התמלול המלא של Pricing in Motion: Michael Wu on AI, Data, and Demand

you search for the same route the price may already change that. It could go up or could go down. But the fact is that, like, they change the price to essentially meet the changing market demand. So they're very dynamic. So and that's something that I would say a lot of companies don't leverage well enough yet. I think they're traditionally the price has always been very static and that works typically I would say fairly well when you have a market that's not hyper volatile, but as you said earlier, we are actually living in, a highly volatile world. We have post pandemic, the war is going on. Everything created a lot of economic uncertainty. So this volatility is going to persist. So what that means is that when this market volatility—if you don't adapt and change with those volatility by changing your pricing strategy to take advantage of these volatility, then you're going to be essentially priced out of the market by your competitor. And yeah the last thing— AD: Yeah MW: I think the last thing is essentially the trust in AI, I think if you have AI price this algorithm, that means you must be able to develop a trust to believe that the decision that they make for you is somehow working, And not just kind of overriding all the time. If you're using an AI algorithm to help you price and the human consistently just overriding it then you are you're using the system, but you're not really taking this recommendation, and no one is getting any benefit from that. AD: So what have you seen in companies adopting AI developed pricing models? Because there's a lot of fear in most organizations of just handing over something as business critical as your pricing to what could essentially be a black box. What have you seen in terms of how companies can adapt to that and kind of how they go through that change management process? MW: Yeah, I think there's a few things. I think the, the technology can certainly help by providing what we call explainability. So black box is typically given to this algorithm that we can make sense of, kind of neural network. You know, it's not really that we cannot make sense of it just because they have such a high number of parameters that we we can't really intelligibly make rationalize what's going on there. Just like we talk about oh, “why is this customer, buying this product get recommended this particular price?” There could be hundreds of factor going into that. And we're not going to be able to kind of rationalize that, okay, this one, this factor, change the price in a certain way. But that one change it maybe and back in a different way. And so there are hundreds of these this factor this going on. So it's really hard for human to actually think about that in a rational way. But it doesn't mean that a machine cannot do that. Machine is actually very good at doing that, that machines actually, can actually look at all your, your data, each one of those factors. And then actually, essentially give you, a kind of the average a driving kind of how strongly each one of these factors driving your price. Right? So these are kind of the explainability that we provide to help people adopt this black box type of algorithm. So it's no longer truly a black box anymore. It's really sometimes we call this glass box. You won't get completely transparency into exactly what the algorithm doing, but you don't really need to because you can't intelligibly make sense of it—millions of parameters in the neural network anyway. But what, you want to do is to get a sense of, like, how does this input drive my output? So you can actually do that using these types of explainable AI. AD: In your experience, is it better to focus on the demand side and improving your demand forecasting capability, or to focus on the supply side and building better predictive models on what your cost inputs are going to be? MW: I think it's both. Because, I mean, the price we know from our basic Econ 101 that the optimal price is when you have this equilibrium condition. And that's when supply matches your demand. And so you kind of have to do both. I think in some situation one maybe easier than the other. I would say that like for the airline customer side, they are actually operating in, rather inelastic kind of supply, regime in the sense that as demand changes, If you have a consumer all of a sudden want to all want to fly somewhere. Airlines don't have the luxury of putting more seats in the plane overnight. I think they can't change the supply of their seats very effectively. So their supply of their seats is rather inelastic. So what they do is instead of changing the supply to matching the demand, they actually change the price of their seats to kind of control the demand level. So then on the B2B side, typically what they do is almost completely different, completely opposite in the sense that you are operating in what we call a near infinite supply regime. When you have this infinite supply, the kind of supply curve is typically very flat. And what that means is that you could treat in most that situation, you could treat the supply as pretty much fixed as a fixed cost. And basically what that means is that you can only you always do just focus on modeling the demand elasticity. So that’s why a lot of the B2B kind of pricing algorithm boils down to essentially some kind of willingness to pay estimation and that's essentially a modeling of the demand. AD: How do you determine— how do you really get to the bottom of identifying what's the willingness to pay? MW: Yeah, willingness to pay. I would say that that's only one step, because the willingness to pay is to look historically what customers have been paying for a particular, product. These type of customer have been paying this much for these type of product. So, that's historically I could look at historical data to essentially build a model to help me essentially predict this willingness to pay. But that doesn't mean that it's

