View transcript : Trainline Fireside Chat with Deutsche Numis

Company

Trainline

Date

14 January 2026

Time

14:30 GMT

Corporate Participants

• Pete Wood – Chief Financial Officer, Trainline

• Mike Hyde – Chief Technology Officer, Trainline

• Nina de Souza – Chief Product Officer, Trainline

Host

• Nick Stockman – Head of UK Equity Sales, Deutsche Numis

[START OF TRANSCRIPT]

Nick Stockman

Great. Hello, everybody, and welcome. Thanks very much indeed for joining this call with the

Trainline Team. Happy New Year to those of you that I've not yet seen. Delighted to be joined by

Trainline's Chief Technology Officer, Mike Hyde, and Chief Product Officer, Nina de Souza, as well as Pete Wood, CFO, sat in the middle there. Thank you all very much indeed for joining.

I think everyone's aware, but just as a reminder, in terms of the format of the call, that we’re going to do this as a fireside chat approach. So, people's mics will be muted, and I'll put all the questions to the team. If you do have a question, do please feel free to try and get a hold of me, and I'll try and get it into the conversation. You can send me an email, [email protected].

If you use Bloomberg, you can find me on there and send me an IB, or there's probably the chat function, which I can keep an eye on as well on the Teams link.

So, that's the way to get hold of me. As I say, we'll try and cover a fair amount of ground in the time allotted. Without further ado, I just wanted to hand over to Pete, who I think was just going to make some introductory remarks before we get into Q&A.

Pete Wood

Great, thanks, Nick, and it's great to have you, Mike, and Nina here today as well. If I think about the themes that we've had with investors over the last couple of months, I think GBR and AI are two that stand out. I'm sure we'll touch on both today, but it'd be great to get into the tech and the product side of things as well.

Just as a reminder on the AI, we have a framework that we've been talking to you, and we use internally, and probably helpful just to level set on that. I'm sure we'll refer to it as we go through.

There's the productisation, the use of AI in developing features and the like for our customers.

We've already set off down that road. There is a much broader ecosystem with Gemini, with

ChatGPT, and other LLMs that is unfolding. Of course, that landscape is important, and how we interface with it, and how that changes, too. Then finally, productivity is something that we're seeing run and unfold, and we're still learning fast there. So, those are the broad buckets that we think about, and I'm sure we'll get a chance to touch on all of those as we get into this.

Nick Stockman

That's great. Look, I think before we get into the meat of the discussion, it would just be worth getting a little bit of a background from Mike and Nina, because probably not as well known to a lot of the people on the call. You both bring unique perspective to the business.

So, can you just maybe briefly share with us your background, your professional journey which brought you to Trainline, and how you fit into the organisation?

Mike Hyde

Yes, certainly. Hi, everybody. Good afternoon. My name is Mike Hyde. I'm the Chief Technology

Officer for Trainline. I joined Trainline in September 2021, so just over four years ago now. I've been working in technology and in data science-type roles for about 20 years now in all sorts of different organisations.

I built out the data team way back for Skype, back in the day, Skype Consumer, when that was a thing. That was bought by Microsoft, and I did a bunch of work building out data science data engineering for Microsoft Office and Skype for Business.

I then spent four or five years at Facebook, as it was at the time, now Meta, of course, where I was building and running all of the data science, and all the data engineering investment that

Facebook had in London.

Then when I joined Trainline, I joined as the Chief Data Officer, and I spent the first few years building out Trainline's data assets and team. Then since, partly prompted by all the revolution we're seeing in AI, there was an opportunity to bring together the data mandate and the overall technology mandate. So, we brought both of those two worlds together.

I switched into being the CTO, so I look after now all of the data estate, all the technology estate.

So, that means all the engineers, including the data engineers, data scientists, and the AI investments that we have.

Nina de Souza

Hi, everyone, nice to meet you. I'm Nina de Souza. I'm our Chief Product Officer. I joined just

June in this past year, so about six months in. I lead our product organisation, working together with Mike and teams, and really looking at how we can translate our technology and data into better customer experiences.

Before Trainline, I was Chief Product Officer for TB Auctions, which is Europe's leading B2B auction marketplace. Prior to that, held senior product leadership roles at Deliveroo, where I was for six years, at Hotels.com, and Expedia, so really building out consumer products at scale and in high-growth environments.

