The Right Offer, The Right Value: The Science Behind Evolving Airline Pricing.

Speaker 1 Hi everyone and welcome to Team Costs May Edition. Let me start by saying that for years airline retailing has largely been discussed as a technology transformation. We've been very focused on the technical side, focus on NDC offers and orders. AP is channel modernization and reducing specially reducing legacy constraints. All of these foundations are fundamental to create a strong modern commercial model. Although by themselves they are not enough to create competitive advantage. It is important that the airlines can generate the right offer, the right value, the right customer and under the right market conditions and with clear commercial intent. And for that, we believe that of a basis of an offer optimization, a strong offer optimization is the enabler of an intelligent and off a generation, which bring us to the topics of today. So we are going to explore pricing, science, willingness to pay and how leading industries are already using data and experimentation to optimise commercial outcomes in real time. My name is Katarina Silva, I'm an engagement manager at Travel in Motion. And to explore this topic I have with me Mark Knitker, partner at ADC Consulting. Mark, welcome. Speaker 2 Hi Katarina, thank you so much for the invitation. Let me give a small introduction. So my name is Mark Nitzker. I'm based in Amsterdam and I started my career in academics as an assistant professor in econometrics. After moving to business where I founded my own company on boutique econometrics focused consultancy where we do a lot of prices assist the estimation and impact measurements across various industries, most notably aviation and retail. We grew quite quickly with that value proposition and more specific positioning and sold 1 1/2 years ago to a larger consultancy ADC, where now I'm a partner and I run the Dutch office and I focus on building gasoline friends and pricing capabilities across other industries such as financial services. Happy to be here. Speaker 1 Thank you, Mark. So let's jump directly to the topics. Let's start with revenue management. This is an area that served airlines extremely well for decades and things are changing. So what is structurally changing today that makes traditional RM and pricing logic no longer sufficient for airlines? Speaker 2 Traditional revenue management was built around fixed booking classes and focuses on forecasting demand for these various booking classes to decide which ones to keep open and which point which ones to close at specific days before departure. And that logic has worked very well and the underlying economic trade off is right whether you want to accept the booking or weight varying the current potential revenue versus the opportunity revenue that that is still very much valid. What's changing, however, is the speed at which technology develops. And nowadays technology has moved forward so much that we have become able to change prices much more granular and much more quickly. And we see that as the industry is moving towards basically modern retailing right away from filed fares and towards offer and order management and continuous pricing. And the research has been, has been done numerously, right. The price is real. There's a lot of upsides to be gained by making your pricing more reactive and more granular so you can optimise more efficiently. Now the interesting question is what's holding airlines back in terms of moving as quickly as possible in that direction and capturing that price? So many of the conversations that I see in airlines focuses a lot on tech capabilities, right, on platforms and moving towards the ability to be able to change prices as as I mentioned. And that work matters a lot, right? It's basically what enables to then decide on what the best price is. But the harder question then becomes, if I can change prices, how do I decide what the best price is? Continuous pricing gives you the opportunity to change prices. Overall, the optimization is typically not easy, but well understood, right. If I know the expected demand for a given price, then I can pretty much optimise to decide what the best price is. But the core question then becomes what is the demands that I see for any given price? And that's the shift I find most interesting. Execution is maturing quickly, but this bottleneck in terms of demands estimation or price elasticity estimation has become the bottleneck in a lot of different airlines. And that's really a measurement question and why I think a lot of this, it'll be very interesting from an ecometric point of view. Speaker 1 Thank you, Mark. And and actually I that's very interesting. Exactly. That's why we are having this conversation because you are absolutely right. There is, there is a lot of, a lot of to explore in it. But you said the old model that was largely about forecasting and demand and managing inventory. It's it appears to be, you know, moving to a need of understanding something on how I understand what you explain is that is moving more to learning a customer behaviour have a real time insight that allows them to create this continuous pricing capability and explore the margins that the airlines can take out of it. So what actually is replacing this traditional approaches and how does it look like the science on the pricing side? Speaker 2 Good, a good question. So, right, traditionally when we have of these booking classes, we try to forecast demand as accurately as possible and forecasting as a well understood problem that's studied a lot in machine learning and machine learning got very popular due to forecasting capabilities such as self driving cars and our generative AI and natural language processing. In business, however, we rarely want to forecast just for the sake of knowing what will happen. We want to forecast to improve our decision making and when we forecast what will happen at a certain price to then change the price. That creates a feedback loop which makes your optimization basically biassed towards the wrong optimal prices. And that's also it can be summarised in terms of correlation doesn't equal a causation, but that's the fundamental difference in terms of it becomes a measurements problem. Right? Not how many boogies will I get at the prices I happened to charge in the past, but rather what would happen if I now charge a different price, something different. And there's many different ways of tackling that problem. The the golden standard or most direct way of doing it is experimentation, where you vary prices in