Willingness-to-Pay: Leveraging AI to Liberate Analysts from Manual Availability Rules Management
Forecasting based on price sensitivity instead of manual availability? View this session to learn the A to Z of willingness-to-pay forecasting. Hear from PROS Senior Director of Product Management Justin Jander, Hawaiian Airlines Manager of RM Systems Justin Mathew, Air Canada Senior Manager, RM Science, AI & Innovation Caroline Dietrich and PROS Consulting Manager Silke Langkitsch on how this innovation helps combat buy-down and drive incremental revenue while laying the foundation for class-free revenue management, airline dynamic pricing and true offer optimization.
Full Transcript
Justin Jander: All right, we're going to get started. And if you've been wondering and you followed me through the last several sessions, this is my last one. And so you get to be done with me for the rest of the conference. I'll still be around to talk if you actually want to talk to me. That's fine too, but no more on stage time. So the next session we have up is on willingness-to-pay. So you've heard that terminology a lot, heard it talked about a lot. And we've got three great experts on willingness-to-pay here to talk to you about that. And so really what we're focused on is more willingness-to-pay as a science. We've talked a lot about that. We understand the science. We've covered that.
Justin Jander: And if you haven't seen it, we're more than happy to give you an overview of that. But the more important thing that we want to talk about today is what it provides from a user perspective is really where we're focused on. The value that it provides as you move from the manual strategies to forced demand to buy up to higher classes and replacing that with automation. So that's really the key component of what we're focused on today. And so we've got a great set of panelists to talk to us about that. And the first step is always an introduction. So tell us a little bit about yourself, who you are, where you work, all of those things.
...
Caroline Dietrich: Caroline Dietrich, working for Air Canada. I joined Air Canada nine years ago and I'm taking care of the revenue management system. So both in terms of steady state but also its enhancement along the way. So including willingness-to-pay. And in addition to that, we're also responsible for some labs to enhance user practice, user day-to-day accuracy also of the different metrics that we're responsible for. So that's it. Glad to be part of the panel.
Justin Jander: Thank you. And we have a full set of Justins on the stage today. So Justin, introduce yourself.
Justin Mathew: We do. We've got Justin quorum here today. So aloha everyone. My name is Justin. I'm from Hawaiian Airlines responsible for the revenue management system with the team that we have here really supporting our revenue management users and experts. I joined Hawaiian Airlines roughly a year and a half ago. And if you're wondering, I do not know how to surf. I also do not know how to swim. My swimming instructor told me that I'm an expert at drowning. So it is still a work in progress but great working with a great team and also a relatively new system for Hawaiian as we can talk about. So really getting the team involved with how we use the revenue management system, how we think about revenue management in general, and making sure that we are keeping up with the industry so that we as Hawaiian Airlines can serve our passengers well.
Justin Jander: I really do appreciate that the Hawaiian team really represents well. You can always spot the Hawaiian Airlines team. They do wear Hawaiian shirts. And so it's very easy to see.
Justin Mathew: Believe it or not, it is company uniform. So our suits are aloha shirts. This is a dressy for us.
Justin Jander: Nobody wears a suit in Hawaii anyways, right? It's too hot.
Justin Mathew: Way too hot.
Justin Jander: And last but not least, of course, is Silke from the PROS team. So Silke, if you can introduce yourself.
Silke Langkitsch: My name is Silke Langkitsch. I'm a consulting manager at PROS for six years now. And before that, I spent all my working life with airlines in Switzerland and Germany in revenue management. So all these airlines had been PROS customer, and this is how I ended up here.
Justin Jander: And we're excited to have the perspective today of, obviously, our airline partners, but also Silke having worked with many different airlines implementing willingness-to-pay, so bringing a different perspective as well across the board. So again, thank you all for being here. So let's dive right in. So I think the first question is why willingness-to-pay? So why is that even a priority for you as the airline? And what is it now? What about it, what's the right timing now? We heard Peter in the previous session say this has been around for a while, but what's the motivation these days now? Sorry, we'll start with Caroline.
