Navigating the Data-Driven Path to Smart Mobility ft. Rob Roemers

WEBVTT
1
00:00:00.100 --> 00:00:10.080
Welcome to the Data Strategy Gurus podcast. In this show, we bring together the brightest minds in the world of data strategy, data management, artificial intelligence, and disruptive technologies.
2
00:00:10.300 --> 00:00:16.840
Thought leaders and experts share their insights, knowledge, and experience on how to stay ahead of the game in an ever-evolving data landscape.
3
00:00:17.140 --> 00:00:23.500
Whether you're a data professional, a business leader, or simply someone who is passionate about the power of data, this podcast is for you.
4
00:00:23.800 --> 00:00:34.520
So sit back, relax, and join us on a journey to explore the world of data, analytics, artificial intelligence, tech, and beyond. Hi, and welcome to the Data Strategy Gurus podcast.
5
00:00:34.620 --> 00:00:50.060
Today, we have Rob Roemers with us, a phenomenon in the data and analytics space, all out of Belgium. So for me, not so far, uh, distant. Rob is the head of data and analytics at the public company STIB-MIVB. Hi, Rob.
6
00:00:50.240 --> 00:01:01.920
Welcome, uh, to the Data Strategy Gurus podcast. Thank you, Iven. Uh, thank you for having me. I was listening to your previous, uh, recordings, and I'm, uh, I'm honored to be in the same list as, uh, Bill Inmon, so.
7
00:01:02.180 --> 00:01:15.250
[laughs] Oh, yeah, [laughs] yeah. Well, finally he became a big friend of mine, so for me it was always a big beacon. Uh, the guy who told us how to build, uh, data warehouses, has a specific, um,
8
00:01:16.240 --> 00:01:23.800
idea about text analytics, so it was a very interesting conversation, uh, with, uh, with Bill, uh, as well.
9
00:01:24.280 --> 00:01:36.880
So Rob, can you tell us a bit more about the STIB-MIVB's journey, uh, to become a data and analytics company in a short way? Uh, we will go through the journey in more depth as well.
10
00:01:36.980 --> 00:01:51.380
But, uh, a bit high level so the listeners, uh, can understand a bit better, uh, how you became data and analytics driven. Well, uh, STIB-MIVB, as you know, it's the Brussels public transport operator. Um,
11
00:01:52.560 --> 00:02:01.580
and I- I'm not gonna say that we are a data analytics company. I think it's, it's a continuing road, uh, that we're on. Um, how, how did we start?
12
00:02:01.620 --> 00:02:13.060
Well, actually, it's- it's a very old company, and it started, uh, already with the horse-driven carts or horse- horse-driven trams. Uh, and from there, we developed into bus, tram, and metro, uh, like we're doing today.
13
00:02:13.180 --> 00:02:23.500
So I- I cannot claim that I invented data for, for the organization. We've been making timetables and planning how many drivers we needed and how many vehicles we needed for many, many years.
14
00:02:24.200 --> 00:02:33.580
Um, and at the core, it's a very engineering-driven company, and we build metros. We, uh, have our own electricity network, so there, there's a lot of, uh, technology there.
15
00:02:33.640 --> 00:02:43.090
So in that sense, it's a fertile, fertile environment to try and sell, uh, data analytics. Um, and I've been there for, uh, a really long time already.
16
00:02:43.140 --> 00:02:51.340
And as you know, I used to have, uh, my own consulting company specializing in data, so STIB was actually one of my last customers before I sold my business.
17
00:02:51.960 --> 00:02:59.980
Uh, it was literally across the street from where I lived in Brussels, the- the main, uh, office. So that's one of the reasons why I kind of hung around, uh, afterwards.
18
00:03:01.170 --> 00:03:07.020
Um, and I started off as, uh, as head of the SAP Business Intelligence department. So we, we started working on that.
19
00:03:07.420 --> 00:03:16.560
Uh, realized fairly early on, uh, when, when I took that job that if we were just gonna be doing SAP data warehousing, that we'd be irrelevant fairly quickly.
20
00:03:17.160 --> 00:03:21.820
I started around two thousand and sixteen internally, I think two thousand and ten, uh, when I was freelancing.
21
00:03:22.480 --> 00:03:36.280
Um, and we, we really turned around the, the environment of we only have SAP data warehousing, and we try to fit everything into that, into a more, uh, holistic idea of what a good data analy- analytics department would look like.
22
00:03:36.340 --> 00:03:44.140
So we do data science now. We have a small data science department. Uh, it's a very federated model at the STIB. So I'm-- we are the central platform team that
23
00:03:45.160 --> 00:03:57.240
services the rest, but the idea is that it's really federated, that every, uh, business unit can have their own full set of, of analysts. Uh, so my focus is mostly platform. Like I said, we have the SAP environment.
24
00:03:57.310 --> 00:04:06.080
We have, uh, an Azure environment. Uh, we have a data science department. We do a integration layer, API manager, uh, that, that, uh, type of stuff that's really with us.
25
00:04:06.120 --> 00:04:17.040
And then we focus with a department that we call the Office of Data Analytics on, on change management, on data literacy to really change that culture. Yeah.
26
00:04:17.170 --> 00:04:27.720
I think you tackled some of the initial, uh, challenges you faced in demonstrating the value of data and analytics. One part was already solved, that they were working with, with the timetables over time.
27
00:04:28.280 --> 00:04:39.600
Uh, but what I hear is as well a, a very engineering-driven company. That's harder to say, "Let's, let's explore the data, test it, fail it, learn from it, and go forward."
28
00:04:39.700 --> 00:04:44.900
Engineers typically try to build the most solid product and move forward for- from that.
29
00:04:44.960 --> 00:05:00.900
So there I feel some friction in, in how you need to transform in, in that typical way of putting up timetables, operational analytics, and then now more in an explorative way and trying to measure kind of everything what you can measure.
30
00:05:01.100 --> 00:05:19.180
I, I recall in one of our previous discussions where you said you were challenging with the people building really the, the trams and putting some sensors on the doors, and that's how far you went to enable the possibility to capture all possible data that could optimize, uh, the operations.
31
00:05:19.980 --> 00:05:31.420
So what were the, the biggest frictions, and how did you overcome those? Well, there, there's a lot of them, to be honest, because STIB is a very large company, and it- it's very diverse. It's ten thousand people.
32
00:05:31.500 --> 00:05:34.340
It's, it's one of the largest employers in, in the Brussels capital region.
33
00:05:34.940 --> 00:05:47.060
Um, and we indeed work with departments that maintain the trams, maintain the metros, that build the, the roads, but you're also talking to marketing, to finance, to HR. So the, the challenges are diverse. Um,
34
00:05:47.980 --> 00:05:55.760
if you're looking at the more engineering-driven ones, the, the problem for me was not to convince them to use data, but to convince them to do it in a structured way.
35
00:05:56.440 --> 00:06:03.719
Because like you said, they're engineers, they like to test and play themselves. They usually come with a fully thought out solution and say, "This is what we're gonna do."
36
00:06:03.860 --> 00:06:12.464
And then you have to like, "ActuallyThat doesn't really fit the [laughs] the roadmap that we have, but could work. Um, but we're gonna do it a little bit differently.
37
00:06:12.524 --> 00:06:18.324
So there's a lot of discussions about, you know, should, should, should we put this data in the cloud or not? And they all have their opinion on that.
38
00:06:18.404 --> 00:06:27.844
Uh, so there, there, there was a lot of work on that early on to say like, we need to have one platform. We make one common investment, and then everybody gets all the freedom they want within the platform.
39
00:06:27.944 --> 00:06:37.514
'Cause they were very much afraid, um, that they would lose access to what they considered their data. Uh, the people from the bus would say, "Well, I need all these sensor data from my vehicle."
40
00:06:37.644 --> 00:06:44.084
Because in, in the end, the bus or tram and metro, it's, it's a, it's a big computer on wheels nowadays. It has an onboard computer, you download all the data.
41
00:06:44.144 --> 00:06:52.214
They would analyze to see how is my vehicle performing, uh, what can we do? How would we change the procurement, uh, strategies with that themselves?
42
00:06:52.364 --> 00:07:01.004
Instead of saying, "Let's make this data available for procurement," something that we do nowadays. And we-- when, when we started buying electrical buses, if we wanna dive into some nice examples, for example.
43
00:07:01.104 --> 00:07:11.344
When we started buying electrical buses, we didn't understand the technology. And we bought thirty buses, three, uh, ten of three different vendors, and we said, "Okay, we'll try and test, and then we'll buy a bunch."
44
00:07:12.124 --> 00:07:19.224
The way s- the way we usually do that is we, we would drive them for four or five years, and then pick the best one and order six hundred of those.
45
00:07:20.084 --> 00:07:26.864
Um, but we started analyzing the data now from the beginning through the onboard computers, specifically related to the battery, 'cause we were very worried.