I was really drawn to Trainline because of the problem that we're solving, and that rail is extremely complicated, and it can be an amazing way to travel, but it also can be really hard and stressful when it doesn't go to plan. So, that's really been coming together over the last six months. You may have seen us launch ‘The Way to Train’, which has been our biggest product release yet, and we're really looking at how we use data and AI to solve customer problems and provide more clarity and confidence. Looking forward to the discussion today.

Nick Stockman

Perfect, that's great. Thank you for that. So, our thinking was just before we get into the tech debate, and the data debate in a little more detail, we're just going to touch on – Pete just alluded there to a lot of the GBR ongoing discussions and questions that we get.

So, it might just be useful just to bring everyone up to speed on the news flow that we've had since November when you last spoke to the market in relation to that.

Pete Wood

Yes, of course, Nick. As a reminder, in November we had two announcements from the government just as we announced our results. The first was the legislation that is required to stand up GBR. That is much less about the retail space, but much more about getting that organisation stood up and actually operating the trains. That is running its course through parliament. There's not really much more to say.

Then the other was this Code of Practice, which essentially is how GBR will compete with other retailers in the market. Again, there's not much more to say then that a process has started, but nothing more to share.

Then, probably two other things in this space to just provide some colour on and update on. The first is that, as part of the budget, the government announced that they would not be pushing through a regulated fare rise this year. This is about just less than 50% of the fares are regulated typically, and historically, we've seen the unregulated fares move in lockstep. So, the base case here is that we won't get any inflationary pressure on fares this year. If I look at consensus, I think that's now being washed through and is reflected in the numbers out in the market.

Then the second is how the government is looking to stand up a website and app. Right now, it's announced a market engagement. We are engaged in that. This is something that happens before the procurement stage to draw out interest and get some input from the wide set of stakeholders in this space.

It's not clear or announced exactly what happens next, but to offer an opinion, we might expect procurement to then follow. That might take 12 months. There will likely then be a period of build before a launch, which could be, I don't know, 9, 12, maybe 18 months. Then once this is stood up, the government has already said that it wants to shut down the other websites and apps that are out there. So, that's probably three years away, if you add all those dates up.

Again, speculating on the dates that haven't been announced, but to offer an opinion.

Nick Stockman

That's great, and we may come back to talk about GBR tangentially through the discussion around tech and the USPs of your business over the course of the discussion. I think that's helpful as a starting point.

Maybe just blending the two a little bit, you made the point there about how long it's likely to be until there is a GBR app, and indeed, website. What's your current thinking around that as a genuine competitive threat? Obviously, there's been speculation historically that you guys might even be involved in a white label capacity in that regard. What do you think you can currently say about what sort of a threat that poses?

Pete Wood

Of course we have a white label business, and we'll look at interest if a procurement process comes our way, but to park that and talk about the mint green business model, we will continue to innovate, and we're confident with how we will show up against GBR.

It's going to take some time before that launches. They've got a variety of work to do to actually reach that stage, and in the meantime, we will continue to invest and innovate, and to grow the gap, if you like. So, we're ready for when they stand up, and we won't stand still in the meantime.

Nick Stockman

I just had somebody asking, and so I might as well ask it now, just around if you've got any comment around the existence of booking fees, which obviously forms part of Trainline’s revenue and how that's likely to play out in the eventuality that that's not a part of GBR from a competitive standpoint.

Pete Wood

Yes, and I think it is unlikely that GBR will charge booking fees. I don't know that for sure, but that's our working assumption. Of course, that is true of all the train operating sites today. None of them have booking fees; they’re actually not allowed to charge them.

Then I look across to Uber, who entered the market, what, four years ago or so now. They not only didn't have a booking fee, they're actually pushing credits out the door. So each train ride you’ve bought, you've got credits back that could be redeemed against rides or eats. They've taken a bit of share, but it's capped off at a few points. That's really been a good proof point for us on the resilience of our model, the trust and the value that people have in using Trainline.

So, that's not to say that we won't look to explore how we'll continue to evolve that model over time, and those of you that follow the stock will know that we've launched insurance and hotels, and there are other verticals that will look and change the mix of monetisation. That's kind of how we think about booking fees and the model at the moment.