Caroline Dietrich: No, that's fine. I can start. Well, actually, if you had attended the session just before, the question was already partially answered. Well, willingness-to-pay, or at least the principles behind willingness-to-pay, have been existing for quite some time. And also, we're moving towards a less restricted first structures for which we have dilution, we have buy-down. And hence, we're having a forecaster that's going to take care about those effects is necessary. In addition to that, class-based forecast is a simplification of the reality. And it's definitely not how passengers are behaving. And so having something that is already known as being sub-optimal has been an aspect that we're very happy to move away from.
Justin Jander: So Justin, when you think about that answer and add to it, when Hawaiian implemented willingness-to-pay, what were kind of the goals behind the, what were you looking to achieve with that?
Justin Mathew: Yeah, absolutely. For us, continuous curves isn't something that's necessarily new to us. Willingness-to-pay wasn't necessarily new. Our old revenue management system that we actually switched from had a different way of trying to solve a similar problem. Because if you think about Hawaii and you think about the little dot that we occupy in the Pacific Ocean, this buy-down nature in really one of the most leisure premium markets in the world has always been something that we've thought about. So willingness-to-pay and that forecaster really was a core consideration for us as we were thinking about revenue management systems and really evolving our practices.
Justin Mathew: So as we switched towards PROS and we implemented the willingness-to-pay forecaster, we wanted to make sure that first and foremost, we were trying to get away from really, as Caroline mentioned, the dilutive nature of a lot of the fares that exist in an unrestricted fare environment. So it was a core consideration for us and definitely something that we wanted to make sure continued forward with our revenue management mindset.
Justin Jander: So Silke, if we add to that, what have you seen when you went into airlines where they weren't using willingness-to-pay, what sort of things were you seeing that were challenges that they were facing that willingness-to-pay was expecting to solve?
Silke Langkitsch: I would say there are quite a lot of challenges and everyone who implemented WTP would most likely agree to that, but there are three major challenges. And really break it down to the analyst perspective. The first one is the O&D level. So the detailed knowledge of O&D level that is needed to assess the WTP forecast. So in most of the cases, this detailed level is not there. So because analysts are still on the lag or they just know an O&D high level, but they don't know the passenger segments by day of week, by TOD, by DCP. So this is one of the biggest challenges.
Silke Langkitsch: The second one is the willingness-to-pay or the independence of the willingness-to-pay and the book load factor. And this is a very common practice we see everywhere in the world that the book load factor is used to force a sell-out. So what is the issue with that? Willingness-to-pay happens or takes place on the O&D. So when you close a class on a certain book load factor, the class is closed for all tourist O&Ds. And all of these classes have a different value. And all these O&Ds have passengers that are traveling on these O&Ds have a different willingness-to-pay.
Silke Langkitsch: So you're completely ignoring these revenue potentials. So this is seen very widely. And the third major challenge is going away from looking into a single class or assessing the forecast on a single class, but now going into a class block and assessing the shape or the distribution of the classes within a class block. So getting, this is also a journey that is quite challenging for most of the analysts.
Justin Jander: So let's double down on the user side of things. I think that's where we want to focus a lot. So when you think about the challenges that, we'll start with you Justin. When you think about the challenges that users faced, what are sort of the things that when you put it in place and I mean, maybe one of the things you asked them was take off some of the rules that they had in place to see how willingness-to-pay behaved. That's a lot there. Can you tell us a little bit about what the reaction was and how the users reacted to that and what the challenges were there?
Justin Mathew: Yeah, absolutely. I mean, for us, this was a challenge that was compounded by the fact that it was a new revenue management system. So first and foremost, as we were moving between different solutions and really some of the challenges that we had in the past was that we had a forecast that we couldn't really rely on. And so we had a general sense that we had to be overriding all the time because we couldn't trust what we had before. That takes a really long time to get away from, especially when we're having to relearn what a forecast actually is and I think Silke very well put, getting away from thinking class to now class group is exceptionally different when we all learn a revenue management of some shape or form starting with classes. And that's something that we had to learn because customers don't think class.
Justin Mathew: When we buy anything else from anywhere else, we're not thinking classes. And so why do in revenue management? It's because it's our history. So that really has been something that we've been trying to get away from really thinking at the class group level, really thinking away from rules. And it's not something that comes overnight. It's definitely something that we are working on. But the more we are able to understand, the less rules we need to be able to understand what a forecast does and essentially how we want to optimize our revenues.