46
00:07:26.984 --> 00:07:35.164
Uh, we as a company were very worried about the performance of the battery. 'Cause we had already calculated for every diesel bus that we replace, we need about one point three electrical buses.
47
00:07:36.044 --> 00:07:40.944
But that was a little bit of a guesstimate based on the statistics, uh, and, and the specs that we got from the suppliers.
48
00:07:41.464 --> 00:07:50.984
So because we had a good relationship with the strategy department, they said, "Can you do data science on these, uh, vehicles and check if this is actually true?" So can we check the performance?
49
00:07:51.004 --> 00:07:56.724
And you started seeing things like, you know, and as you probably all realize that it's snowing outside today, for the people that are listening later.
50
00:07:57.243 --> 00:08:02.604
[laughs] Uh, when it's cold, your battery performance is less than, uh, than on a warm day.
51
00:08:02.644 --> 00:08:10.743
And if you drive uphill, and Brussels has a lot of hills, you use more battery than if you drive downhill, which may have an impact on how we want to plan our lines in the future.
52
00:08:10.844 --> 00:08:18.624
Um, heat, not only cold, but also, uh, warm. 'Cause when it's too hot, the air conditioning kicks in, which also burns a lot of battery.
53
00:08:18.724 --> 00:08:28.804
Um, so there's a lot of stuff that we found next to just the basic stuff that we were looking for, like what's the, um, uh, the basic, uh, lifetime of, of a battery?
54
00:08:28.874 --> 00:08:37.984
'Cause it degrades a little bit over time, which was the main thing we were looking for. Um, and actually what changed is, to come back to my story, what changes, we changed the procurement, uh, cycle for that.
55
00:08:38.044 --> 00:08:44.904
So instead of saying we keep those thirty buses, we drive them for five, six years, and then we buy six hundred, we're now buying a hundred every year.
56
00:08:45.544 --> 00:08:55.324
Analyzing, adding that data into the pool, keep analyzing it and changing our procurement requirements, and we want this sort of warranty and this sort of guarantee, um, because we see certain things.
57
00:08:55.384 --> 00:09:06.464
So it's really getting a lot smarter. So the complete approach also changed the way the company is operating as well on the, uh, procurement side.
58
00:09:06.704 --> 00:09:18.744
Uh, so that's, that's a big achievement if I understand how you implemented that by analyzing the data and, and learning more from that than what you were initially expecting to look for, uh, in, in, in this respect.
59
00:09:19.244 --> 00:09:29.724
So, so that must have been a hard time to, to explain to procurement that you're changing the ways of, "Hey, we just tested for five years, and then we buy a bunch of six hundred of that."
60
00:09:29.784 --> 00:09:40.124
But now we do it, uh, every, uh, every, every year one hundred, if I understood. And then- But actually- Looking- Actually it was not that hard because it was the business themselves that changed the procedure.
61
00:09:40.184 --> 00:09:50.324
It was not us as IT saying, "We did this magic analysis, and you guys should work this way now." We do it, and that's one of the things that we try to do, uh, consistently, is we do the work with the business.
62
00:09:51.024 --> 00:09:58.554
So we assist where is needed. If they can do their own analytics, we check the homework. Um, but it's them that are doing the, um, the smart things.
63
00:09:58.604 --> 00:10:02.424
So everything that I'm telling you, I never said any-- to anybody, "You need to buy a hundred buses every year."
64
00:10:02.484 --> 00:10:12.364
They figured that out on their own, which is actually part of what our, our roadmap and our strategy is based on. I always say we're, we're not a data-driven company, we're an insight-driven company.
65
00:10:12.424 --> 00:10:20.444
'Cause I don't care about... I mean, I do care, but I, I don't care about the data 'cause we don't need to collect a bunch of data. That's not the point. The point is to do something with it.
66
00:10:20.904 --> 00:10:29.344
And it's also one of the jokes that we keep making ear- uh, early on was when we started doing data science and AI and advanced analytics, everything you wanna call it.
67
00:10:29.384 --> 00:10:36.544
It's been a complete failure if you look at it from the IT department, 'cause we have not put anything into production or hardly anything into production.
68
00:10:37.384 --> 00:10:45.564
Um, but we learned so much because a lot of the stuff was just let's do analytics, let's do exploration. We find something. Ah, there's a procedure that's broken.
69
00:10:45.904 --> 00:10:56.144
Let's fix the procedure, and now we don't need, you know, I don't need a learning model that keeps telling me that now it's working fine. Then we do another exploration analysis, and we, we really did early on.
70
00:10:56.304 --> 00:11:06.264
We were, we're a public company, so we're using tax money. We're trying to be very [laughs] very conservative in, in how we spend that. I don't need to build a Google. I don't need to build a Facebook.
71
00:11:06.644 --> 00:11:13.083
We do a sm- a small, um, a small test, maybe two, uh, two, three weeks.
72
00:11:13.624 --> 00:11:24.424
We check that it works and then if we want to continue, we do another two, three weeks and, uh, uh, and based on that, lots of times what we learn is as you do two, three, four cycles, you're actually done.
73
00:11:24.464 --> 00:11:32.964
You solved your problem, and you don't need to put anything in production technically. So you have another approach where you say this is success by not...
74
00:11:33.164 --> 00:11:39.624
Typically what I, what I've heard about a lot of companies, they have their data science teams. They are in the explorative way.
75
00:11:39.684 --> 00:11:48.344
They, uh, can experiment for two years, and then management comes up and say, "Hey guys, it's time to see some results. Uh, you have to put something in production."
76
00:11:48.854 --> 00:11:54.864
And that's why we call eighty percent of the, of the data science projects, they do fail because nothing is into production.
77
00:11:55.164 --> 00:12:09.116
But you look at it a, at a, at a-From a different perspective where you say, "We've, we've learned the insights from the exploration, what we've did, and we did communicate that with business, so they get the insights from the exploration and the experimenting.
78
00:12:09.136 --> 00:12:18.336
And this is what we just did at, at that time. We don't need it in a sustainable way, in a consistent way. We just need it to solve that, uh, particular problem."
79
00:12:18.816 --> 00:12:24.936
And it's more in the win- This is specifically interesting when you're, when you're looking at data, because what, what-- approach it-- you have two approaches to doing data.
80
00:12:24.996 --> 00:12:38.356
You have the approach where, where the Facebook and the Googles go, where you have a lot of generated data, you let your AI look at that, and it magically finds spots of gold everywhere, which is a very expensive way to work, and it works very well on machine-generated data.
81
00:12:38.396 --> 00:12:47.276
The data at the STIB is mostly human-generated. Now, there is some vehicle data and stuff like that, but a lot of it is interactions. It's very difficult to do it that way. So what we do is we work with hypotheses.
82
00:12:48.316 --> 00:12:59.066
Um, to give you an idea what we usually do when we go work for a business or a new business unit, we say like, "Give us some of the data and give us like ten days to just analyze that first set, and don't tell us anything."
83
00:12:59.916 --> 00:13:05.716
Um, and we tell them early on, "And we're gonna come back, and you're gonna make fun of us because we're gonna find a whole bunch of stuff that you already know."
84
00:13:05.726 --> 00:13:16.456
[laughs] But that's really good because that means we actually-- we're actually reading things, uh, that, that you already know, and that means that we can ask the data questions and it's, it's, uh, it gives a little bit of confidence to the business that we can actually do something with this data.
85
00:13:17.336 --> 00:13:19.816
Um, so early on, they make a lot of fun of us.
86
00:13:19.856 --> 00:13:30.816
Uh, we, we did-- When we started with data science, and we were naive like everybody else, we said, "We're gonna find, uh, a link between, uh, the wear and tear of brake pads and the number of times that the, the driver pushes on the brake."
87
00:13:32.076 --> 00:13:41.216
Uh, seems obvious that you can get all that data. We looked and we couldn't find any link between that [laughs] strangely enough. [laughs] Uh, we went to the business and said, "That's really weird.
88
00:13:41.276 --> 00:13:49.856
Like you sh- we should be able to see, you know, the, the wear and tear." And they go like, "No, because, you know, it's scheduled to be every like three months." I don't know how much time there, uh, is in between.
89
00:13:49.896 --> 00:14:02.836
"Every three months, we replace all the brake pads because there's so much pressure on our, uh, uh, on our network, and we need all the vehicles, so we don't have time to like check that maybe it would, you know, uh, maybe it can run for another month or something.
90
00:14:02.896 --> 00:14:11.556
We need them to run all the time, so we, we accept that risk." Um, so we found a lot of stuff like that where we say, "Can we experiment with going maybe to four months?"
91
00:14:12.276 --> 00:14:22.876
Uh, I think there's, there was a lot of discussions like that, but that's actually how we got to most of our stuff, because then we started seeing weird little signals in the data where we said like, "That's strange."