Nick Stockman

Yes, that makes sense. Great. So, let's dig in then to the technology side of things a bit more. So,

Mike, maybe just to start with you. Can you bring to life – and appreciating that you've generally got generalists on this call, I would say, rather than technologists – the relative strength of

Trainline’s, tech stack versus what else is out there or other things that you've seen in equivalent businesses?

Mike Hyde

Yes, certainly. The way to think about Trainline's tech stack is that it's a full-stack platform.

That's the key differentiator. We're primarily about rail aggregation, but we also extended to e- commerce and a fulfilment platform all in one piece.

We're quite big. We're the number 1 travel app in the UK. We're the number 1 rail app all across

Europe. We have a single global platform and that supports both our B2C apps and websites, and it also supports all the B2B business that we have through our API and also through some white label apps and services as well.

So, all of that runs through this one single aggregation platform. Like I said, we go full lifecycle.

In rail, that is quite complex. So, it's all the way from customer acquisition through to search and discovery, e-commerce, fulfilment, customer service, self-serve, and we're even able to open the barrier with the app itself, which requires quite a lot of backend integration.

In terms of scale, speaking about the backend, we integrate with pretty much all of the rail and bus carriers across Europe. There's 270 different transportation carriers that we integrate with directly in the backend, through which we process about eight terabytes of data a day. At peak time, at the sort of 8:00 am commute peak, we're handling about 350 searches a second.

So, the platform really has to scale to be able to handle those spikes in the morning, in particular during that commuter rush hour. The other thing, as I say, is that real-time fulfilment is critical. Again, another big differentiator with a typical e-commerce platform is that quite often a

Trainline customer is rushing up to the gate, and people are often buying their ticket as they're approaching the gate, so they need the whole thing to work in the back end, the ticket to land almost real-time so they can unlock it and run again on the train.

So, the whole thing has to work fast. So, to do all this, we have about 600 technical staff across main software engineering, data science, data engineering, machine learning, AI engineering.

So, we operate at scale.

In terms of how we maintain all this, we have a evergreening approach, as most modern tech companies do. So, we're constantly modernising and maintaining elements of our tech stack.

We're also constantly looking at what the newest technology is to take off the shelf and integrate into our stack. So we're constantly modernising, innovating, and making sure that we're staying at the cutting edge of where we want to be.

Nick Stockman

Is it even relevant to think about what it might cost for somebody else to replicate what you've got?

Mike Hyde

I think it would take a long time. We've been at this a long time. If you look at the effort needed to replicate those kinds of 270 integrations, Trainline's been at that backend for many, many years to build up that kind of footprint.

So, yes, the integrations, the backend, and increasingly the AI investment, all of which take time to build up. The other major asset, of course, that sits behind all this, the size and scale of our datasets that also take size and scale to build up and to accumulate.

So, a lot of the features that we're able to build now are off the power of those unique data assets that we've got, and that comes from the scale, and the size of our footprint, and the user base that we've got.

Nick Stockman

Got it. No, that's great. I guess you've touched on this, but there’s been a degree of concern around a lot of the tech-led business in the UK recently that the advent of AI might result in a bit of catch-up investment required.

I mean, it sounds like you are already investing in that area, but maybe just to bring that to life in terms of your confidence in the sustainability of the current rate of tech spend without needing to raise it, and perhaps to bring that to life in terms of how you've reallocated resources towards things like AI from historical areas.

Mike Hyde

Yeah, no, certainly. As I say, the way I think about it is, look, ever since I joined in 2021, I've been building out those data assets, and the capabilities and the deep technical skills that we need to be ready to play and compete in the AI era that we find ourselves now. So, we're feeling good about that.

We think about the AI opportunity in front of us, as Pete said, broadly in these three buckets, and we actively play in all of them at the moment. The first one is our ability to build intelligent product, and to build features for our customers, and that's changed remarkably. We're able to innovate in ways that we just couldn't do a few years ago, and that's very exciting.

The second one is our opportunity to connect to this emerging AI ecosystem, the nature of the digital internet. We can talk a bit more about that. Then the third thing, like a lot of businesses, is it also creates opportunities to automate and create efficiency in terms of how we work, the things that we do as a business.

