Damien Robert is currently the Chief Solution and Delivery Officer at Pricemoov. He’s been at Simon-Kucher & Partners (SKP) for 17 years of his career so he actually understands pricing really well. Damien also took a three-year stint at Disneyland Resorts Paris. Damien develops and implements tailor-made pricing solutions, ensuring easy price steering across the organization.
In this episode, Damien talks about Machine Learning and how it helps optimize inputs in the pricing work as he shares insights in relation to Pricemoov’s models.
Podcast: Play in new window | Download
Why you have to check out today’s podcast:
- Find out what Machine Learning (ML) and Artificial Intelligence (AI) are all about and how these two are used in pricing
- Discover simple and advanced pricing strategies you can implement to reach a bigger market
- Understand why looking at price in the perspective of your client is a good move towards success
“Look at your prices with the client view.”
– Damien Robert
01:32 – How Damien got into pricing
02:34 – What does pricing mean for Damien
04:28 – Damien’s pricing story when he was still in Disneyland
06:25 – What makes Pricemoov different in the pricing market and how Damien got involved with them
08:50 – Simple and advanced pricing strategies people can implement
10:57 – Collecting competitive pricing information with car rental companies
13:00 – Defining Machine Learning (ML) and Artificial Intelligence (AI) alongside the things they do
16:45 – Using ML to forecast the effects of the decisions made for pricing
19:54 – Mark’s Amazon story: his $20 book reaching a price $225 in Amazon
22:00 – Damien working in a theme park, the existence of black box
24:57 – The things Pricemoov considers when making pricing models
27:00 – Damien’s pricing advice for today’s listeners
“Pricing is not only a price, but also the way you message it and what you can achieve.” – Damien Robert
“Our software is allowing you to configure the rules that we provide. So, you have a set of rules, you, yourself, decide which one you want to use and implement for your product portfolio, because you may want to use one or another one, and we can basically train you on which one you want to use.” – Damien Robert
“That’s what people misunderstand. Sometimes in machine learning and artificial intelligence, they think that the model by itself can be smart. The reality is that the model is not smart, but the model can be much more granular and fit much more kind of situations because his granularity and the computer allows, basically, to have so many computations and so many different alternatives.” – Damien Robert
“Somehow, machine learning can be used to optimize part of the dimension, some of the inputs, but basically, at the end of the day, you need a skeleton of decision making, which is important to have. Otherwise, you do not take the long-term impact of price changes. You do not take the potential competition reaction. So, you’re missing a lot of the points.” – Damien Robert
“Very frequently, you just optimize, set a price, but you do not check the consistency, you do not realize some of the price points, and you won’t even realize that the way your prices are featured on the web is sometimes a catastrophe. So really, just look at the end of the tunnel and look at it because you will have a lot of surprise, especially on promotion. 50% of promotional faders are usually related to execution. It’s not at all related to the design of the promotion itself.” – Damien Robert
People / Resources Mentioned:
- Pricemoov: https://pricemoov.com/
- Simon-Kucher & Partners: https://www.simon-kucher.com
- Tesla: https://www.tesla.com/
Connect with Damien Robert:
- LinkedIn: https://www.linkedin.com/in/damienrobertpricemoov/#
- Email: [email protected]
Connect with Mark Stiving:
- LinkedIn: https://www.linkedin.com/in/stiving/
- Email: [email protected]
Full Interview Transcript
(Note: This transcript was created with an AI transcription service. Please forgive any transcription or grammatical errors. We probably sounded better in real life.)
Look at your prices with the client view.
Today’s podcast is sponsored by Jennings Executive Search. I had a great conversation with John Jennings about the skills needed in different pricing roles. He and I think a lot alike.
If you’re looking for a new pricing role or if you’re trying to hire just the right pricing person, I strongly suggest you reach out to Jennings Executive Search. They specialize in placing pricing people. Say that three times fast.
Welcome to Impact Pricing, the podcast where we discuss pricing, value, and the automatic relationship between them. I’m Mark Stiving and my mission is to help your company win more business at higher prices, and helping us do that today is Damien Robert. Here are three things you’d want to know about Damien before we start.
He is currently the Chief Solution and Delivery Officer at Pricemoov. He’s been at SKP for 17 years of his career – that’s Simon-Kucher & Partners – so he actually understands pricing really well. And by the way, he took a three-year stint at Disneyland Resorts Paris. Maybe we’ll ask him about that too.
Hi, Mark. Nice meeting you.
It’s wonderful to meet you too.
Hey, how’d you get into pricing in the first place?