92
00:14:22.896 --> 00:14:34.376
Like, you know, all the time this happens, and then the business like, "We don't know either why that would happen." And then you go start doing a hypothesis. We think that drivers are doing this, and that's causing it.
93
00:14:34.446 --> 00:14:40.496
And then you try to look for data sets that help you support that, um, [lip smacks] uh, that hypothesis.
94
00:14:40.616 --> 00:14:56.236
We found one early on that we, we talk about a lot, where we saw that, uh, if you have over-revving on your engine, if you hit your, uh, your gas too hard and the engine makes noise, uh, we saw a lot of over-revving, and with the over-revving, uh, those buses would break down, which is really nice in your scatter plot.
95
00:14:56.316 --> 00:15:00.296
You could see like all, uh, uh, over, over-revving always the, always a problem.
96
00:15:00.836 --> 00:15:10.536
So we started early on with a-analysis, like it's probably the driver that is, you know, having a bad behavior, um, which is not something we usually obviously say. We have very nice drivers.
97
00:15:11.096 --> 00:15:18.516
[laughs] Uh, um, so we started looking at the different things, and you start saying, "Actually, it's not that because the, the, you know, it's different drivers, and it's at different times."
98
00:15:19.296 --> 00:15:23.176
Um, to make a long story short, uh, we started adding a whole bunch of data in there.
99
00:15:23.316 --> 00:15:31.056
So we started adding, uh, the, the timetable stuff, what, what lines they were driving, how many people were in the vehicle, because you can see all those different things based on different sets.
100
00:15:31.476 --> 00:15:39.796
But eventually, what we found is it's actually not a driver that is-- has a bad driving style. It was, uh, vehicles. Vehicle-- We have four zones in Brussels, uh- Mm-hmm...
101
00:15:39.806 --> 00:15:49.816
uh, for, for STIB, and one zone for some reason has a lot of very popular routes, and they go uphill. So the over-revving was not the fact that, um, you were-- you had bad behavior.
102
00:15:49.856 --> 00:15:54.696
You were driving uphill with a full double, uh, bus, which was really pushing the bus to its maximum.
103
00:15:54.736 --> 00:16:02.356
Now, one of the funny things at STIB, uh, and it's just a funny type business processes, we assign the buses for life into one of those zones.
104
00:16:03.056 --> 00:16:14.486
So the buses that were breaking down were basically [laughs], you know, out of that department constantly being tortured uphill, and then obviously over their lifetime, they would break down more than the other department-- than, than the other zones.
105
00:16:15.396 --> 00:16:22.486
So there we, uh, they, they changed the business process a little bit where they rotate. I think they rotate the engine, uh, out now after a couple of years.
106
00:16:22.516 --> 00:16:30.556
So you actually distribute the load a little bit over the different vehicles. Uh, but then to come back to the fact we don't put anything in production, you know, we figured it out.
107
00:16:30.636 --> 00:16:44.116
We don't need to put that model or that analysis into, into a production system. Yeah. Fu-fully, fully understand. Uh, I was just thinking of this is specifically for Brussels.
108
00:16:44.256 --> 00:16:55.536
Do you interact as well with other public sectors for, uh, public, uh, transport? Because I hear sometimes different stories where they run very old material, old buses.
109
00:16:55.896 --> 00:17:08.596
So they could benefit of, of your approach on looking at the data, the analytics to optimize as well, um, the complete infrastructure, what they're running. Mm-hmm. Well, um, it, it depends how, how you ask the question.
110
00:17:08.636 --> 00:17:14.476
So if you're talking in Brussels, do we talk with the other, uh, public services? Yes, we do.
111
00:17:14.616 --> 00:17:22.836
Um, [lip smacks] mainly because our focus up to now or up to a couple of years ago has been very internal, but we realized that we're gonna need a lot more context.
112
00:17:23.636 --> 00:17:30.316
Um, and Brussels, as you probably know, it's politically quite fragmented. So for example, we are the public transport operator.
113
00:17:30.376 --> 00:17:37.736
You have Brussels Mobility, which is the transport authority, which plans how we do, um, different things. Uh, there's different departments.
114
00:17:37.776 --> 00:17:48.816
The IT department for Brussels is under a different, uh, political, uh, group or a political leader. Uh, so yeah, we do coordinate a lot. One of the actions that we took, because we are one of the larger--
115
00:17:49.796 --> 00:17:59.440
one of the largest IT purchasers within, uh, Brussels, so we have, um, quite some influence. So we started a group called Data Moves Brussels-Where we bring together,
116
00:18:00.420 --> 00:18:08.360
uh, different departments like, uh, for example, Brussels Environnement who measures the air quality and the weather. Uh, for example, Brussels Parking who arranges parking.
117
00:18:08.870 --> 00:18:16.640
Um, we, we have some very nice, uh, cross, um, cross-selling type projects that we could use, and we, we work a lot...
118
00:18:16.700 --> 00:18:24.700
Uh, we look-- doing something right now with the parking, for example, uh, because lots of times people park in the wrong place, and they block the tram lane, uh, tram lines.
119
00:18:25.160 --> 00:18:37.440
And we found a, an AI company that has a camera that you can put in your tram that can automatically scan, uh, with an AMPR camera, uh, read the license plate and automatically do a warning or a, uh, a fine.
120
00:18:37.520 --> 00:18:47.640
Uh, so we're looking at... It's very embryo, uh, very, uh, early days on that, but we're looking like can we do something like that 'cause it would help both of us. Um, if you're talking-- That's what...
121
00:18:47.680 --> 00:18:59.480
If you're talking Brussels, there we, we really try to, uh, use our, uh, use our superpower responsibly, a-and help the others. [chuckles] If we bring, we bring purchasing, and we, we have a lot of volume.
122
00:18:59.920 --> 00:19:05.420
Uh, we try to bring the others with us. Uh, we have Paradigm, which is the central ID in Brussels that we work with a lot.
123
00:19:06.140 --> 00:19:14.480
Um, and we try to do the same thing there to really say, "Okay, we will bring, uh, our volume and our knowledge, uh, for the others to, to enjoy that as well."
124
00:19:14.980 --> 00:19:20.280
Um, you were talking more about the, the other, uh, public transport operators, and th-those are not in Brussels.
125
00:19:20.940 --> 00:19:31.080
Um, obviously we're in touch with them as well, and there's a lot of initiatives there to, uh, work together with data. I don't know if you know the, the Brussels-- uh, the Belgian Mobility Company, BMC. Yeah.
126
00:19:31.180 --> 00:19:36.280
Uh, it's, uh, it, it owns the MOBIB card, which is the, like, the Oyster card, but for, for Belgium.
127
00:19:36.440 --> 00:19:44.239
Um, we actually bought with the four public transport operator in Brussels, uh, a small consulting company that focused in public transport data.
128
00:19:44.620 --> 00:19:55.520
So they're now helping us harmonize the different data flows that we have for timetables and real time into one, uh, central API. Because right now that was still a problem.
129
00:19:55.540 --> 00:20:01.780
You can travel in Brussels, you can travel in Flanders, you can travel in Wallonia, you can take the train all over Belgium, uh, but they're not connected.
130
00:20:01.840 --> 00:20:12.559
So we're, we're, we're actively trying to fix or untangle a little bit the spaghetti that, uh, that is there politically. Um, but yeah, we, we talk with them, we see how they work, but the...
131
00:20:12.660 --> 00:20:22.839
it's, it's very difficult to compare. Yeah. In, in one sense, the business model is different 'cause we are, um, we're an urban operator, so our, our, our-- the population density is very close.
132
00:20:23.420 --> 00:20:31.140
It's, it's, it's a lot simpler for us to be on time or to, to give a, um, to give the idea that the service that is really, really available.
133
00:20:31.190 --> 00:20:35.500
'Cause if one of the vehicles is late on a stop, it doesn't really matter as long as there's a vehicle every five minutes.
134
00:20:35.540 --> 00:20:46.700
Where if you're, uh, De Lijn in Flanders, for example, and, and we're just talking, I live in Belte in the middle of nowhere, uh, in a very nice, uh, forest, but here I have a bus once an hour, so if that one is late, I'm worried.
135
00:20:46.740 --> 00:20:55.839
And if it's not there, it's a big problem 'cause it's another hour. So the, the, the challenges are really different. Um, and also funding-wise, it's very different. In Brussels, we, we are very lucky.
136
00:20:55.880 --> 00:21:00.440
We, we have a very, um, a very respected, very loved brand within Brussels.
137
00:21:00.520 --> 00:21:13.680
Our political, uh, um, parties are very, very happy with the work we do and are willing to invest in mobility, which is different in the other parts of the country or, or nationally. Yeah.