If I take the first part, building intelligent product, I mean, that's huge. We've had a machine learning capability, powering things like search and journey routing for years. In the last year, we've really stepped it up. We've built and launched our own agentic AI travel assistance for the first time. It's a first for rail, and it brings together all of the years of experience and domain knowledge and unique data sets that we've got, so that you've got the experience of a personalised expert rail assistant built right into the app that can help you when you're on your journeys.

That's leveraging that in-house expertise and those kinds of deep, unique data sets that we've got. It's even able to process refunds for customers now and start to get things done for people.

So, we feel like in our space, we're right at the cutting edge of trying to push that forwards.

Just to hit the other two really quickly, the AI ecosystem, yes, we are - that same AI intelligence platform that we're building, we're extending so that we can partner and work with third parties.

So, example, we're in conversation with OpenAI and others in terms of how our technology can integrate into these new chat experiences.

For the third part, productivity and automation, again, we can come back to it later perhaps, but we've rolled out ChatGPT Enterprise across the whole organisation. Every engineer at Trainline has got an AI-powered coding assistant that helps with it.

As an industry, I honestly think we're somewhat in the foothills of this modernisation piece, but we're excited about it. We're moving quite quickly, and we're starting to see real velocity and productivity wins from those investments already.

Nick Stockman

That's great. Thank you so much. Then you mentioned in your introduction that you were previously Chief Data Officer. Can you just talk, perhaps, about the uniqueness of Trainline's data? Because I think perhaps first-party data and its uniqueness, well understood to be an important underpin to successful use of AI, but maybe not something that's as associated with

Trainline in terms of that data set. So, it'd just be good to talk about that.

Mike Hyde

Yes, no, you're completely right. I think a lot of the modern AI experiences that any business can build are mostly as good as the data that powers it, really. Trainline's got some very unique data sets that we've been, as I say, building now for several years.

We think about it in three big buckets

the user, the trip, and the network. So, the user is basically our ability to understand the 27 million active customers that we have all across

Europe. We have a unique view of the rail-using public, if you like, who they are, what their desires are and how they show up and use the rail network.

The second bucket is trips. So we have this unique historical database of an enormous number of tickets and journeys that have been bought on our system and searched for, importantly, over many years. So, you can really start to understand travel patterns. What are the major corridors?

What are the major movements of people and populations? How is that evolving and trending over time? There's also a lot of insight to be gathered in these situations from search, for example. What is the demand for travel, and where does it come from?

Then the third big bucket, as I mentioned, is the network. That data set comes from the aggregation of all of those 270 integrations across carriers. So, we have probably the largest, most aggregated set of rail data, probably across Europe. That includes things like all of the timetables, the pricing, the real-time information, the disruption information, and so on. That really lets us build machine learning and AI models that deeply understand network itself.

Joining the dots between those three things, between the user, the trips they want to take, and that how the network itself is performing, is kind of our superpower. That's what unlocks a lot of quite unique AI ML power features in our product. So, we're excited about what we've done so far with that, and I honestly think there's more mileage to come from it.

Nick Stockman

That's perfect. Thank you very much, Mike. So, Nina, maybe just to build on that a little bit, then, in terms of how we can use these USPs of the business to drive product and product development. Can you just talk us through some of the specific examples, I guess, where you’ve used machine learning and AI within the product already?

Nina de Souza

Absolutely. So, we've been using machine learning AI for quite some time now. I think one of the best examples is with SplitSave. So, folks might be familiar, but every time a customer searches for a journey, our algorithms are checking whether we could create a cheaper combination by looking at splitting that journey up. So, looking at different combinations.

It sounds quite simple, but what Mike described about the complexity of the network, looking at that availability and pricing across all of the different combinations you could create, is actually incredibly complex. It's a really great example of where using AI and machine learning can create real value for our customers. So, really, the power of our scale, and the power of the technical teams that Mike and others have built.

More broadly, we're using machine learning across search and ranking. So, in what order do we show things? What do we recommend to the customer? Really helping them make choices, especially in aggregated markets. We use it to work out pricing, to work out personalisation, how do we personalise across app and web, especially as we get into selling things like hotels or insurance, when is it relevant, and even to power things like fraud detection, which benefits

Trainline and our cost structure.

It also really benefits the industry, and we're uniquely positioned to do that. There's a lot of interesting things around fraud protection as well. Then more recently with our Signalbox technology, which is our geospatial technology, we are using that to really power AI-driven disruption experiences. So, understanding the network and predicting what will happen is a newer area for us that we've launched with ‘The Way to Train’ and one we're really excited about for the future potential of AI as well.