Justin Jander: Caroline, anything you would add there from your side?
Caroline Dietrich: The fact that you want to work with the revenue management system, not against. And well, also something that we have to acknowledge is the fact that even currently with the traditional forecaster, you may have less than optimal practices, which you have to combat and go beyond and adopt willingness-to-pay. So it's also some of those practices that you have to take care of at the same time. And then also for us, what's helping tremendously is the visibility of the outputs, the different outputs. And actually yesterday, some of them were presented. So in the UI, the alpha visualization, eventually also transformed fares. That is also something that we needed to gain confidence and to better understand what was happening. And so that, well, what can we trust and what can we help so that it works to its best.
Justin Jander: I think that's a key point, right? Nobody's suggesting that the users aren't necessary anymore. Obviously, that's not at all what's being said. What we're looking to do is automate some of the decisions that have been in place and remove those rules that were there to give the analysts more time to focus on real revenue opportunity and complementing the system instead of replacing the system. Obviously, there's an element of that. So we kind of continue on this track. Silke, when you went in, how familiar are analysts with the idea of willingness-to-pay buy-down? Is that something, when they use the load factor strategy or a day's prior rule, are they thinking of it as buy-down or is it just someone told them to do that and they just do it? Maybe I'm leading the question there.
Silke Langkitsch: Yeah, no, no, I guess you're right. So as I said, it's a very widely seen concept to use book load factor to force a sell-up. And of course, there's a broad idea to sell-up because everyone has noticed that the bid price is not selling up. And this is the easiest way, and I can fully understand it. It's easy, it's quick, and it is most likely many people are familiar with that because they know it from the time of the leg optimization. But it has nothing to do with the concept that WTP is following.
Justin Jander: How about you, Justin, on your side? What's the kind of connecting the dots between why they're closing a class versus the concept of willingness-to-pay?
Justin Mathew: It really comes down, at least for us, just about understanding, right? Understanding and being able to challenge practice. Something that we've had to work on, and really, again, because willingness-to-pay is fundamentally different, the way that you need to approach it has to be different. Going back to basics, as has been mentioned a lot during the past couple of sessions that we've had, of what is the forecaster actually doing? What is the value of the forecast? When it comes to willingness-to-pay, what is the value of a transformed fare?
Justin Mathew: Why are fares being transformed? Why is demand being transformed? What is the effect of that? Being able to better understand what's happening allows and empowers users to be able to make adjustments that they need when they need to do it, and ultimately be able to get back to a lot of their work that they love to do, which is understanding customers and understanding what's happening in our markets.
Justin Jander: That's a great point. I think it really gets to understanding the behavior rather than, like, it's nine days before departure. I need to close Q class. I mean, none of that matters to a passenger, right? So, Caroline?
Caroline Dietrich: Yeah, to add on top of what Justin said, well, buy-down is a phenomenon that is, well, it's passenger behavior, but still, the quantification is the real challenge, as it's extremely difficult to know if I'm opening a class, if I'm going to dilute more than I'm actually going to stimulate. And hence, that's the power to us of willingness-to-pay, because at least we're going to give analysts that visibility about what to expect, and does it make sense to, well, have this type of availability or steering mechanisms over time. And after that, the buy-down effect combined with low load factor, overcapacity that's ended up with, like, spiral down in a traditional environment is definitely a challenge to understand for the analysts.
Justin Jander: That makes sense. So then, I think the natural next question is, how do you convince the analysts to stop using those rules that they've had in place forever? Justin, you talked about it being education is a big aspect of that. Anything else that kind of goes into it? Is it a trial and error or something along those lines? How does that work?
Justin Mathew: Yeah, we see it less as convincing and more about encouraging our analysts. It comes down really from the start of setting good examples of having our leadership make sure that they're promoting the right message and guiding our users. We take it by example. Our directors, our managers, they've had the opportunity to get their feet wet, to get their hands dirty, which isn't always common, to go in and actually actively look at forecasts, actively work on forecasts to be able to see what is happening. And we've had great results of the same leaders who are providing guidance and instruction, going in and actually showing with results of what is happening and what's the benefit of actually trusting a forecast and letting willingness-to-pay do what it does best.