Pretty funny. I was an Aerospace Engineer as a background then I realized that it might be a bit boring because projects were for something like seven years, so I decided to join consulting. End of 2000, basically, internet bubble burst and not so many doors open at that time, and I found a very nice place and a very nice door open, which was Simon-Kucher & Partners – a German company that was still growing at that time despite the crisis. I entered the door and I realized afterwards; it was about pricing. That’s the story. So, almost by chance, so to say, but yeah, great lot of chance of doing this.
I think almost all of us got into pricing by chance, right? Something locked us into it and then we realized the power and the fun.
So out of curiosity, when I say the word pricing and you say “I’ve been in pricing”, what does that mean to you? I mean, it’s this huge thing in my mind, but what does it mean to you?
What it means to me is, basically, I would say two things. First, pricing is very broad and what I loved is how central it is, because a pricing strategy of a company means basically putting in front the value delivered by a company, the right product structures, the right pricing, and also all the day to day processes to steer this performance. So, its really fun stuff, both strategic and very operational. And at the same time, I found a place where I do find the right balance between quantitative stuff on one side – so where I do use my skills of engineers to model things and very quantitative stuff. So, having to understand that phenomenon, how a consumer is behaving, finding the right mix between both was like magical for me. So, that’s what it means to me.
Yeah. I think I agree with you 100%. And then price ends up being a number so it truly is quantitative, and there’s lots of reports and studies and analytics we can do, but we also have to think qualitatively. We have to think about the people and the way they think and stuff that doesn’t come natural to me, at least, and maybe not to you either. So, it’s kind of fun to learn.
Yeah. It’s fun to know that somehow, pricers are very left-brained, if that makes sense, but consumer, when they buy the product, they say, “Does the way you feature a price can drive completely different behaviors?” There’s even a book on that. Pretty interesting to read.
Just one book on that?
Yeah, probably lots.
Okay. I really want to talk to you about Pricemoov AI, but before we do that, can you tell me about your experience at Disneyland? Give us a pricing story that we’ll enjoy.
There’s so many pricing stories, but just to show you basically how pricing is essential, what we did is a huge segmentation survey, as well as pricing survey together to understand how to unlock demand at that time, and what we realized when we did the survey, is that families with young kids – so kids below seven – were not coming to the resort from Spain, Italy, UK because they were waiting for people to be able to go to Space Mountain and the full stuff. And based on that, we said, “Okay, but what if we just, for the celebration of the 50th anniversary, we’re offering for free that gets to come? Wouldn’t it unlock demand?” So, we did the research on that, and based on that, we fully, basically, communicated everything around 50th anniversary, kids come for free, it’s a present from Mickey Mouse.
Thanks to that, we grew the demand like hell, and basically, things and people came. Three years later back, again, when the kids were old enough basically to do Space Mountains, has been very successful, and of course at the same time, we had pushed prices for adults up so overall, it was balancing. It was a huge revenue game, thanks to this.
So, it shows, basically, how pricing is not only a price, but also the way you message it and what you can achieve. It has been a huge success and it was really nice to be there at that time.
I think that was brilliant. So, although you didn’t change the product per se, you really truly looked at the market segments as you changed your pricing. What are you going to charge for, who you’re going to charge, how much you’re going to charge, and how you communicate it? So that was brilliant. Love that story.
Okay. So, let’s jump into Pricemoov. Tell us, first off, what does Pricemoov do? How did you get involved with them?
So, Pricemoov is basically a startup that has been created in 2016 which I joined recently, March, so now, almost a year ago. And the reason why I joined Pricemoov, it’s because they have really great tech guys, great data guys that do offer basically a pricing platform that lows any kind of company basically to optimize our prices.
So, we are quite different from other pricing software company in the sense that we are hosting all the data through the latest technologies like Snowflake, and basically, based on that, you have your product portfolio and pricing, but you can build any kind of pricing strategy on our platform. Being very simple, you have some basically pricing rules which are already embedded, and then we can develop for you much more advanced pricing structures, but somehow, if you need a new strategy two years later, you can still build it on it. So, it’s not that you have a price engine which is set in stone with your very relative solution. We have even an SDK approach so that if your company is having data scientist, they can basically develop the strategies by themselves.
So, it’s really an open platform in terms of pricing where usually, you’re more buying a price engine and you have something a little bit too much setting and so on. That’s the big difference. And we try, also, to have an interface which makes it much simpler for any kind of user. Some people in car rental agencies are even able to monitor how their bookings are doing, how the competition is doing, and by themselves change the prices if they want to. It allows, also, very much less advanced people to manage pricings as well. So, simplicity and relativity, I would say, makes us quite different in the market.