actually a good price or even optimal price. So knowing that what you're historically willing to pay is a good starting place, then you need to what we call go through the second stage of optimization. The optimization is a process of looking at your win loss and win rate and then determining what is the truly optimal price, and then from what you historically had been willing to pay, your willingness to pay, how do you get them from what they historically are willing to pay to this optimal price? AD: The airlines have been operating in essentially near real time for decades now and basically every other industry is trying to get towards real time, a quick time at least, or something even approximately faster than that. What are the lessons that people could take from how airlines operate and the, the time frequency? What can they and other industries apply to getting towards a faster kind of quick time operation? MW: Yeah I think one, one thing is obviously you know, to trust a—I mean because no human can actually price in a real time manner at the scale of an airline. Now, even if you have a large team of analysts, you can never do it. You just can't. Human just cannot do this. I mean people need to sleep. So you can't have them 24-hour there nudging price up and down and all that stuff is just not possible. You have to have an AI do this. That means you need to trust it. You need to that price so and not kind of always kind of the “oh, AI says this is the price? I don't think it is. I believe what I believe.” And a lot of times, people kind of overestimate their experience and their knowledge because their experience and knowledge is important, but so does everyone else. So does every other salesperson in your team. And what AI does is able to collectively learn from all those people. And they give you a more holistic picture rather than just one person's experience. So you actually gets essentially be able to price more optimally by leveraging the knowledge from everybody. AD: So for all the talk of artificial intelligence in a lot of companies, a lot of industries, just generating a sales quote is still a friction-laden, timely, sweat-inducing kind of process. What can companies do to shrink just simple quoting time, particularly in industries where you have very complex, super quotes? MW: So there's a few things. Obviously technology will help. I mean, if you look at actually create some quotes using the more traditional way, then obviously there's no hope for that. I mean, so you have to use CPQ type of technology to help you streamline these so whatever you need approval, wherever you need a signature here and there. Then you streamline those processes. So it makes it essentially, remove those friction that you talk about. So that's one thing. The other thing is that the CPQ, the price part of it is actually the key. It's actually also crucially important, because that's where the AI comes in, actually comes, rather than just making it more efficient and actually making it more beneficial to the business. Because you could price it in a way that optimizes your margin, optimizes your revenue. So not only is it's a faster, but it's also better. AD: One of the other things you mentioned is the use of third-party data. What have you seen being done effectively in other industries outside of the airlines, in terms of how companies have ingested third party data? What types of data? How do you mediate that data, all those problems, what have you seen as kind of successful approaches to bringing in more data into the picture? MW: Yeah, I would say that a lot of the I would say consumer, retail, or B2C type of business are actually using more data then some of the airlines do. And actually, I think because these B2C industry, they have what we call the demands actually highly volatile. They're probably more volatile than, than the airline. So they can't forecast far out into the future as an airline airline can forecast or demand very far out up to a year in advance. And that's actually why you can actually buy a ticket for travel a year from today because they have to be able to forecast the demand a year from today. But most retail B2B company they can't forecast that far out. They can only forecast you know a few months maybe a quarter or a couple quarters out. And what they need to do often is leverage more different variety of data. So if you are a retail B2C you probably use everything from clickstream to historical transaction, social media, or anything that could help you kind of gauge, predict the demand for your product and services. To help you make that demand forecast. So in terms of some of the use for data, that's for a lot of these B2B company, it's very domain specific. But if you are in a logistics company maybe the weather is really important. Maybe the storm then is actually important to know about that. But if you are in chemical or maybe the petroleum, under the oil and gas index, it's actually really important. If you are more in the consumer space and manufacturing, then the commodity price index may be really important. So and so these are very domain specific. But by leveraging these data you can actually help them price better because those are the data that their analysts use to make those pricing decisions anyway. AD: A skeptic would look at AI and say, well, this is just a shiny new hammer out in search of nails to be pounded down. How do you avoid that kind of problem, if you will? MW: Yeah, I think that as Chief AI Strategist or someone in this role, you need to not fall in love with the technology, okay? I think that's something really hard because you are in that position because you probably love the technology, you probably love AI. But you need to kind of restrain yourself or, kind of resist falling in love with the technology is that you need to fall in love with your customer's problem. You need to think about how are you going to use this