138
00:21:13.740 --> 00:21:23.720
That's a, a lot of exploration what you, what you do with the, with the data. How do you align that a bit more with, with the data governance and the standard BI reporting tasks in the traditional way?
139
00:21:24.330 --> 00:21:34.600
Well, we- You said you have a central data platform. That's one of the achievements and where everybody keeps their liberty to, to do whatever they, uh, wanna do, have access to the data.
140
00:21:34.680 --> 00:21:46.060
One of the principles of data literacy, accessibility to the data. So how do you balance out the explorative part and more the, the, the standard way of, uh, the needed reporting?
141
00:21:47.240 --> 00:21:57.860
We, we, um, early on, what we did is we had two sets of reports. We had exactly that. We had what we call standardized reporting, which were completely under our control. Uh, we built them, we put them in production.
142
00:21:57.960 --> 00:22:05.800
The, the questions went to us. Obviously, they're done with the business who, who formulate the, um, definitions, but we would be responsible for that.
143
00:22:05.860 --> 00:22:13.500
And we had what we called user community, which was the same data flows, but they could build whatever they wanted, and it was unsupported.
144
00:22:14.420 --> 00:22:23.880
Um, we would help them obviously, and then if they had something that they wanted, and they really wanted to, like, now we're gonna spread it throughout the organization, we would make it into a standardized report.
145
00:22:24.680 --> 00:22:34.500
Um, that was before when we just had SAP. Now, with the data science and the, the data lake, uh, coming up in Azure, we're giving them a lot more freedom in Azure.
146
00:22:34.560 --> 00:22:44.470
So on the data lake side, we're saying we are the guardian of the, the raw data layer. Nobody can touch that. That has to stay immutable. But the layers on top, we can discuss.
147
00:22:44.560 --> 00:22:53.700
Either you can build your own stuff, uh, if they know how to do that, or we build some, um, some data marts or some small data lakes for you where you can experiment with everything you want.
148
00:22:54.100 --> 00:23:04.700
And later, we can validate whether this makes sense or not. We use SAP mostly now for the standardized type reporting that goes to our-- that go to our top management. Right. So we really split. If you're looking at...
149
00:23:05.040 --> 00:23:13.900
I don't have any slides to show you, but we have a really nice slide, uh, where we say on the left side, we have really... We, we actually we need to get back to the story because then I can explain it better.
150
00:23:14.180 --> 00:23:20.060
We created personas. So we have four personas, uh, for, uh, analytics and data use.
151
00:23:20.100 --> 00:23:30.920
The, the top one is really, like, the CEO, the-- our CEO and our, our top management, uh, that really just wants structured reporting and, you know, what's my- Yeah... what's my P&L look like?
152
00:23:30.960 --> 00:23:41.740
You know, the, the basic stuff. Um, you have a sort of middle manager, um, that, that works for the-- He, he works for the dispatcher. We call him Hicham. He works for the dispatch. He has some people working for him.
153
00:23:41.800 --> 00:23:52.020
You have, uh, what we call Alice, which is somebody that works in the dispatch as a dispatcher. She really dives into the data. And then you have Carine, which is like a data scientist that needs complete freedom.
154
00:23:52.560 --> 00:23:59.840
Uh, and from those, we created, um, uh, a nice overview where we say, "Okay, you have on the left side those top management profiles.
155
00:23:59.900 --> 00:24:07.820
They just use SAP."On the other side of the spectrum, on the right, you have that data scientist, and that's... It's a persona for a data scientist.
156
00:24:07.960 --> 00:24:21.040
So it can be somebody that is really, uh, into, uh, statistics and Excel to the, the triple PhD in astrophysics. Um, they get complete freedom in the, um, uh, in the data lake. Yes. Up to some limits.
157
00:24:21.060 --> 00:24:29.920
They can't spend all the money in one day. [laughs] And then in the middle, you have self-service, and we s- we split that into self-service visualization and self-service BI. What's the difference?
158
00:24:29.950 --> 00:24:37.900
That self-service visualization is for managers like me. I need to go to the, the management. I need to show some data. I wanna have my, uh, pie chart in a column chart.
159
00:24:38.240 --> 00:24:50.300
Okay, I can change that stuff in Power BI or Analytics Cloud myself. Uh, visualization is really for that dispatcher. She needs to dive into the data, find the problem, why is the metro, uh, running late the last time.
160
00:24:50.340 --> 00:25:00.600
We don't really know why, what's going on. She really understands her model, so she has a little bit more freedom, uh, or she has maybe prepared data, but she has more freedom to dive in and, and find a new problem.
161
00:25:00.620 --> 00:25:05.620
Where the data scientist all the way on the right, yeah, you just open up everything for them, and they can really explore.
162
00:25:06.940 --> 00:25:17.310
So, so this approach of the data governance, where you control the data lake, the raw data, where you say, "We secure that this is in a certain format that it's consumable for everybody."
163
00:25:17.880 --> 00:25:19.960
You defined the personas as well.
164
00:25:20.440 --> 00:25:34.400
Did that help on becoming more data-focused, but you'd try-- well, you prefer to use the insight-focused [chuckles] approach, which, which I tend to believe more in than a purely data-focused because that's only the raw material.
165
00:25:34.460 --> 00:25:42.820
But the insight focus that people understand, hey, we do have the data, and we can do something with it, and this is the type of insights we can get from that.
166
00:25:42.940 --> 00:25:55.360
How has that, uh, changed that, that, that culture over time, and how did you achieve that? It's, it's an ongoing discussion. I don't think we have achieved it yet. Um, but it's also, I think, a sign of the times.
167
00:25:55.420 --> 00:26:00.980
Everybody is interested in data now. Uh, before, I was talking to somebody else who was laughing, like, before they left you alone.
168
00:26:01.020 --> 00:26:08.360
It was kind of this little backwater, and sometimes you came out with some money, and they said, "Great," and go back. Now they're all there going, "Why aren't you making more money for us?"
169
00:26:08.960 --> 00:26:17.660
Um, so the pressure is definitely a little bit more on. Um, what, what we did a lot since I started as an internal is really focus on the change management.
170
00:26:17.720 --> 00:26:28.100
So we build a community within STIB, uh, of everybody that uses data, and we call it the data lab community. Um, and once a month, they come together. There's about, I think, two hundred and fifty people in that.
171
00:26:28.180 --> 00:26:39.500
They come once a month, we come together, forty, fifty people show up. Uh, we do some presentations. We do some hackathons. We, you know, they, they, they, they ask each other questions and, and, and give, uh, solutions.
172
00:26:39.560 --> 00:26:44.329
We have a little, uh, Teams chat for them where they can ask their questions and, and talk to each other.
173
00:26:44.340 --> 00:26:54.200
But really creating that community, creating that awareness that you can do something with data, and especially the cross-pollination of ideas. So we did, for example, a, a project for, um,
174
00:26:55.140 --> 00:27:05.160
the vending machines, the ticket vending machines. And then we realized, well, actually what they're doing there, you know, you have a machine, it has sensors. There's a subcontractor that is managing that.
175
00:27:05.220 --> 00:27:15.040
We manage the SLAs of that subcontractor. It's actually the same if it's a lift or an escalator or another machine. So we went with that business unit early on before we had the data lab community.
176
00:27:15.080 --> 00:27:23.480
We would do the tour of all the different departments that we knew, saying like, "Actually, an escalator, it's kind of like a, a ticket vending machine. This is what they do. Would you be interested?"
177
00:27:24.000 --> 00:27:34.820
And then we would do the same project there. Now they do that on their own. [laughs] Um, also because, you know, there's, there's examples, uh, galore these days. Um, so that's one of the things we did.
178
00:27:34.860 --> 00:27:44.320
The other thing is, um, I think governance is, is really something that we still have to work on, especially the, the official like data owner, uh, data steward type stuff.
179
00:27:44.780 --> 00:27:48.540
We didn't start there, but it was a conscious choice because when...
180
00:27:48.560 --> 00:27:54.220
Right after I started, they, you know, they, when I started, they presented me this really nice thing, and there's all this budget, and it was gonna be great.
181
00:27:54.250 --> 00:28:01.810
And then six months later, they said, "Ah, actually, this data thing, we don't really believe in it. We're gonna cut your budget by fifty percent." Um, so [laughs] we- [laughs]...
182
00:28:01.810 --> 00:28:10.140
we kind of went into survival mode, which actually was a hidden present. Um, but, um, we went into survival mode and said, "Let's show value quickly."
183
00:28:10.240 --> 00:28:13.540
And, uh, you know, I had a company before, so we really went into selling mode.
184
00:28:14.120 --> 00:28:23.420
Uh, and we started focusing on things like the ticket vending machines because we figured revenue, uh, assurance is probably a good thing, and we can show how many euros we're saving, [laughs] uh, which is there.
185
00:28:23.460 --> 00:28:26.480
We did invoice flows, a whole bunch of stuff we can talk about later if you want.