Okay. So, let’s jump back to the very beginning of what you said. I remember back when I first started looking at pricing systems, I was lost. I didn’t even know what, I couldn’t really describe what it is they do. And so, you start talking about what are some different pricing strategies we could implement. Can you give us a couple of examples of pricing strategies? Let’s do the simple ones first. So, a couple simple pricing strategies that some people might be able to implement.
Simple pricing strategies, for example, for an online retailer would be combining, basically, competition prices and margin information to basically find the right balance and you would basically already define from the beginning what kind of behaviors you want to be having, depending on how you’re positioned versus competition, and what are your margin levels. That’s one example of basic strategies. Some people want to be even lower the competition more systematically in certain SK use. So, you can have a set of simple strategies like that and you do affect, you do apply it basically to your product portfolio or your different locations very easily.
And the more advanced one of the classical, we work for CMA, CGM which is basically you know, a large ship operator, large vessels; it’s the fourth on the market. You have Maersk, the Danish companies, and you have MSC, and then you have here at CGM. So, it’s a 30 billion plus revenue company, and based on the feeling factor, so as a head or as a late in terms of bookings, based on that, basically, we’d make different pricing recommendations in terms of price adjustments that’s much more advanced and that would even be considering forecasting methods, demand forecasts, that will be basically supported by machine learning in terms of optimization. This is what I would consider basically more advanced strategy.
So, any kind of decision tree approach as well between conversion, traffic information, sales volume, price versus competition would be also quite advanced. Even so, that would still be very transparent, which is something we really like. So, something is still where you decide and you know what you want in terms of behavior, in terms of pricing behavior.
Pretty fascinating. So, if I go back to the first one that we talked about, you’re also then collecting, somehow, competitive pricing information, and in the car rental example, you’re even collecting availability of vehicles?
Yeah. In the case of car rental company, we do basically directly connect through with data from our customers. So, they do instantaneously provide us access to availability, to level of bookings from the different segments. So, we know, basically, lively on top of competition information what is the level of performance, how does it compare to historical data, and this falls for the different segment. So, if you had the corporate segments at a fixed price but you can’t still tell them yes or no in terms of bookings, then you have the classical B2C channel where you’re fully dynamic pricers. And so, we make recommendations basically on prices for the different segments and availability in our engine.
Yeah, okay. So, when we talk about fully dynamic pricing, then, we’re talking about your software setting the prices. Is that correct?
Our software is allowing you to configure the rules that we provide. So, you have a set of rules, you do yourself, decide which one you want to use and implement, basically, for your product portfolio, because you may want to use one or another one, and we can basically train you on which one you want to use.
Right. And so, as in the real car example you gave earlier, I could change the price because I’m seeing what’s going on or I could have set up a set of rules where you’re just automatically changing the price for me and I don’t have to worry.
And when you have 300 agencies and you have only two people to do that, you need to implement rules. You cannot do it by yourself. So, you’re somehow setting the rules and then you will receive alerts if there are specific situations where it’s even going out of the road and you should be considering and watching specifically those states.
Yeah. So real quickly, define AI and ML for us, and then I’m going to ask another hard question.
Sure. That’s very interesting because this is also what I always do when I want to explain to people what ML can do.
Basically, what I usually say is that historically, when you wanted to define a model, you had to run an analysis, and based on this analysis, statistical analysis, you were understanding a phenomenon, and based on this understanding of the phenomenon, you were deriving a model, and this is how I want to change things based on this model.
What machine learning brings is, because you have so much computational capabilities, we can directly fit a model to your extensive number of input and try to fit this based on live data and retrain the model very frequently. And this can be much more granular, thanks to that. You have a neural network, you have a model which is waiving different models, so you’re able to be very, very granular. If you have a random forecast, same thing. Even so you don’t have all data because you have different decision trees, working and being trained on a subset of variables and a subset of data, even so you don’t have all the data, you are able to still find a recommendation. So, this is really what machine learning brings.
Okay, so pause for just a second. When I did my dissertation at Berkeley, I created statistical models using scanner panel data from grocery stores and I was trying to figure out whether nine cents works and how it works in making people’s decisions. So, I had to create different equations, different models, and test the fits of those models. Now what you’re saying with machine learning is I would never have created the models. The software itself is essentially creating the model. Is that true?
