Innovation Roadshow - FLYR Labs

Transcript

Dominic Matthews:Good afternoon everyone. Special thanks to CAPA for allowing us to be here. Of course it helps for a sponsor. I've got, I've got 10 minutes to take you through our product and vision at FLYR Labs. 10 minutes as a comparison, that's about the average injury time in a World Cup game nowadays, so Hopefully this can be just as exciting. Any Polish people in the audience will probably know that 10 minutes is a very long time if you were watching the game last night. So a little bit about FLYR Labs. So we're a 400-plus people data science company based out of the US and Europe. We've been helping airlines to optimize revenues through our artificial intelligence-based revenue operating system. As we've heard this morning, a lot of talk about the disruption obviously caused by the pandemic. But also, I think one thing that stood out for me when the gentleman from Air Serbia said, we ignored the forecast and we just reacted. I think that was basically how airlines had to operate when it came to making decisions on network and on pricing, etc. And clearly that was necessary, but that's not sustainable, that's not scalable. So we believe at FLYR that our mission is to apply the most advanced and intuitive technologies that help transportation leaders unlock their ultimate potential. So let's talk about that potential. So McKinsey commissioned a study— this was pre-pandemic— that showed the estimated half a trillion dollar opportunity in value that can be unlocked in the travel industry alone. Through using AI, artificial intelligence. And 90% of that was in marketing and sales use cases, whether that be optimizing which product we should put in front of the customer, improvements in customer service, churn reduction, less cancellations, etc., and very importantly, optimizing pricing and, and promotions to customers. And Artificial intelligence has a real unique opportunity because with the latest technologies that we can embrace, they can— we have this continuous improvement cycle. So let's take an example. We sell flights through various different digital channels. We create a huge amount of data, and we can use that data. We have the computing capability nowadays to actually store that data, to feed that data into a machine learning model, and And we can create predictions. And we can predict consumer behavior. We can predict demand. We can see trends. We can predict willingness to pay of customers. And once we can use this technology to produce better offers, more appropriate pricing, more adaptive, reactive revenue strategies, we increase conversion. We create yet more data. And the machine learning models can get even better, even more accurate. even more optimal. That's the opportunity with, with AI. We create this flywheel look or virtuous, virtuous cycle. Let's talk a little bit about terminology. People throw around terms AI, artificial intelligence, ML, machine learning, deep learning. So artificial intelligence is not new at all. It's been around for over 50 years. A simple rules engine can be considered AI, but that requires a human to basically say, if this happens, do that. Machine learning is much more advanced, where we're using historical data to allow the machine to work out its own behavior. And deep learning is an extension of that, where we can really take into account a lot more data. And we use what we call artificial neural networks, which are designed to mimic the most impressive machine known to man, which is the human brain. And it's only really been possible in the last 5 to 10 years due to the massive improvements and advances in cloud computing, which enable us to run these kind of processes. So airlines operate in a very complex environment, as you know, and what's really important for us is to understand how the decisions that are taken, whether that's human decisions or automated decisions, how those decisions impact the outcome. And every day you're taking thousands of decisions and there's thousands of outcomes. You could consider that the thousands of prices and thousands of flights, etc. We have to understand the correlation between the decisions and the outcomes to be able to optimize the decisions which will lead to optimal outcomes. So we use technology which is very similar— We'll just go back. which is very similar to that in self-driving or navigation tools, self-driving cars and navigation tools. And I really like that analogy of a navigation tool because it's important not only to understand where I am now, what I need to do next, but to anticipate what's going to happen further down the road. And that's exactly what we're trying to do at FLYR in terms of improving the forecast. I use the F word a lot here, forecast. It's super, super important. The people of I've lost a bit of faith in that, and we're trying to bring, bring back confidence in being able to accurately forecast what's going to happen. And we're in excellent company when it comes to leveraging— click, click— leveraging deep learning. Companies such as— voilà— companies such as Google, Netflix, Tesla, etc., are all leveraging deep learning. And the great thing about this is a lot of this innovation happens in the open-source community, so we can leverage the thousands of