a controlled but random way to see how customers response. However, experimentation is difficult in an airline setting because there is different types of operational constraints. Firstly, there is commercial risk, right? You can't just randomly overcharge some customers and others not without risking your branding and general relationship with your customers. Another one is cultural challenges, right? Commercial teams are trained to defend revenue per flight, so telling them to change prices deliberately to, for instance, A potentially wrong 1 can be difficult to get through. And 3rd, and that's where experimentation in the airline space is really different than for instance, in a digital space. And that's that the feedback is slow, right? It's not about changing the colour or the design of a website, but you're really changing physical processes and booking curves are weeks or even months long. And so AB testing can become quite complex as well as expensive in terms of calendar time. Now experimentation is is the golden standard and all these challenges can be overcome. However, what you also see is movements towards elasticity estimation from observational data, right? So data that you have already captured in the past that contains price variation and that price variation you can use to estimate elasticities using a variety of different technometric methods because those prices weren't sets randomly. You basically then have to take care of this correlation doesn't equal causation challenge. So basically experimentation or observational methods. Both are strong ways of thinking about measuring price elasticity. Speaker 1 I see all of this as being growing past towards on how the airlines have been doing RM and and potentially all of these challenges you mentioned are what keeping airlines also being a little bit shy on going on moving forward or a little bit slower than than we expected. Because even though this is a popular topic, let's say, and there is a lot of airlines talking about it. We see that this is something that is for them still seen as highly innovative. So that there is a lot of things that they need to explore and that they need to bring in that today they are not doing and especially using experimentation as you mentioned in a, in a higher a broader scale. But in reality, and I would get out a little bit of airline world here. This is the type of techniques that are being used in other industries. Talking with you before we talk about a ride hailing e-commerce, all of these other sectors that are already using that. So what have you seen in terms of models operating in these different sectors that can be used for the airlines and that can learn with it so they can actually overcome all of these challenges in a more confident way? Speaker 2 So even though every industry obviously has its own nuances and complications, the general shift of technology improving and therefore being able to change prices more often and more granular is quite common. And we see this, for instance, very much in in e-commerce or in retail, right, where the the Walmarts, Amazon's, the salon, those they all have online prices that automatically update based on estimated elasticities. We see more and more digital signs in supermarkets. So they canmore quickly adapt based on expected demands at different prices. You also mentioned right hailing, right. I think that's the most famous example. Be a search pricing is essentially life price elasticity estimation based on the number of bookings and the available demands and all of these industries. The idea is the same, right? You have the opportunity to change prices through Francis Digital shelves. And once you know what will happen, it's pretty approachable to determine what the best price is. But then the core challenge is what will now happen when I change a price to a certain new value that I haven't tried in the past and the science is quite mature because it is so present in different industries. So I would say in that sense, the bottleneck for airlines is not invention in this space, but rather adoption and building the data foundations to enable this type of science. One caveat though, right? We should not oversimplify. So some of these industries have advantages that airlines don't. There's fewer regulatory constraints and distribution can be simpler as well as feedback being faster. So it's not directly copy paste, but the underlying logic and science of estimating a price elasticity or estimating A willingness to pay very much transfers between different industries. Speaker 1 It's great that you, you finish with that sentence because it takes me to my, my next. My next question that I was thinking about while you were talking is it's actually, let's say it's quite clear what is the ambition? It's quite clear what is the, the, the destination, But it's, it's a little bit different when we think about how to operationalize that and how to bring that to reality. And, and I think that's that's the main challenge you were mentioning, which is it's not just a copy paste, it's using the logic, but adapting it to the reality of the industry. And I wanted to ask your opinion and bring it back to airlines again on how do you think how can airlines apply that in a tangible way? How can they they make this new step deliver credible results for them? Because again, with the challenges it may they may have some concerns on how actually our personalised these real time observations. Speaker 2 Yeah. I mean, so in terms of movements towards this direction, there are a few things that that I typically see. Maybe the most common one is experimentation on ancillaries or another more constrained subspace, right, bags, seats, lounge access, priority boarding or other upsides. And This is why because it's one of the easiest places to start where you have a very clearly defined problem space because you can upgrade to certain to certain subsets and it's more easy to experiment and randomise. These customers don't see the experience of another one here. The feedback is also faster and it is easier to get internal buy in. So that's why I believe this is where a lot of airlines begin. Do also see movements towards elasticity modelling at finer granularity, right? Really moving from broader market level elasticities to going all the way down to routes and point of sale level for different dates towards departure or different time frames and different customer context. And this is where a lot of the the real revenue is right? Moving towards more granular price elasticities allows you to do your pricing more