186
00:28:27.060 --> 00:28:35.680
Um, but we specifically said, "Let's not go to the business and say, 'First, you need to organize yourself, and then maybe we can deliver some value.' Let's do it the other way around.
187
00:28:35.780 --> 00:28:48.000
Let's deliver value, realizing that it's not so secure, uh, or, you know, it's not so well-governed. Um, but we can flag that to the business saying like, 'Look how valuable this is.
188
00:28:48.140 --> 00:28:50.500
Wouldn't it make sense that you make a small investment from your side?'"
189
00:28:50.940 --> 00:29:04.340
And we actually met each other in the middle because before I got the time to say wouldn't we, we already got people coming back saying like, "Ah, you know, this, uh, we, we did do a very nice set of dashboards for, uh, operations at the bus terminal and the metro, and they run out.
190
00:29:04.360 --> 00:29:13.380
They're managing the fleet. It's about, I think, a hundred and fifty people that use that on a daily basis to manage the fleet. Um, they came back to us and said, like, "Don't be scared.
191
00:29:13.480 --> 00:29:21.280
We wanna do an internal audit because, you know, we don't understand how the formula works and who can change stuff. So we're not at all angry with you, but we really wanna secure the things."
192
00:29:21.300 --> 00:29:28.260
We're like, "Okay, well, [laughs] you know, welcome. That's actually what we wanted to do, so I'm glad that the question comes from your side." Um, that's one department.
193
00:29:28.460 --> 00:29:32.500
Then there's other departments that are saying, you know, the data quality is really bad and IT should fix it all.
194
00:29:33.020 --> 00:29:37.400
Um, there's other ones that are hiring external firms to come tell them the same thing that I told them already.
195
00:29:38.000 --> 00:29:48.600
Um, it's a very big depart- It's a very big company, so what I say is correct for some pieces but not for others. Um, but, uh, yeah, yeah, that's, that's basically how we got there.
196
00:29:48.640 --> 00:29:55.260
But I think governance for me early on was really not my, my main focus. Yeah, I understand.
197
00:29:55.320 --> 00:30:05.020
And a-also the different approaches, that's intriguing, where you allow them to exist, not trying to go for one standard way where you say, "Yes, IT needs to fix everything."
198
00:30:05.440 --> 00:30:16.920
For one department that works, for the other one, you're trying toBring them in because it's not only, uh, IT's problem, it's, it's, uh, a company-wide, uh, responsibility what you have to drive for, for data quality.
199
00:30:17.060 --> 00:30:26.860
Yeah. And it's one of the anal-- Uh, a-actually, we, we do analytics on our own data as well. So when we looked at the budget data early on, we saw fifty percent, when I started, fifty percent was report creation.
200
00:30:27.300 --> 00:30:37.600
And we realized we were doing only, uh, data analytics for a really small sliver of the company. So we said, [laughs] "If we're gonna do this for the entire company in this way, our budget is going to go exponentially."
201
00:30:38.060 --> 00:30:46.960
Which is, which is not-- I mean, which would be okay for me. We just hire more people. But, you know, it's, um, it's not something politically or, or organizational-wise that we wanted to do.
202
00:30:47.020 --> 00:30:54.400
So we said, "We really have to focus on this self-service. We really have to invest there if we wanna keep growing." The growing makes sense on the platform team side.
203
00:30:54.900 --> 00:31:04.270
It doesn't necessarily make sense on the report-building side. So for me, in the beginning, we, we joked about it, it's all people that are secretly working for us. Uh, those- Mm...
204
00:31:04.280 --> 00:31:11.270
two hundred people in the community, actually some way or another, they're, you know, working together. Let's not say they're working for me. Uh, we're, we're working together.
205
00:31:11.550 --> 00:31:13.800
[laughs] Data is a team sport, so it's not right.
206
00:31:13.810 --> 00:31:31.220
Uh, [laughs] so, so that-that's the way we really try to push that, and it's also nice because we're seeing a lot of times one of those or some of those people from that community go to their boss and say, "Can you go talk to the data team 'cause we're doing really nice stuff, and I wanna do something where I need this other department to help me.
207
00:31:31.280 --> 00:31:40.840
Could we make this into a more official part?" So they're really, uh, advocating for us. Yes. So did that help your background as a, as an entrepreneur?
208
00:31:40.940 --> 00:31:52.560
Or hear some, some things coming back where you say, um, it's our money, so you're very responsible on, upon the budget and what you can achieve, showing value. Typically,
209
00:31:53.500 --> 00:32:04.160
what you would expect a different, uh, attitude in a, in a public company where people are not so, um, mindful of the resources they're, they're using in, in such a way.
210
00:32:05.460 --> 00:32:12.360
I, I, I don't know if that's correct, if people are not so mindful of... I think we all are very, uh, conscious that we're using public money. Um,
211
00:32:13.580 --> 00:32:17.980
I join you in concluding that there are definitely efficiency gains to be, to be had.
212
00:32:18.600 --> 00:32:28.940
Um, I think I'm a bit of an atypical profile for, for a public company, um, in, in the sense that I, you know, I had a consulting company before. We had about seventy people that worked for us.
213
00:32:29.000 --> 00:32:39.840
When I started at STIB, I had about seventeen, eighteen people. So for me, it was really just-- it, it's one big, uh, one big fixed price project, and I have to spend all the money- [laughs]... at the end of the year.
214
00:32:40.129 --> 00:32:47.500
And usually, I try to keep some [laughs] when I was, uh, when I was still in the private sector. So in that sense, it really helped to make investments.
215
00:32:47.530 --> 00:32:54.700
So if you had a little contingency left and right, you can invest in continuous improvement, and that's really what we did early on. We said we really have to run this.
216
00:32:54.760 --> 00:33:00.560
When, when they cut our budget, we said, "Let's run this as a small internal consulting firm, and we're gonna sell our services."
217
00:33:00.620 --> 00:33:07.840
So I spent a lot of time early on really selling, just doing account management, going to talk and say, "Well, we could do this. We could do that. Would you be interested?
218
00:33:07.880 --> 00:33:16.980
And I can get you a little deal because I'm doing something for this department, and if you invest it, then this- Mm-hmm... I will tack it on to this other project, and we'll deliver this huge value for you."
219
00:33:17.000 --> 00:33:29.220
And, you know. So it was really selling. I do that a lot less now. Um, so I think there, there's definitely something there. And also what I think I did differently from, from others is I went outside a lot.
220
00:33:29.280 --> 00:33:38.510
So I went to, uh, the DECA, the, the Wallonian company, the transport company. I went to De Lijn. I went to NMBS, all the different ones in Belgium, the transport companies, and we said, "What are you guys doing?"
221
00:33:39.140 --> 00:33:45.030
Especially when we started doing data science early on. Like, "What did you [laughs] guys already analyze? What worked and what didn't work?"
222
00:33:45.100 --> 00:33:55.340
And pragmatically, I picked all the stuff, uh, for what they have already experimented with and said, "There, we found some, some gains. Okay, we can do the same thing. We probably will find, uh, the same results."
223
00:33:55.380 --> 00:34:05.380
And they always say, "Fail fast." They always say, "Fail fast, but don't fail first. Have a couple of wins and then start failing." [laughs] Um, and it's especially true in, in, in public sector.
224
00:34:05.460 --> 00:34:12.280
Like, like you're saying, it's, it's company money. It's, uh, it's community money. I don't think we should innovate with that.
225
00:34:12.900 --> 00:34:22.760
Innovation-- I mean, it don- doesn't mean that you can't do innovative things, but I don't think we should innovate in the pure sense because it's very difficult to explain to the public that we used this money and it didn't bring anything.
226
00:34:23.180 --> 00:34:27.240
We learned a couple of things, but, you know. So what we try to do is be a fast follower.
227
00:34:27.300 --> 00:34:37.980
So when, when we innovate, and that's why I try to correct myself there, we do innovate, but we innovate by taking proven technologies and putting them together in an innovative way instead of trying to invent a new technology.
228
00:34:38.080 --> 00:34:48.420
To give you an example of that, um, we have overhead lines in the tram, those electrical lines that run on top of the tram. Um, they're very difficult to monitor. We want to monitor them.
229
00:34:48.460 --> 00:34:58.120
There's a way to monitor them, which is, uh, you have to hire an external firm. They put, uh, an extra antenna, a pantograph, they call that, on top of the tram that has sensors on it.
230
00:34:58.160 --> 00:35:09.799
You cannot really measure directly because it's all high voltage. So if you touch it, you're dead, but your sensor also burns off. So it's very difficult to do. Um, that sensor is that, uh, sensor box that they have.
231
00:35:10.240 --> 00:35:16.680
Uh, it's also very sensitive, so you have to drive very slow, uh, to do that. You can only do it at night with a special tram, or you have it on top.