No, that’s the limit where people don’t always understand it. It’s that you could try, basically, if you could define exactly four simple things that does work, basically like this, so if you want to identify if words are positive or negative, this is how it works. If you’re trying to do something much more complex, you still need to build an architecture as some halls, machine running, and the model will try to optimize the skeleton. And then yes, it will give you an answer automatically, much more easily. But you still need to kind of build this skeleton if you want it to be successful. And that’s what people misunderstand.
Sometimes in machine learning and artificial intelligence, they think that the model by itself can be smart. The reality is that the model is not smart, but the main model can be much more granular and fit much more kind of situations because his granularity and the computer allows, basically, to have so many computations and so many different alternatives. Or in your case, you need to take an assumption of what is the structure and you had much less basically situations that you could model, somehow.
So, it brings more granularity, it does work even so you don’t have all data, but it is not really good enough to solve a complex problem. It’s not like the magic wand button on your picture to make the picture look better. Some people still think, “Let’s press the price button and it works.” That doesn’t work.
Okay. So, there’s supposed to be intelligence, structure, design around all the machine learning and AI we put in.
I can just picture people saying, “Yes, but it’s a black box. I don’t know where that price came from.” How do you address that in the world of machine learning?
In this role, for example, I did work for a ferry company, and for this ferry company, what we did is we split the mall into two pieces. One of the critical input is basically to demand forecast. And on this demand forecast side, basically, we identify the right variables. So, how late are we since one week, four weeks, one day, since the opening of the sales? You build the right variables. And by training the model basically on this, you get the right for them. So, it is good at basically building the right forecast. But then, the decision tree about pricing is a fully transparent model where one of the first dimension that you consider is always going to be my occupancy level. This is considering a machine learning input which is forecast but taking it into a transparent decision tree which afterwards we’ll get is always conversion. If conversion is good, it shows that maybe your price’s too low. Conversion is very bad, maybe your price’s too high. Then you consider also the last dimension which is competition pricing. So, machine learning is one input but the decision tree is transparent.
And if I have to give one example, which I frequently use, it’s the Tesla. Why Tesla? Because basically machine learning and AI are used to take all the information from the different sensors to know if there is an obstacle or not, but afterwards, all the interpretation to say is the obstacle moving? Is the obstacle on the road? Should I drive left, or right? It’s a fully transparent model. An engineer, an automotive engineer, has designed basically where to go, and pricing is the same.
Somehow, machine learning can be used to optimize part of the dimension, some of the inputs, but basically, at the end of the day, you need a skeleton of decision making, which is important to have. Otherwise, you do not take the long-term impact of price changes. You do not take the potential competition reaction. So, you’re missing a lot of the points.
Okay. So, I think I understand what you just said. And I think what you said was, we’re going to use machine learning to forecast the results of decisions that we make. And so, if I change the price from $1 to $1.20, I can forecast what I think is going to happen, and machine learning will help us get to that forecast. But I still have to make a decision to go from $1 to $1.20.
Somehow, yes, exactly, but when I was saying forecast, whether forecast as a demand or forecast how the consumer should react, but based on that, do you want the automation to change the price or not? It’s still a transparent decision making.
Right. So, I could still write rules the same
It can still be automated, but basically, it’s a transparent rule.
Okay, awesome. Let me tell a quick story that happened to me eight years ago, seven years ago now, and I just want to hear what you have to say.
I wrote, my very first book was called Impact Pricing. And I looked online, I’d gotten letters from my publisher saying, “Hey, we’re out of print.” And I said, “That’s fine. I’m not going to pay you to print more books. I don’t really care.” And then I watched on Amazon, and the price of my book which is normally a $20 book was it $225, and I was convinced that that’s because of automation. And so just for fun, I had 50 books here in my office. So, I published or I built my own Amazon store and put them back up on the market for $25. And I just watched that $225 price walk back down. It was pretty fascinating. So, what do you think happened?
It’s very interesting. I haven’t worked at Amazon, but at least, the initial one has been that the model couldn’t find, basically, any rational or any change in volume following the price change, but there was probably not enough volume change so that they can interpret anything, or they were doing some price testing with your book. I don’t know exactly what was them were making. But then, when they did realize basically that there was another book at $25, and I don’t know if they had access to your sales at that time, they did understand, basically, they understood that there was another case and there were higher revenues made being a $25 and based on that, they took the decision to follow it and go further to your price, but I have no clue. I’m just making guessing here, but maybe you have a response?
No. I think it’s just; I would guess the same. I don’t think it was Amazon people making decisions. I think it was companies who had Amazon stores, and they were running some type of software automation, and it looked exactly like you said where there’s no volume change when we change price, let’s keep raising it until there’s a volume change.