Justin Mathew: Oftentimes it's scary, especially when we think of this new method and model, but ultimately it's scary because we are replicating what we would have done in other places, oftentimes in more aggressive solutions, with now something that's a bit more automated that allows us to be able to do other beneficial activities.
Justin Jander: Yeah, the term practice what you preach, I think comes to mind there, right? To be able to confidently speak about how to use it really fits in with using it, right? So practicing what you preach. And so Caroline, anything you would add there from when you're thinking about how you would have the users start turning off those rules in place?
Caroline Dietrich: Yeah, I think we need to consider that it's a leap of faith. So we'll have to take risks and if we don't, there's nothing that we can see and learn from. Also something that we realize is, well, that actually we know, is the fact that if you have positive revenue benefits to show, it's going to be the greatest help you're going to have to get the adoption from the users and the stakeholders. So the revenue evaluation method, its strengths and its accuracy is key for anything that you want to do.
Justin Jander: So I mean, notably one outcome of willingness-to-pay can be lower load factors because if you have users in place that even have an eye towards load factor as an objective, then they may open up more than they need to allowing to get higher load factors. When the tool comes in, doesn't care about load factor, its interest is in maximizing revenue. So that's a big factor when you see that, to think about the scary part in the leap of faith, that somebody has to trust you that you're going to see a bit of a decrease and potentially a bit of a decrease in load factor, but the gains are on the other side. So Silke, when you're convincing these analysts to remove the rules, what are you going through? What's the process that you use for that discussion?
Silke Langkitsch: So I try to step away a little bit from the scientific background of the WTP forecast because the experience I had is that analysts are sometimes scared about these alphas and lambdas and the class blocks. So there are so many new things that are coming with WTP that people step away and are afraid to do something. So what I started to do and that is, I have to admit I also had a learning curve like all of us here at PROS the last two years with WTP, I guess. So I focus on what everyone is used to and familiar with. So these are booking intakes and what else, and the time frame. So what we are doing is we forget lambdas, we forget alphas and we are looking into the outputs, so the transformed fares.
Silke Langkitsch: And usually, so in traditional we always say, oh don't look into the outputs because the output is a result of a good forecast. In this case we say no, start looking into the transformed fares. So we take in an O&D, break it down to day of week TUD, when analyst knows, well, a passenger segment, the booking behavior, the willingness-to-pay. And what we do then is we are looking into the transformed fares from the first to the last DCP and we see how the transformed fares are changing for each booking class over time. And this tells a story. So when is the booking class increasing or is there maybe a wave? So and what we are doing is we are looking into these DCPs and into the transformed fares and we try to read the story, what happened.
Silke Langkitsch: So in the early DCPs most likely the transformed fare is very high, so it means all the classes are open. So we are familiar with that. So and then we see over time, so the transformed fares are decreasing, willingness-to-pay is increasing. The first classes get inefficient, means the class is closed. So it looks very familiar and what then happened, the analysts are saying when I ask, so is that what you expect, is that what you do? Oh yeah, this is what I do. And then there are the other cases where we see waves or blocks. And then yes, some customers are familiar with these blocks and waves. And we went through this journey. And then it's important that we find out why is the system, for example, decreasing, again, the transformed fares and in fact open up again.
Silke Langkitsch: So what happened? And the simple answer, and it's really simplified here, so please to all the scientists, apologize that, but this is really to bring that to the user and to understand the concept behind WTP. So why is it opening again? Because it didn't receive observations. And then why didn't it receive observations? Maybe in this DCP there is simply no demand. But then we come to the user interventions with all the settings we know. We have an availability setting, maybe the book load factor was reached, the class was closed, but it was too much for this O&D. Maybe a minimum bid price was set.
Silke Langkitsch: Maybe the flight was too soon to full. So there was no chance to book anymore. And these root causes need to be identified because this is the only way how you can assess if the forecast is right or not and if you can be happy with that forecast or not. And you can be confident. And it's not about to tweak it immediately, it's really more work on the root causes first and then gain the trust. And I have to say this way was really, and then of course we go back to the forecast and then we look into the class distribution, DCP by DCP into the demand curves and we see, okay, what we have seen, what we have analyzed, how is it looking in the system? Because the overall goal is not looking into your transformed fares everyday, the goal is look into your forecast. But for that you need to understand the principle. And this is how at least I guess some are convinced.