more granular and hence gets get more out of your pricing algorithm pricing implementations. Now the most interesting technical challenge there is dealing with confounders, where a confounder is a variable that impacts both price and sales. And confounders can be numerous, right? Examples are oil prices, macroeconomic show shocks, geopolitical tension, competitor pricing, or even marketing activities within your own company that you have not tracked well over time. And if you don't correct for those, then the elasticity that you measure is underestimated, which will lead to mispricing systematically. Basically the system will recommend too high, too high prices. And we've worked with various airlines already on this space. And what I usually see is that some form of machine learning algorithm has been implemented to measure these elasticities, but they don't take in consideration this complexity with unobserved confounders. Yeah. And that that's the two main applications that I see, right, experimentation, ancillaries and more granular price elasticity estimation. But I think the the oldest map where most airlines are today is that they are still at the platform ambition or at the move towards the new technology to enable this type of price variation. Speaker 1 Then let me ask you this because when I hear you explaining what are the benefits of it and we are living that right now right. The the airline industry has been living in the past years. A lot of have fueling these Co founders having a lot of influence in the way they do their pricings and the way they also guarantee their load factors and so on and so forth. Because of course they want to get more revenues and get the customers to buy more fights and so on and so forth. But we still, even though the benefits are clear, we still see a little bit of difficulty sometimes. I'm having the buy insurance, the internal buy insurance of the sea level suites that see also the benefits, but somehow struggles to know on what to prioritise. Where do I start? So if you if you were talking right now to let's say CCO or CIO of an airline, what would be the mind shift or what would be the action you would recommend them to prioritise in order to start benefiting of all of these new techniques that are more and more within the technology available for the airlines? Speaker 2 Right. I think there's a few ways to start, but the most important one that I start with this to not wait for the full transformation, right. Many new systems are being developed and they're all further improving. And the temptation is to scope a multi year programme before being able to then change prices and actually implementing it. And that's how a lot of time passes with small progress. So I would say start small in a place where the risk is bounded. And a really good example would be in ancillaries. And you can run many models, but you really should run real experiments and learn how to measure those properly and build experience as a team as no one gets it right the first time immediately. And if you invest in measurements before optimization, then you enable basically learning from steps in the right direction as well as steps in the wrong direction. These you can test every hypothesis. So measurements basically enables continuous improvement as you cannot improve when what you cannot measure. And it's not as glamorous, but every successful pricing transformation I've seen focuses on call measurement first, so you can understand whether something works well or not and how well it works. Secondly, there is a lot of data already, right? Most airlines sit on air on years of price variation across their network. And with the right causal methods, that's really a goldmine for elasticity estimation. And it's considerably cheaper to start with because you have the date already available for these analysis and many airlines are leaving this on the table in general, to summarise right, you don't need to bet on a single platform or a single programme. You you need to prioritise learning in a disciplined way from the data and experiments that you can run today. And the execution side right is becoming more and more mature. I think the teams have get serious about understanding elasticity or willingness to pay will build a real advantage over the ones that are still waiting for the perfect system. Speaker 1 Mark, I think first of all, thank you so much. I think we can, we can take really good outcomes of, of this conversation and and also really good advices for, for the ones that are listening to you. If I may put in some of what come out of our discussion today is the winners, let's say of of this journey, especially on what we talking about on the right offer and optimization and having the right price are the ones that are going to learn faster price smarter and continuously improve commercial decision. So what you were saying about, don't hesitate to experiment to try it out to understand what it goes right, what it goes wrong. Combine all of this data that you have already do the experience, try to understand how your customers behave. And then and then with that, you will have the the right basis to take a real time decisions and, and, and take the benefit out of that. So I think that's really, really important to make sure that we pass through to the to the audience that is listening to us, because it is important that everyone understands that this is not just a shift on turning on something new. Is is also something that it needs to be built and experimented and make sure that you learn with your mistakes so you can be better in a controlled way than. Disciplined way as you mentioned. So thank you so much again for joining. Thank you everyone that has been listening to us. Before I give you the word marks so you can say goodbye to, I just want to say that you can find for the ones that are listening to us, you can find more thought leadership from travel in motion and our guests on the on about airline retailing commercial transformation on our website or in our LinkedIn posts. So look for us. We, we have a lot to share with you. And if you enjoy this episode, please follow us, follow Tim cast on Spotify and share with your colleagues. It's important that we create awareness so we can move forward with this transformation. And Mark, again, thank you so much for joining me and and share with everyone your knowledge on this topic. Speaker 2 Yeah. Thank you so much for the invitation, Katherine. I'm very happy to be here. Speaker 1 Thank you so and I will say with this goodbye and until next time, thank you everyone.