What I can just tell you is that I have been working for a theme park which is operating in Orlando, and basically, there is one which is smaller than the other one that has fixed capacity. I was working as consultants at that time, and I had basically said yes to a software provider, fully AI ML based, zero pricing expertise, and I told them, “Okay, we will put that in place and we’ll take a bottle of revenue increase that you will get from us.” And all of a sudden, the prices started basically to go up even so the sales were not good. And basically, we asked the people, “but why is it so? What’s happening? Why is the model telling us to increase the rates?” And they told us, “We don’t know. Set the model.” And the only way out has been to unplug the model and they went back home.
So that just tells us that if the model is fully black box, I fully understand the first command to do our makings. There was really no backbone in terms of decision making based on which parameters, which data is taken into consideration, one is predicting the demand and the other one is making the decision on pricing itself. So then, we put something back a more rational approach and it did work. Just to tell you, black box does exist.
Yeah. So, I really liked the idea that says we still have to have people thinking through what pricing decisions we’re going to make. Does it make sense or no, and how to make those. So, I approve of that.
I even so think, I don’t know if you’re the same in the US, but in Europe, they are really strong. They really look strongly as if there any companies that are aligning on pricing, and issues done having something which is completely black box. How can you prove that your algorithm did not agree with the other algorithm to make prices go up? And there has been something in Europe where car rental players were using the same company basically to make pricing decisions on their box and the algorithm were really looking at going into the same direction.
So, I think that the very important to prove is there is an engine, there is a software that human beings deciding how to set up the software, and basically, they have to be quite independent on that. So that’s also the beauty of having one platform where still, the pricers and the product owners decide how to set up prices.
Nice. So, let’s talk about the last topic we got. We’re running out of time. Do you think about customer perceived value when you create these models?
That’s a very complex one in terms of putting rules. What we do at least on this is we take that into consideration on one first module which is a price building module, which is setting the rules between SK use. Because in reality, we do not optimize all SK use independently. We decide that some SK use will be related to the other one. So, for example, if you have different sizes of bottles for the same drink, basically, there will be still some indexation happening on that. Same thing on some other products’ packaging, or sometimes the different SK use are just the different colors or the different versions, somehow.
And then, thanks to that, basically, when we do optimize the price points, which are free, then all the other price points keep the consistency. So that’s one dimension where we take that into consideration.
The second is that even in the pricing rules, we put basically boundaries because there is a way to say, you start from this price but never go up than this price and never go below this price. So, we take into consideration psychological threat folds and so that the model has the freedom to evolve with interest rates also. That’s the second dimension where we take that into consideration.
And also, in the frequency of price change, you can define how frequently you’re accepting to change prices, because there are also some businesses where you know that you don’t want to price change every single minute or every hour because it will drive too much stress on a very high-priced product. You will never do it. That’s another dimension that we take into consideration.
It’s not perfect, but that’s how we do it.
Yeah. So those are actually mostly psychological aspects that you built into the model, which is kind of interesting, actually.
Damien, this has been fantastic. We’re going to have to wrap it up, though. Last question. What’s one piece of pricing advice you would give our listeners that you think could have a big impact on their business?
Look at your prices with a client view. Don’t look just as at prices, but think about what they really on the websites, what they see on their prices, what they see in their contract, and look at it with their eyes. Because very frequently, you just optimize, set a price, but you do not check the consistency, you do not realize some of the price points, and you won’t even realize that the way your prices are featured on the web is sometimes a catastrophe. So really, just look at the end of the tunnel and look at it because you will have a lot of surprise, especially on promotion. 50% of promotional faders are usually related to execution. It’s not at all related to the design of the promotion itself.
So that would be my piece of advice.
I could buy that completely. Just put yourself in the customer’s shoes and look back at your pricing. What are they seeing? Does it make sense or not?
Damien, thank you so much for your time today. If anybody wants to contact you, how can they do that?
Let’s go on LinkedIn. Message me. I will be super happy to respond to you.
Okay. I will have the link to his LinkedIn page on the show notes.
Episode 158 is all done. Thank you very much for listening. If you enjoyed this, would you please leave us a rating and a review? Those are very valuable to us. And if you have any questions or comments about this podcast or pricing in general, feel free to email me [email protected] Now, go make an impact.
Thanks again to Jennings Executive Search for sponsoring our podcasts. If you’re looking to hire someone in pricing, I suggest you contact someone who knows pricing people. Contact Jennings Executive Search.
Tags: Accelerate Your Subscription Business, ask a pricing expert, pricing metrics, pricing strategy