engineers, the expertise that these, these big guys have, and apply that into our own applications. And so how do we do this in practice for airlines? Well, quite simply, we take all of the commercial data that's available and we put it into one data model, and we try to understand the patterns and the correlations between the decisions and the outcomes. And we, we try to— we produce effectively 3 things: highly accurate forecast of revenue, various aggregation levels, a highly accurate forecast of load factor, and an optimal price that we should put into the market today to optimize revenue, to maximize revenue. And we're always looking at the context, whereas other systems may look at the past to predict the future, year-on-year benchmarks, things like this. Deep learning is much more advanced than that. We look at the full context of the network. That means we can even provide a forecast and an optimal price for a market you've never flown before because we will be able to find patterns in other markets and similarities to be able to, to predict that. And over time, the model will just improve and improve, as I mentioned before, due to the AI flywheel. We call this the revenue operating system where we produce quite simply accurate forecasts of demand and revenue, real-time revenue optimizing decisions with escalation to analysts where required. We recognize that a system that can run on autopilot should not always run on autopilot and requires a close, a close collaboration between analysts and leadership and the platform itself, and all commercial data and KPIs within one easy-to-use user interface. The platform is underpinned by 4 key pillars. As I mentioned, a very robust data platform, extremely accurate forecasting capabilities, which is sort of the linchpin behind the capabilities, Pricing that reacts extremely quickly to changes across the network, changes with competitors, changes in capacity, etc., and the intelligence platform to be able to present that information to analysts, leadership, and other decision makers across the, across the airline. And we don't just stop at revenue management or flight revenue management. We really believe in the journey to total revenue optimization, whether this be optimizing of ancillaries where we see a huge potential. Many of you doing— many of you are doing very little in terms of optimizing ancillaries. We see a huge potential there. Fare family optimization, offer optimization, helping airlines to present the best offer to customers through their channels. We're even investing in cargo revenue management as well to give that total view of revenue to airline leadership. And of course, I talk a lot about outcomes. Well, the most important outcome is, is on the revenue side. So these, these are real figures. We have demonstrated substantial revenue uplift due to adoption of our technology, load factor improvements, improvements in forecast accuracy, and quite interestingly, a reduction in the need for analyst influence. So freeing up staff to focus on more strategic discussions or to focus on the markets or the flights that require the most attention. And this isn't just theory. We actually perform a real-life A/B test. So we will run the FLYR platform in parallel to your incumbent systems and processes, and we'll run a competition on both to measure revenue uplift, forecast accuracy, or any other KPI that would make sense for the experiment. So what's it like working with FLYR as a partner? So in the last 12 months, we've raised a substantial amount of capital. And what this has meant that we can— we've scaled up our engineering teams, our data science teams, and also our delivery teams. We can offer an extremely fast implementation where we can supplement your resources to be able to go through that data gathering exercise. And then we can also scale up and scale down. We continuously improve our platform. We ship new functionality every 2 weeks. It's deployed seamlessly. There's no version upgrades or anything like this. Deployed seamlessly to our, to our partners. And we really act more as a partner than a vendor. And this is really baked into our commercial model as well. So we offer a free implementation in just 3 to 4 months. Yes, I said free. That's zero cost. We'll even subsidize your project teams. If resourcing is a challenge, we can put our resources on the ground to help you with the adoption. You can walk away if you're not happy, if the performance is not proven. We really put our money where our mouth is. And we offer a performance-driven commercial model where, quite frankly, we only— we feel we only deserve to get paid if we demonstrate quantifiable revenue improvement for the, for the ally. I'm actually being told, Red, I'm 50 seconds over time. So thank you very much. If you'd like to hear more, then please get in touch. Thank you, Dominic.

Copyright policy: All transcripts on this site are the copyright of CAPA - Centre for Aviation. Our reproduction policy is as follows: you may quote up to 400 words of any transcript on the condition that you attribute the transcript to CAPA - Centre for Aviation and link to the original video page. All other use is prohibited. While we aim for 100% accuracy in the transcript, there may be some minor transcribing errors.