Nick Stockman

Thanks, and yes, that's interesting because the disruption features, and the talk about delay repay notifications, I've noticed those have started cropping up, and it's been something that the team had talked about to investors for a while, that they'd love to be able to do more in that delay repay space. Can you just bring to life how those products have been developed and how well they've landed with consumers?

Nina de Souza

Yes, absolutely. So, ‘The Way to Train’ is essentially our AI disruption suite of features and really powered by all of the data and our scaled platform. You can think about three outcomes. So, there's anticipate, resolve, and support.

With anticipate, we have a feature called Travel Forecast, which is essentially us helping customers understand when their trains might be delayed or might be cancelled, giving them real-time notifications for that, showing on a map where is their train in relation to where it should be, and really trying to improve customers' confidence about what is happening to them.

So, it's a feature we're really excited about. I can't overstate how this is extremely differentiated in UK rail, being able to predict instead of just reacting to what is happening and telling customers once the train is delayed, providing that information in advance is something that we're uniquely positioned to do, and something we're really excited about. The 18 million customers we have in the UK are going to make that technology better over time. So we're constantly learning from our customer data and from our journeys.

The second area, resolve, that's delay or repay, as you mentioned, being able to notify customers when they're owed compensation, being able to estimate what that compensation is, and directing them to the TOC website to then fill in that form for compensation. Now, this is an interim measure until the industry gives access to third-party retailers for automatic delay repay, but we are doing a really nice job in the product of making it easy enough for customers to copy and paste into that experience at the moment.

Then the third area is support. So, the AI-powered chat assistant that Mike mentioned, enabling for our customers an experience where the AI assistant is journey aware, it knows what the customer is doing or intends to do and can provide personalised recommendations and support.

Those are three specific features, but they're really more than that. We believe they are a part of our competitive moat, and taken together, represent an end-to-end disruption experience that is unlike anything else in the industry, and can continue to get better from there. So, yes, really excited about that.

From the customer launch and the feedback, it’s been really positive, looking at usage and engagement stats, as well as our qualitative feedback from our customers. It is a newly launched feature, so we will continue to see how that plays out as we expect it to show up in retention and frequency of our customers as well.

Nick Stockman

That's great. Maybe just tying into that - and this might be for both of you to have a word on, but tying products such as that back to the importance of a powerful data set.

If you think about launching those disruption features into your European markets, where you've obviously got smaller data sets at a relatively early stage, is it possible to launch them there, or will you have to wait until we've got more scale, we've got more data, we've got more history for that to be really effective?

Nina de Souza

Yes, it is possible. I think it's true, Nick, that more data helps, especially for the predictive features, and so we have that data and that scale in the UK, and that's why we're investing there in the UK.

There are other features like the delay, repay, support, and the chat assistant that are relying less on scale. We see Europe primarily as an opportunity right now to improve the product from an aggregation perspective. So, what does that mean? That really means getting that supply onto our platform, so building those integrations into the product so that customers have more choice, and then better communicating that choice and helping customers make a decision on what to purchase.

Our focus is there at the moment, but these features have been built on our European, Pan-

European platform, as Mike said, it is one platform, and so we believe in the future they will be highly relevant and differentiated. As we capture on the aggregation opportunity, and as we get more scale, we expect that we will want to have those features in market as well.

Nick Stockman

That's great. So, then just as you think into future products roadmap and things coming down the line, are there any teasers that you can give us around additional, I guess, in particular, AI- powered tools and products that your guys are working on?

Nina de Souza

Yes, there's two areas we're super excited that I would call out and they're for slightly different travellers. The first one is about making journey discovery more intuitive, and that's really for our less frequent travellers.

It is a smaller part of our business, but if you can imagine my mom coming from Michigan and trying to navigate the European rail network, and what station do I need? That experience is really clunky today and has a high cognitive load on the customer.

We're prototyping what that might look like to have a more chat conversational style interface, thinking about who we may partner with as ChatGPT and others come to Europe, and really building for a use case, which is a smaller use case for us, but one where we think we can build a better consumer product by enabling discovery, potentially through other platforms. That's the first area, really, for our infrequent smaller set of customers, make that more intuitive.