technology to solve your customer's problem, rather than just, I love this technology, I'm going to use it everywhere. Then you run into that exact same problem, and you try to hit everything with this new hammer, a new shiny hammer. And that's not necessarily, I would say the best use cases. Sometimes you will find a use case that will work, but it may not stand the test of time. A few years later, maybe it doesn't scale. Maybe you find it to be too expensive. So a lot of these things, you're not truly leveraging the for example, generative AI is very, very hot now. Everybody is thinking about different ways to use generative AI. But a few years later, you may find that, hey, it may be too costly. Maybe there's the data center don't have enough kind of energy for you to support your use case. So, there are lots of things that that I would say are not necessarily the best use case of this new technology. So you need to really think about how you use this technology to actually solve a customer problem instead. AD: Fall in love with your customer's problem. MW: Yes. AD: Great advice. And what advice do you give to the CFO, who one is fearful of just handing things over and who also wants auditability and the ability to kind of trace how a pricing, how pricing is calculated and delivered, how do you help them overcome that fear? MW: I think the CFO, the one thing that they need to do there's more important is to first, I would say, recognize the power of pricing. I think pricing pricing is actually, I would say, the fastest and the most effective driver for margin or revenue. Okay. First I need to recognize that, okay. Because in fact there's numerous, academic papers and studies. You know, I think first with a very famous Harvard Business Review article actually illustrating that the power of pricing is really, really something that people overlook. For a long time, actually, people have demonstrated that one-percent change in pricing can lead to as significant as 11% margin improvement. And subsequently, people have actually demonstrated that one-percent change in price. It could be up or down, whatever the optimal direction is. One-percent change in price is actually has more significant impact than one-percent change in pretty much anything you can do in the business. You can have one-percent cost reduction, one-percent efficiency improvement, one-percent increase in headcount, anything you could do in the business. One-percent price has more impact on revenue than any of those. AD: It's the most single, most cost-effective lever you can pull. MW: Yeah. AD: To have a material impact on the business. MW: Yes. And it's also fast. It’s the fastest and the most effective driver. So recognizing that right you know is the first step. And then the second one is obviously to you need to invest in these AI pricing technology. To help you adopt these kind of pricing disciplines. Because these like I said, these pricing disciplines, they are becoming much, much more real time, And real time dynamic. So you can't possibly have human kind of work 24 hours every second and be up to date and tracking what's going on in the market and price accordingly. You just have to have an AI system to help you do that. AD: So pricing is one piece of one letter in the CPQ equation. MW: Yes. AD: Or three-letter acronym as it were. Another key piece of that though, is quoting him for a lot of complex B2B businesses, just generating a quote can take days, weeks, even longer. How do you help companies get to a faster quote time? And what does that end up resulting in in terms of outcomes? MW: Yeah, I think the quoting process, this is more like a workflow orchestration. So that typically result in using technology like CPQ. Will get you efficiency gain. So you get much faster kind of quote turnaround if you need like signature or approval from someone. They automatically get notified. So it's orchestrating this entire workflow to get a quote to be able to present that quote in front of a customer. Lots of sales have to happen and if you can orchestrate all that in a in a very seamless fashion, then you get efficiency. AD: So in closing, let's nut this out. What are the three things that every CFO can do to optimize their pricing and make a material difference in the business? MW: So one thing that I think every CFO can do right now is I would say, get your data ready. Because a lot of these pricing algorithm, they're all AI based. And that means you need data to train them, so if you don't have those data then there's no way out. So as I said before, data essentially is the field for your AI. So and certainly having pricing management tool will help you do that. But there are many other ways for you to collect data as well. So I think the second thing is to get familiar with AI. Get educated. I think both for yourself as a leader and also for your team, to understand how these AI work, how this I actually help them do their work better and not just merely there to, to replace them. And, I think this includes, I would say, some change management, which they could always start early because when you introducing, a new technology, there's always, I would say some change management that's required because it changes the way, people work and especially with AI because remember, I, as I said before, is machine mimicry that both automate human decision and learn. So the fact that it's able to automate human decision is actually rather something new. There's very few technology before I that could automate decision. Now you can automate lots of physical tasks using robots assembly line. But decision is something that always it's you know, humans always do that, And traditionally. So some people are not familiar with that and not comfortable with that. So you need to start that change management early. And then I would say lastly I know just you know, talk to vendor. The provider is AI solution. Talk to reputable, vendor that have proven track record of success and scalability and, and all that. And that will get you started. AD: Doctor Michael Wu, Chief Scientist at PROS, thanks so much. MW: Thank you. It's my pleasure. AD: So what did we learn here? The key things that we picked up. One one-percent change in price could be up, could be down. Can have up to an 11% impact on the business. That's a tremendous single lever to pull for any company. The second thing we learned here is lean into data. Getting your data in order. Lean into change management and lean into being willing to change and adapt pricing models on a more frequent basis. And I think the third thing that we heard is that moving towards real time quoting can have a significant impact in changing in a positive way. The competitive differentiation of a business. What a difference. So thanks for joining us. Thank you for listening to The Margin. If you have questions about today's episode or if you'd like to schedule a call with an analyst, reach out to us at insights@mgiresearch.com. You can also reach us on LinkedIn, Facebook, and X. You can find more information about our research and advisory work at mgiresearch.com. Until next time.