Justin Jander: So I think what you're starting to hit at is you go from the initial assessment, getting comfort, but then that sort of spills into best practices and creating those and that's where you kind of went next. So Justin, I think that's a good segue then. So what would you say, kind of how did you establish best practices with the users under the new construct? What did that look like for you guys?
Justin Mathew: It's definitely a work in progress for us and one that I'd say is something that we didn't have much of a choice in. We had to be full on adopting willingness-to-pay because it was a new system but also fundamentally our entire network is using willingness-to-pay. So because of it, because of our liberal use, we wanted to make sure that we're using it properly. And that comes down to, again, understanding what's happening, being able to know when to adjust, how to adjust, ensuring that we've got training in place and this is something that we are continuously working on with different aspects of the system, also different aspects of usage.
Justin Mathew: Making sure that fundamentally we are listening to our users, we are listening to the people who are in the system day in, day out because we won't be able to make effective change if we don't understand what's happening. So fundamentally we invest a lot in training and also making sure that we are adjusting in the right way at the right time. Knowing when, how, what to use and especially with some of the changes that have been made in the most recent version, thank you, Rachel, allows us to be able to better manipulate the system and the forecasts so that we don't have to resort to alternatives that are not good for overall system health.
Justin Jander: That makes a lot of sense. So Caroline, I think under your purview is best practices, if I recall correctly?
Caroline Dietrich: No, but that's all right, I can still talk about it.
Justin Jander: You're expert in it anyways, I think.
Caroline Dietrich: Thank you. Well, for the best practices, the involvement of users along the adoption and the transformation is key. Creating champions, having already them included in already even the training preparation and so that as we're all learning together, it's something that's completely new and that we don't know how it's going to behave in all the parts of the network. Still, partnering and having them asset as champion will be key to build proper best practices but then also to convince the rest of the community to follow along.
Justin Jander: So that makes sense. I think the one thing we've heard pervasive across the discussion this week is dynamic pricing. Caroline, can we start with you on where do you see the tie-in between willingness-to-pay and dynamic pricing? How do you see those things fitting together?
Caroline Dietrich: So at Air Canada, we have a transformation plan that will last for several years that includes willingness-to-pay and dynamic continuous pricing but goes also beyond that considering ancillaries, considering offer generation and beyond. And so the two modules are complementary. Doesn't have to be first willingness-to-pay or the other one but still, it's clear to us that they will work together and they will also multiply the benefits together. So we're already thinking continuous with willingness-to-pay and so going to a continuous price is just a natural step after that.
Justin Jander: And Silke, what would you add on the dynamic pricing connection to willingness-to-pay?
Silke Langkitsch: Yeah, the prerequisite for optimal prices that are the result of the dynamic pricing are the transformed fares and the transformed fares are the result of the WTP forecast and of course, the bid prices. So the prerequisite is to have a really stable and reliable WTP forecast that is not interrupted by user adjustments that should not be in place. So the better the quality of your transformed fares, the better the quality of the optimal price that is calculated by the dynamic pricing.
Justin Jander: Justin, any words on that?
Justin Mathew: I'd say fundamentally, it comes back to building foundations. Willingness-to-pay and at least the way that you think with willingness-to-pay, the mindset that comes with willingness-to-pay transfers very well when we think about continuous pricing and going to a continuous world. So fundamentally, if you're moving your mindset there, we'd hope that the systems and everything else that comes with it will come.
Justin Jander: Well, I think that's a perfect way to wrap things up. I think as we look at the previous session where we talked about going from revenue management to offer optimization, willingness-to-pay is a fundamental part of that and we just tied the picture together of how that step gets you to continuous pricing. So with that, thank you all very much for participating and I'll speak again on behalf of all of our panelists. They will be around for the rest of today and tomorrow. I'm sure they would love it to be, to talk about willingness-to-pay at tonight's reception or maybe just wait until tomorrow, either way. But I'll leave it up to them to decide if they prefer to talk about work or not, this evening. But again, thank you all very much and we really appreciate it. So thank you all.
[applause]