Then the second area is going further on disruption support. So really, a bulk of our customers, a bulk of the business is from highly frequent repeat customers. Really, those are the people who those disruption features matter. They know where they're going. They just want it to be simpler.

They want to understand their options. So, you could imagine us - Nick's taking a train journey and his train's cancelled, us actually giving you options and doing the rerouting for you.

So, features like that, where we're really able to take the disruption support more from information, and light resolution to deeper resolutions, and doing that for our more frequent customer basis, an area that we're excited about.

Nick Stockman

Perfect. Just a couple of questions I've had in on product. So, one was just, is there anything more that can be done from a technological standpoint to improve the SplitSave offering versus the likes of TrainPal, for example, that would potentially strengthen the competitive advantage that you have?

Nina de Souza

Yes, it's a good question. I think there is, and we're looking at that. So, for folks who aren't familiar, TrainPal is doing splits in a slightly different way, probably covering more edge cases than we are covering.

As a scaled market player, we really have to think about how that is going to work for the customers overall and for the industry. So, when you're thinking about doing multiple splits and customers potentially changing seats multiple times, that's okay if it's 2% of customers, but if it's 30% of customers, that's quite an unwieldy experience physically on the train.

So we're balancing customer experience, value to customer, convenience, and our scale.

Absolutely, value is important to our customers. We're thinking about how and where we play, and how we sit against TrainPal for customers that might be more price sensitive.

Mike Hyde

Yes. I’ll just say on that, just for clarity, there isn't really a technical limitation for us not adding more sophistication to SplitSave. It's more a product customer experience, user experience trade-off, if you like. So, we're constantly looking at it.

We could keep going down the rabbit hole, it's fine, but also we're a mainstream product. We've got 18 million customers [in the UK], and there's also a point at which too much complexity can also cause a bit of confusion with some of our customers as well. So, there's an optimisation point for us. So, we're constantly looking at that and evaluating where's exactly the right place for us to play.

Nick Stockman

One for you, Pete, but just while we asked about TrainPal there, or Trip.com, someone’s just asked about whether you're aware of them making any inroads from a market share perspective?

Pete Wood

We monitor a broad set of competitors, and TrainPal is certainly one of those we watch. We've seen downloads grow as one of the more direct things that we can measure over time. They appeal to a younger, more youthful audience, strong value messaging, and that's something we watch.

We've talked a bit about some of the SplitSave. At times, they may even trip over the line as well.

There's a bit in the press about whether they are compliant with all of the advertising requirements of them as well.

Yes, look, they are growing a bit, but they are also a couple of points of segment share. I think they are actually smaller than Uber at this point by our best reckoning and we watch both of those as slightly better known brands that play in the third-party retailer space.

Nick Stockman

Okay. Pete, again probably one for you, and - well, possibly a combination of you both, sorry, that I missed earlier when we were talking GBR and white labelling potentially. The question was, would a white label Trainline solution effectively be the best value for money that GBR can ultimately spend on that on this?

Then, as a follow-up question, if they do decide to go to another provider instead, have you got any views about who else would be a relevant competitor in that procurement process?

Pete Wood

Yes, we haven't got the procurement to really see what the government is after at this point, so it's difficult to know how well we would stand up against that.

I do believe we offer a great scaled solution; we've got all the benefits of the scale that Mike has talked to don't just exist for Trainline, but our white label providers also benefit from the security aspects of our estate, et cetera, et cetera.

So, yes, I think there would be a strong value proposition, and look, assuming the procurement comes, once we get to see that, we will look at it with the intent of putting in a bid, and we'll have to see how that goes.

In terms of other people that are in this space, I think through the market engagement, we expect there's quite a broad church here, that there's the big IT integrators that would build from the ground up, all the way through to some of the smaller participants in the market today who are perhaps providing technology in a white label sense that Trainline isn't to some of those train operators.

So, there's quite a broad set of options that the government has available, and again, depending on how any procurement gets shaped up, some of those will fall away and others will come to the fore. So, yes, it's a watching brief, really.

Nick Stockman

Great, and then there were just a couple of follow-up questions on specific product things that you guys have mentioned. So, have you got any visibility on a timeline as to when automatic delay repay could become available?