232
00:35:17.280 --> 00:35:20.799
Uh, long story short, costs about two hundred and fifty thousand euro every time we do it.
233
00:35:21.400 --> 00:35:28.940
Because it's so expensive, we only do it every, like, four or five years, which is not really great if you wanna [laughs] have an updated view of the status of your network.
234
00:35:29.000 --> 00:35:39.140
It's metal, hot, cold, you know, that, that has an impact on the tension. Um, so already early on when I started, that department came to us and said, "Can we use video and do something?"
235
00:35:39.180 --> 00:35:44.700
And, you know, we're not ready for that right now. A couple of years ago, we said, "Okay, now we are ready. If you want, we can do it."
236
00:35:45.150 --> 00:35:54.220
So the nice thing was they already had those tests from, from the previous years, so they had video and they had sensor readings. Then we said, uh, again, coming back to my story, we made a hypothesis.
237
00:35:54.280 --> 00:36:03.960
We said we can, with machine vision, deduct certain variables. Like there's some obvious ones like the, the pantograph has to be, or the line has to be in the middle of the pantograph.
238
00:36:04.060 --> 00:36:11.460
Uh, there cannot be electrical sparks because then the distance is too high. So we had like five or six that, that we did, and we used, um,
239
00:36:12.360 --> 00:36:22.172
the video to see can we, can we deduct it?Um, long story short, we can do it, uh, then [laughs] based on the video. Then we said, "Okay, now we can do the next step, huh?"
240
00:36:22.552 --> 00:36:35.072
So the next step is we build our own little sensor box. Really dumb, like weatherproof box, Raspberry Pi, high definition camera, infrared light, G-GPS unit. We put it on top of a test tram, uh, where the, the...
241
00:36:35.092 --> 00:36:38.872
No, it's not a test tram. It's, uh, where they learn how to drive, huh, th-those type of trams.
242
00:36:38.912 --> 00:36:45.912
We have two of those, uh, 'cause you cannot put a Raspberry Pi on top of, uh, a tram with actual people on it due to insurance reasons.
243
00:36:46.032 --> 00:36:51.182
[laughs] There are some regulations there, uh, that, that you have to pay attention to. Um,
244
00:36:52.312 --> 00:37:06.412
and with those two, uh, trams actually just driving around, uh, training, uh, new drivers once a month, we already had an updated, uh, version of the network. Um, now we're expanding that into, um, into ten trams.
245
00:37:06.472 --> 00:37:16.772
If we have ten trams where we have that, we're still in the process right now of certifying the sensor box. Um, we actually can do the entire network once a week. And it's really cool 'cause it's a camera.
246
00:37:16.832 --> 00:37:27.412
It's pointing upward. We talked to the DPO for all the people that are gonna... worried about how we, uh, handle video in public spaces. Was the main part of the project, [laughs] um, but it-it's all, uh, covered.
247
00:37:27.452 --> 00:37:36.232
We, we anonymize and all that, but we basically look, and they, they, they capture the data. When they-- when you see the anomaly, they take five minutes before, five minutes after, it makes a video.
248
00:37:36.302 --> 00:37:44.912
It goes onto a portal in the cloud. Um, it's all in the cloud, by the way, huh. It's encrypted and everything, but, uh, the, the video's in the cloud. We do machine vision on it, and it creates a little portal
249
00:37:45.832 --> 00:37:54.432
where the, uh, the, um, um, the, the technical guys can just click on it and see what the problem is. Before they had to drive there every time, wait for a tram to drive by, see what the problem was.
250
00:37:54.912 --> 00:38:01.492
Now they can do that automatically. They can schedule urgent, non-urgent, um, and we can save because we anonymize everything.
251
00:38:01.572 --> 00:38:11.392
We can save the videos, and we can train the model to also pick up, um, other anomalies, which we're gonna do later. We first wanna implement the whole thing. But there, we've taken a lot of very,
252
00:38:12.612 --> 00:38:23.972
um, yeah, very proven technology. It's, it's a machine vision, um, it's a facial recognition, uh, algorithm that we use for, for, for the machine vision. It runs in the, in the Azure cloud. It's just basic.
253
00:38:24.032 --> 00:38:33.532
We use the, the face blurring. It's just basic Microsoft algorithm, Raspberry Pi camera. But the way we put it together was very innovative. Yeah.
254
00:38:33.632 --> 00:38:53.892
So fully understand that you say your innovation is, is in a certain way where you create new things, but not really some thought-provoking new initiative, uh, what you're developing like a Google or a Microsoft is, is developing, uh, with venture capital, uh, because you're using the public, uh, capital.
255
00:38:53.932 --> 00:39:01.572
No, but that's still very innovative. I mean, trying to, to use whatever already exists is proven and then show value out of that.
256
00:39:01.932 --> 00:39:11.072
So in fact, you already created your digital twin from the infrastructure, from the lines, uh, for the STIB and MIVB, if I understand. It's, it's one step.
257
00:39:11.152 --> 00:39:28.132
We, we're actually working on digital twin, and it's one of the things where we're using the, the power of the region, uh, to really see can we do it together 'cause there's a lot of, uh, work where, for example, uh, Brussels, if you, if you don't know this, uh, there's a thing called Urbis in Brussels, which has the entire outside of the buildings are, are available.
258
00:39:28.152 --> 00:39:31.792
And they did really cool things with it, like putting it into, uh, the Unreal Engine.
259
00:39:31.852 --> 00:39:40.772
So now if you build an apartment building in Brussels, you can run the, the whole, uh, cycle of the sun and stuff like that, see where the shadow falls, but they only have it for the outside.
260
00:39:41.692 --> 00:39:46.412
So we were saying, like, we would be interested in inside, for example, like the tunnels and maintenance for that.
261
00:39:46.792 --> 00:39:53.512
So there's a initiative going on together with Paradigm where we have LiDAR backpacks, and they run through our infrastructure, and we map it that way.
262
00:39:53.902 --> 00:39:58.992
'Cause also for us, that gives extra value if you're just talking digital twin and, and 3D visualization, for example.
263
00:39:59.472 --> 00:40:08.212
Now that we're building the new metro to show, um, the, uh, the different people that are going to be impacted with shops, what, what works will be, will be done in their area.
264
00:40:08.272 --> 00:40:14.992
We can give them VR helmets and say like, "Okay, your shop will be here, and, you know, phase one will be like this, and phase two will be like that."
265
00:40:15.672 --> 00:40:25.632
You can visualize the non-existent new metro station and already go like, "Okay, do you wanna buy a shop? You know, here you can stand inside of your shop. It doesn't exist yet." Um,
266
00:40:26.532 --> 00:40:31.492
other cases we have there is, um, we, we work a lot of data is for people with disabilities.
267
00:40:31.552 --> 00:40:43.172
If you have problems with sight, for example, that we would create a sort of digital environment for you to train you how to go to the metro, uh, at home 'cause you would have the environment available.
268
00:40:43.212 --> 00:40:55.412
That's all future stuff. We don't do that today, uh, but we're preparing for that. And we have very small, um, sets where we have it now for three, uh, stations, for example. Uh, we have that digital twin.
269
00:40:55.452 --> 00:41:01.472
We're mostly interested when you're talking digital twins, we're very interested in people flow to understand how people...
270
00:41:01.502 --> 00:41:07.792
And that's why we're trying to do it with the region 'cause for us, we see people once they pass the metro gate, but at that point, it's too late.
271
00:41:07.832 --> 00:41:13.312
If there's a lot of people coming through the metro gate, I cannot plan an extra metro. Where- Mm-hmm...
272
00:41:13.372 --> 00:41:23.892
if I can see a little bit further down, uh, or, or back in the pipe, I could see that, okay, the, the opera is finished, and there's a lot of people coming out of the opera, and I see that they're not going to the parking.
273
00:41:23.952 --> 00:41:30.852
So that still gives me time to change my, uh, change my planning. So that's why we're working together with, um, with people.
274
00:41:30.912 --> 00:41:38.752
Again, there's, uh, one of the reasons we tested that was 'cause we are aware of the privacy issues, uh, involved there. So we always say we're interested in the flow of the river.
275
00:41:38.812 --> 00:41:49.172
We're not interested in the sing- the, the individual drops. Um, it's the flow of the river that's, that's important for us when we're talking about that. The other thing, um, is like, uh, electricity.
276
00:41:49.332 --> 00:42:00.512
We're very-- We have our own electricity network, as you may know. At Metro Tram, we, we run our own network next to, uh, the, the official public one. Um, but we are linked to it.
277
00:42:00.692 --> 00:42:09.792
Uh, early on when we started thinking about electrical buses, we said, "Okay, if we're gonna charge all the buses at night, [laughs] the rest of Brussels will have to unplug their refrigerator 'cause there won't be enough electricity."
