076
council with representatives
from across the organization.
1598
01:25:19,516 --> 01:25:22,637
This forum provided regular
opportunities to discuss how
1599
01:25:22,700 --> 01:25:25,239
changing organizational
priorities or market conditions
1600
01:25:25,403 --> 01:25:28,043
might affect alignment with the
data science Strategy.
1601
01:25:29,129 --> 01:25:32,090
Successful alignment also
requires adapting communication
1602
01:25:32,145 --> 01:25:35,332
styles to different audiences.
When discussing Strategy with
1603
01:25:35,364 --> 01:25:39,551
executives, focus on competitive
outcomes. With partner teams,
1604
01:25:39,801 --> 01:25:42,864
emphasize collaborative
workflows. With team members,
1605
01:25:43,223 --> 01:25:45,848
connect strategic behaviors to
professional growth and
1606
01:25:45,942 --> 01:25:46,723
meaningful impact.
1607
01:25:47,946 --> 01:25:50,988
Throughout this journey, we've
transformed our understanding of
1608
01:25:51,008 --> 01:25:53,691
data science Strategy from a
mere project plan to a
1609
01:25:53,711 --> 01:25:56,918
commitment of behavioral
patterns that create competitive
1610
01:25:56,973 --> 01:26:01,637
advantage. This paradigm shift
enables teams of any size to be
1611
01:26:01,723 --> 01:26:04,582
more adaptable while maintaining
strategic integrity.
1612
01:26:05,019 --> 01:26:07,660
Your immediate task is
straightforward yet profound.
1613
01:26:08,285 --> 01:26:11,535
Examine your current approach
using our strategic framework.
1614
01:26:12,207 --> 01:26:15,519
Ask yourself whether your team
has clear behavioral commitments
1615
01:26:15,566 --> 01:26:18,379
that directly contribute to
winning in your competitive
1616
01:26:18,419 --> 01:26:22,643
context. Remember that Strategy
manifests through daily actions,
1617
01:26:22,983 --> 01:26:24,045
not PowerPoint slides.
1618
01:26:24,405 --> 01:26:27,670
The most successful data science
teams, regardless of size or
1619
01:26:27,671 --> 01:26:30,592
resources, are those who
consistently embody their
1620
01:26:30,631 --> 01:26:34,139
strategic behaviors, adapting
tactics while maintaining their
1621
01:26:34,202 --> 01:26:37,655
competitive positioning. The
true power of Strategy lies not
1622
01:26:37,717 --> 01:26:41,077
in predicting the future, but in
building a team that can thrive
1623
01:26:41,123 --> 01:26:41,467
within it.

Transkript bearbeiten

VERGANGENE FOLGEN SUCHEN

Vorhandene Folgen von DataScience Show Podcast durchsuchen.

Betten Sie diese Suchleiste in Ihre Website ein

Rechtliche Hinweise: Der auf dieser Seite eingebettete Podcast und das Bildmaterial stammen von Mirko Peters, das Eigentum seines Eigentümers ist und nicht mit Listen Notes, Inc. verbunden ist oder von diesen unterstützt wird.