Pete Wood

It's still a live conversation. In short, no, we don't. As Nina referenced, we have taken a half step forward in terms of notifications and making that easier for customers, but that is definitely not the end game that we're after, and it's something that we're pressing hard on. It's the top ask of our customers, and we really want to get it out to them, and we need to continue to work to find a way forward there.

Nick Stockman

Great. Then Nina, I have a question, just asking to dig into the fraud protection side of things a little bit more, just in terms of exactly the problem that you're solving for in the market there and what you guys are offering?

Nina de Souza

Yes. So, I guess there's two elements, right? There's payment fraud, which we partner with other companies that payment fraud is their bread and butter, and we layer on top what we know about customers to do that better.

Then you have the rail fraud space, where we work together with the industry to help them investigate fraud solutions. So we are identifying cases where people are short-faring, or skipping the barriers, or just doing a number of tricks that people have to get around paying. I'm really working with them on that.

Also, around our digital pay-as-you-go trial. So, there's potential for fraud as you have to say where you are, and we have the geospatial technology to help understand the network, and understand where a customer is.

So, we're very well positioned to have the information to identify that fraud on behalf of the industry. So it really is a collaboration with them, and product work going on in the background, and maybe there's a world in the future where we want to offer more as a service and as a product, but right now, really a collaboration on solving the problem.

I don't know if you’d add anything, Pete.

Pete Wood

No, no, you’ve got it.

Nick Stockman

Great. I’ve had some questions in on pay-as-you-go, but I might just leave those for a second, just while we carry on with the AI discussions. I just want to make sure that we cover those areas in some detail. So, maybe if we move on from AI within the product perspective, and then on to the broader ecosystem. Maybe one for you, Mike.

Just how are you thinking about the changing backdrop and the potential disintermediation of search? What are the challenges like in terms of winning in that world versus the SEO world? Or how whole-scale is the landscape shifting?

Mike Hyde

Yes, it's a great point. I do think we're living in an interesting era. The large public LLMs are certainly starting to change how people are just starting their online journeys, if you like.

So, I do think that we're starting to see people navigate the digital world slightly differently because we're starting to see a little bit of traffic switch from starting with a Google search into starting with a conversation with a ChatGPT or something like that, and just in terms of mapping out what you want to get done online.

How we think about it, there's two parts to this topic. We think about these LLMs almost like the new search, they're helping people with queries. In effect, that creates a new source of lead generation for us, because people talk about trains and rail travel, and then they get recommended often to link onto Trainline.

Then secondly, there is this question, which is live in discussion at the moment, around how

LLMs are going to get into automating other tasks, like how far down the e-commerce funnel are the big LLM players going to go.

So, if I just take those two points. On that first point, we have seen a small but noticeable switch from traffic, which has been pretty religiously Google search for the last decade or two in pretty much every internet company I've worked in. It is starting to move, which is interesting.

In a sense, that's helpful to us. A bit of competition for Google in that space is not a bad thing, probably for everybody involved. That said, it is still quite small. It only represents less than 1% of all our international web traffic at the moment. So, we see it, but it's small in that sense. In terms of how it's going for us, it's going well.

We see it in a very similar way to SEO. The industry is largely referring to this as GEO now. The things that make you good at SEO tend to make you good at GEO. So having great, unique content that these services can index, having a well-known brand that people trust and people like, and having just generally positive sentimental conversations about you elsewhere on the internet. These are the things that are picked up by those services and mean that you're ranked well.

We're ranked very well at the moment. We have tooling that measures how often we get referenced in a conversation in ChatGPT, for example. If you're talking to ChatGPT about buying rail tickets, where does it recommend that you go to? We're the number 1 cited place to buy those rail tickets in pretty much all of the markets we operate in. I think we're number 2 in

France at the moment, yes, but number 1 everywhere else.

So, we're feeling good about that, something we watch closely, and we've built it into our wider emerging SEO strategy. So, we feel good about that.

In terms of the second point, will LLMs get into e-commerce? I think that's clearly quite an emerging space at the moment. There's lots of talk at the moment about standards of how LLMs and agentic systems might be able to pay for things in the future. Our position, really, is that we believe there is always going to be a space for strong, vertical players like us in that market.

Our position is largely that we want to show up wherever anyone wants to buy a rail ticket. So, we will look to integrate with these platforms and use them as a place to drive traffic. We don't see those LLM platforms replacing vertical players. As we've talked about, there's an awful lot to do when you sell a rail ticket, like with what we do on Trainline.