278
00:42:10.412 --> 00:42:19.296
Um, same with, uh, charging your, your, your bu- your buses duringUh, your trip right now, we've been very pragmatic with how we do our buses.
279
00:42:19.336 --> 00:42:27.226
Our, our electrical bus lines actually run on top of metro lines right now because we have the electrical substations underneath, so we can have the charging.
280
00:42:27.256 --> 00:42:36.976
But we're gonna have to talk to the region eventually about where are we going to charge. Same, and it's the same problem the lane has, that you have to electrify, but it also means your depot has to be, has to change.
281
00:42:37.016 --> 00:42:47.016
You have to be able to charge, you have to have a charging infrastructure. So we're really building all that monitoring, [lip smacks] but also, um, has an impact and then we're really going farther every year.
282
00:42:47.116 --> 00:42:56.496
It has an impact on your engineering and the way your depot, your, your maintenance, uh, facilities are built. 'Cause right now you work under the bus with internal combustion, but electricity is on top of the bus.
283
00:42:57.276 --> 00:43:03.336
So all those places that we have with the holes where you drive your vehicle over and you start working doesn't work anymore. You need to be able to get your people on top.
284
00:43:03.736 --> 00:43:12.876
So there's a lot, a lot, a lot, a lot of work there. You need to never stop. You need to train your people. Uh, you need to have electrical, uh, maybe not engineers, but at least, uh, technicians.
285
00:43:12.916 --> 00:43:22.036
At that point, you start competing with, uh, with Audi in Brussels that have also started building electrical vehicles, and you need the same people. So there is a lot of... It's not a bad thing, eh?
286
00:43:22.156 --> 00:43:33.956
It-- but it, it's a discussion- No... that is based on a regional discussions, and it's one of the things that makes it very interesting. It's a very varied [laughs] environment that you get into and that you talk to.
287
00:43:34.016 --> 00:43:46.076
Yeah. Int-interesting. I mean, it's, it's so inspiring to understand that you're really trying to collaborate with a, a lot of external parties as well and see how can we manage that best as well.
288
00:43:46.156 --> 00:43:59.016
As well on skill set, on people available, resources available, and building your own infrastructure. Rob, you're so thoughtful and knowledgeable. What, what keeps you driving in this world of data and analytics?
289
00:44:00.316 --> 00:44:07.316
I'm just really interested in the, in the whole thing. I like stuff that works together, the, the whole network thing that, that really in...
290
00:44:07.376 --> 00:44:14.916
As a child, I was, I was always really intrigued by the fact that, and now you don't notice it anymore, but I'm a little bit older, and the phone still had a wire.
291
00:44:15.016 --> 00:44:24.196
Uh, when you called America or anywhere else in the world, it's actually a copper wire that connects you physically from one place to the other. I, I found that really, really intriguing.
292
00:44:24.236 --> 00:44:30.436
Now with, with wireless phones, you don't really have that visual anymore. And it's the same with like the post office. I found that really weird also.
293
00:44:30.476 --> 00:44:37.136
Like, I put something in a box somewhere, and it shows up in another box somewhere else, and there's people that arrange all of that. So for me, data is a little bit the same.
294
00:44:38.196 --> 00:44:45.656
Something generates data, and I can look at it, I can change something else, and I can see something change. So for me, that, that's really what, what keeps me going in the whole thing.
295
00:44:45.756 --> 00:44:57.396
Uh, I'm not, I'm not really the, uh, the technical guru that understands the, the deep bits and bytes of, of all the solutions, but just the whole and how I can make something better. That's really what, what drives me.
296
00:44:57.956 --> 00:44:59.936
Yeah. I think it's a, it's a valuable skill set.
297
00:45:00.016 --> 00:45:14.176
It's the system thinking, the design thinking, and, and the holistic thinking, which is lacking very often with technical people because they're so into it and, and that's harder to understand how everything, uh, works together.
298
00:45:14.216 --> 00:45:23.936
So how do you envision the future of data management and a bit more specific, the, the role in public transportation? I think it will keep getting more and more important.
299
00:45:24.096 --> 00:45:27.076
You, you see that movement, I, I think on a European level, if you're just...
300
00:45:27.136 --> 00:45:40.516
if we're not talking globally, but just Europe, there's a lot of awareness from, from the EU that transport, maybe not just public, but also private, at the, the intersection between multi-mobility, uh, should really be a European, uh, thing.
301
00:45:40.556 --> 00:45:52.636
So there's a lot of work with like, uh, the NetEx formats, the CD formats that Europe is really pushing to say you should be able to, uh, leave, uh, your house in Brussels and go to Barcelona and do that in one contract.
302
00:45:53.136 --> 00:45:56.896
Not like you're doing now. Like at first I have to take... I have to walk a little bit. Okay, you wouldn't plan that.
303
00:45:56.936 --> 00:46:06.576
But, uh, then I have to buy a metro ticket, then maybe a train ticket to the airport, then I need the plane, and I get in Barcelona, I need to, uh, rent a car. It's all different transactions, and they're independent.
304
00:46:07.416 --> 00:46:12.656
[lip smacks] So there's a lot of movement from Europe to say, "Can we do that in one contract?" Um- Yeah...
305
00:46:12.696 --> 00:46:18.596
but that brings out a lot of very interesting questions like, say that you're, uh, you're disabled and you're in a wheelchair.
306
00:46:18.756 --> 00:46:30.696
If the Uber drops you off at the, uh, train station, is it the driver's responsibility to get you to the next fa- uh, uh, phase, or is it the receiving, uh, one or the one where you arrive that is supposed to do that?
307
00:46:30.756 --> 00:46:36.535
How does that work? What if one of the chains is late, who is responsible for paying all the other stuff?
308
00:46:36.576 --> 00:46:47.736
So it creates a whole bunch of extra questions, but I, I see a lot of, um, potential in, uh, what Europe is doing with the, with the data spaces, eh? We're involved with the mobility data space. Uh, at STIB also the...
309
00:46:47.876 --> 00:46:58.055
we have the, the AI, uh, testing and experimentation facility. We brought that to Belgium together with IMEC. Uh, there's one facility that, uh, it's really for AI to test on mobility data.
310
00:46:58.156 --> 00:47:07.656
There can be one in Europe for different, uh, there's different high-value data sets. Mobility is one of them. So we work together with IMEC, uh, with, uh, um, with Paradigm in Brussels.
311
00:47:07.796 --> 00:47:16.036
Uh, there's, there's a bunch of partners that are there, but those are the main ones, uh, to make sure that we have that facility in Europe, uh, in, uh, in Belgium, and that we, we can test with this.
312
00:47:16.136 --> 00:47:23.236
We're one of the drivers there and really try to use, like I said, our, our superpower for, uh, the benefit of everybody. Yeah.
313
00:47:23.296 --> 00:47:30.856
Now that you touch upon, uh, AI, how do you see the, the large language models, uh, shaping the, the future of data management and analytics?
314
00:47:30.996 --> 00:47:41.435
And do you see something more, um, value for, uh, large language models coming up? [lip smacks] Oh, that's the, the question of the day, eh? Um, I'm, I'm a big fan.
315
00:47:41.666 --> 00:47:51.116
[laughs] I'm a big fan for my normal work already to, uh, for, for ChatGPT and, and, and, and those things. Um, s- do I see value for data management?
316
00:47:51.216 --> 00:47:59.896
I think there, there, there might be a, a, a really nice job for, uh, for solving the data cataloging problem, to really be able to talk to your data and understand.
317
00:47:59.936 --> 00:48:08.356
The main problem will be that you need to feed it something, and from the crap stuff that we have to input them, I don't think we're gonna make them very smart.
318
00:48:08.396 --> 00:48:20.984
But, uh, I think for just understanding this, you know, why is this formula this way, and then you're sending me to a wiki, and I have to understand all the stuff that is written there to have that in a chat formatUh, I think could be really interesting.
319
00:48:21.344 --> 00:48:34.464
One of the things that we were looking at, um, to test a little bit, uh, mostly through the, um, OpenAI API, yeah, 'cause we obviously don't put sen-sensitive data into the ChatGPT front end for [laughs] people that are worried about that.
320
00:48:34.984 --> 00:48:43.943
Um, [laughs] we, we were thinking about, um, the government gives us a whole bunch of, uh, measures that we have to do, and they're, they're very detailed, and they're very broad.
321
00:48:44.064 --> 00:48:55.304
Um, and, and they're not always as easy to, to present, um, especially if you want it in a sort of standardized way where you just click a couple of buttons, and, you know, like, your top management, and me as well, like.
322
00:48:55.844 --> 00:49:00.724
Um, it, it's not so easy. So we were dreaming a little bit, and we're, we were exploring that way.
323
00:49:00.804 --> 00:49:09.434
Like, can we take those, uh, measures which are standardized but very broad, uh, and then are very deep once you want to go into the an- analytics?
324
00:49:09.504 --> 00:49:18.764
Can we do something like that, saying like for example, "What's the cleanliness of this station over this period if you took this line into consideration?" And then give you that value.
325
00:49:18.884 --> 00:49:28.544
So right now it's always a question that comes back to us or to the specific departments that have to analyze that. They prepare the data, and they send it back, and it, it's a back and forth, and it takes a long time.
326
00:49:28.944 --> 00:49:38.264
So we're thinking there's, there's some value there, uh, for us. [lip smack] We already explored the, the obvious ones, the, the, the chatbot, the chatbot for, for customer support.
327
00:49:38.304 --> 00:49:43.404
We're, we're a bit hesitant there, not necessarily for the reasons you would think, like hallucinations.
328
00:49:43.924 --> 00:49:54.404
Uh, we, we, we think there's a lot of value, um, in, uh, using it on our customer support database 'cause Brussels is very multinational. Our support database reflects that.
329
00:49:54.644 --> 00:49:58.434
There's a lot of tickets in different languages. You have to understand all the languages.
330
00:49:58.454 --> 00:50:09.074
So if you could have, like, an agent that you could ask the question, and it could just grab the Romanian answer that we gave once, uh, about, uh, where, uh, where I can buy a ticket at a certain station, it comes back to you.
331
00:50:09.304 --> 00:50:21.064
You still have the human agent to do the check. Um, that could be interesting. We, we're not so, um, h- right now actually it's, it's not something that we're really analyzing to put that available for the public.
332
00:50:21.844 --> 00:50:24.484
Um, not because we're afraid that it would give some weird answer.
333
00:50:24.544 --> 00:50:33.944
We, we are conscious of that, but the main thing is, like, if it takes off, it's gonna cost a lot of money as well, and it's something that we cannot really budget. It's consumption-based, uh, behind that.
334
00:50:33.984 --> 00:50:43.814
So if, say, we are the first that does it, and it goes into the news, and everybody starts to play with it just for fun, it's gonna cost us [laughs] a lot of money. So it's not the first thing that we wanna do.
335
00:50:43.964 --> 00:50:54.244
Just to show you we're conscious with how we're spending the money. Um, it's something that we always keep into consideration, and right now it's very difficult to box that, that budget. Yeah.
336
00:50:54.284 --> 00:51:02.284
That's, that's one of the, the, the problems that these, uh, large language, uh, model companies, uh, face for the time being because it's consumption-based.
337
00:51:02.304 --> 00:51:11.464
They're bringing a lot of value as well, uh, but the profit margin is, is very low, although that's an API call, doesn't cost, uh, cost that much.
338
00:51:11.604 --> 00:51:19.404
That's, that's what I understood from, from the challenges what they have on the business models, uh, for the time being. Mm-hmm. Um, Rob, we're coming up to an end.
339
00:51:19.644 --> 00:51:34.834
I have some, uh, rapid-fire questions for you just, uh, to challenge you and see, uh, what you would prefer. Public transport data or ride-sharing data? Uh, public transport data. Mostly because ride-sharing data is...
340
00:51:34.884 --> 00:51:41.964
We have seen it before, and it's not super, super interesting 'cause you... Their models run, run their own things.
341
00:51:42.024 --> 00:51:50.344
What is very interesting about ride-sharing data, and then I'll let you get back to your rapid fire, is about 30% of them end up at a public transport stop. [laughs] Yeah.
342
00:51:50.444 --> 00:52:04.604
Interesting to know that there is a transfer of, of type of data. And then big data or quality data? Oh, God, quality data. 'Cause also 'cause I'm a really big fan of little data. Uh. [laughs] But you said it.
343
00:52:04.704 --> 00:52:15.684
Uh, it's capturing the right data and then building the insights out of that. So real-time data or the historical data analysis? Ooh, that is a difficult one.
344
00:52:16.884 --> 00:52:26.184
Um, I would say from, from, with my STIB hat, historical data analysis, uh, 'cause we're still having some issues with real time, but I'm, uh, I'm not against real time.
345
00:52:26.244 --> 00:52:39.843
I think there's great things to be done there as well. Data integration or data segmentation? Integration. Not just 'cause I run the integration department at STIB as well, but I think [laughs]
346
00:52:39.924 --> 00:52:45.144
separation we did in the past, and I think it's, it's a cyclical thing. It will split back up at some point.
347
00:52:45.184 --> 00:52:52.774
But I think right now integration is more important, especially 'cause you're seeing, like we were talking before with the European contracts, we need more standards.
348
00:52:52.844 --> 00:53:07.824
We need more integration to really play nice together. And then customer-centric data or operational-centric data? Um, well, the right answer would be to say customer-centric, I think.
349
00:53:07.924 --> 00:53:18.804
But for STIB, uh, I, I would focus on operational-centric 'cause I think th-those are very, uh, closely linked for us. If the operations run very well, the customer will, will be happy.
350
00:53:18.924 --> 00:53:28.824
If we focus only on the customer because we're a public service, for me, it's a bit of a paradox. You can't over-customize. The, the point is to have a centralized service, right?
351
00:53:28.864 --> 00:53:41.343
And I've got some more, uh, personal questions. Are you a morning person or a night owl? Definitely a morning person. I, uh, I get up really early, and I do my reading and my meditation, uh, early on, so.
352
00:53:41.424 --> 00:53:52.244
But it's also 'cause I have small kids that keep me up. [laughs] So I'm really tired at the end of the day. I used to be a night person, but that shifted. [laughs] It's, it's a practical, uh, observation.
353
00:53:53.004 --> 00:54:00.664
Uh, do you love cooking at home or dining out? Uh, well, we were talking about it before. I used to live in the center of Brussels, and we dined out a lot.
354
00:54:00.824 --> 00:54:13.824
Then we had kids, and we moved to, uh, the beautiful countryside, so now it's more cooking at home. And are you more fond of the mystery of space or the secrets of the deep sea? The mysteries of space for me.
355
00:54:15.804 --> 00:54:28.136
And if you w- could choose, do you prefer mind-reading ability or time manipulation power? Oh, God, that's a really good oneI don't think I wanna do mind reading.
356
00:54:28.196 --> 00:54:39.386
I'm very conscious of what other people think of me and others, so I don't think I wanna know. I'd like to live in my own reality, so I'll take the time warping. [laughs] But last but not least, those doors.
357
00:54:39.406 --> 00:54:49.576
Rob, uh, it's all right. As we wrap up, you know, data analytics, it connects us all, but there is, uh, some, uh, major power as well called music what connects us, uh, as well.
358
00:54:50.136 --> 00:55:00.376
What is your favorite band or type of music? I'm not as into music as you. I was listening to your podcast, and I heard you were a DJ before. Uh, but for me, uh, it's, it's Pearl Jam.
359
00:55:01.656 --> 00:55:08.376
Um, I, uh, partly nostalgic 'cause it was my, my younger days, but I, uh, I lived in Seattle for a while. I did an exchange, and I lived there.
360
00:55:08.406 --> 00:55:16.596
And just to give you this funny anecdote, I came there and I said, "Ah, Pearl Jam, Nirvana." It's after-- Uh, it's like '95, and Nirvana was not there anymore. I said, "Ah, Pearl Jam, we're gonna see them."
361
00:55:16.616 --> 00:55:25.016
All my friends were like, "Yeah, they play every year. We're gonna go." And then they didn't play the entire year, uh, that I was there. And then afterwards, they played. They said, "You have to come."
362
00:55:25.076 --> 00:55:28.896
So I bought a ticket, and I went there, and it became, it became a tradi- a tradition.
363
00:55:28.976 --> 00:55:39.216
So whenever Pearl Jam played, uh, in Belgium, my friends would come here, and then when they played there, I would go there and we'd buy each other tickets. [laughs] So, uh, it- Good, good excuse. [laughs] Good excuse.
364
00:55:39.316 --> 00:55:48.476
Rob, it was so nice having you on the, on the show. I think you beat, uh, Bill Innman on the time we spent together on discussing about data an- analytics.
365
00:55:48.856 --> 00:55:59.216
Thank you so much for joining me on this, uh, really insightful discussion on data an- analytics. Thank you for having me. [outro music] Thank you for joining us on this awesome podcast.
366
00:55:59.496 --> 00:56:06.956
As senior executives, data and analytics architects, and AI professionals, your time is valuable and we appreciate you choosing to hang out with us.
367
00:56:07.216 --> 00:56:13.496
If you like what you heard, please give us a thumbs up, hit the subscribe button, and leave a comment. We love hearing from our audience.
368
00:56:13.736 --> 00:56:29.236
Don't forget to spread the word on social media, and let's continue to drive innovation in the industry together. Thanks for listening, and we'll catch you on the next episode. [outro music]