It's not a case of just search and discovery, and it's not even really a question of e-commerce.

You've then got to actually fulfil the ticket, and open the barrier, and do the customer service on all the things that we've been talking about.

So, we think there's a strong place for us in this future ecosystem, and our strategy at the moment is really to figure out how we can build relationships with all these different providers, and look at them primarily as new sources of traffic.

Nick Stockman

In a world where, as you touched on there, we start to see agents booking directly on behalf of the end customer, and therefore, all of the transaction of the journey happens through ChatGPT, or whichever AI platform a customer chooses to use.

How disruptive could that be, do you think, and maybe it’s one for you as well, Pete, from an economic model perspective? I guess people have questions around the importance of people being on the Trainline app or website as regards, advertising revenue, ancillary revenue, that sort of thing. Is there a risk to some of those sources of revenue if we don't see people come through your app or your website?

Pete Wood

Yes, I think we clearly have opportunity to do more with a customer when we own them. To add a bit more to how Mike started from a technical perspective, the vast majority of our business is repeat users making the same journey, and they are looking to get their ticket, on a journey they know, as quickly as possible.

To Mike's point on the vertical specialist, we have honed a flow that does exactly that, and we have a tech stack that can deliver against that, and the conversational style to try and get the ticket that you have is not necessarily the natural port of call to go to that.

Nina spoke earlier, maybe her mum in Michigan doing a more discovery route will do that, but that is not where the majority of our business lies. We are looking to stand up technology and capabilities that could play in that space as well.

Then, if I just take another step back, we have a licence to sell, we have to pass accreditation, we have bonding agreements with the industry because we hold on to a large sum of cash.

There is a much larger set of considerations, including post-sale, where we have customer support. At some point, you need a person at the end of the phone, and those are difficult to scale.

Of course, we're all learning with this emerging technology, and in the very long run, who knows, but for the foreseeable future, there is a lot that vertical specialists have to do and will continue to do. Rail is no different to that. So we expect the majority of customers to still come through our branded properties for the foreseeable.

Nick Stockman

That's super helpful. I've certainly had this discussion with one person on the call in relation to another business recently. Can you even get a hold of a sense of somebody at OpenAI who talks about how they envisage the future in this regard? Do you think they're even thinking, yes, we want to sell train tickets in the future? How do you think they're thinking about this world?

Mike Hyde

I can talk to that. Yes. I'm quite well-connected into several senior folks in OpenAI. I don't want to overly speak for them, but honestly, it feels like the way they are seeing the world is that they want to create an ecosystem of partnerships. Their objective is not to put everybody else out of business. Their vision is more about creating sort of a discovery layer at the top, where, a little bit like I described earlier, competing with what - in many ways, the job that has been done by

Google Search over the last decade or two, really.

Your jumping off point, where you can triage what you want to get done in your session online, but then more and more, they want to figure out how they integrate and pass you into the right place to get that done, like with the smoothest handoff, if you like, for the customer.

You can see the evidence of that, as they're building their latest integration SDKs at the moment.

Those SDKs are all about enabling other brands to show up in ChatGPT, but they're not taking over ownership of the customer. They're passing the customer into that brand and allowing them to then transact and get fulfilment from that brand. So, it does feel at the moment like that partnership is shaping up in exactly that way.

Pete Wood

Yes, and to offer one other thought, Nick, the margins in rail are not big. At times, I wish they were bigger, but it ultimately is somewhat a moat. So, for OpenAI to decide rail in Europe is the next move is again, to what end, and with what reward, if you like. So, again, another consideration here.

Nick Stockman

Yes, 100%. I think that's probably covered that area pretty effectively. So, maybe we'll just come on to talk about AI as regards to productivity and efficiency.

So, you've talked about some of the additional functionality that people in the tech side of the business benefit from, from an AI perspective already, Mike, but can you just dig into that a bit more in terms of areas where you've already really leveraged AI to get more efficiency out of the business?

Mike Hyde

Yes, certainly. So one of the things we've been doing over the last almost two years now is actively experimenting, deploying AI assistant tools across the business, and learning like everybody else about where they're most effective, and how we can automate what we do.

Where we've got to at the moment is where most of our developers use GitHub Copilot. We also have a portfolio of other AI engineering tools. So, every engineer gets access to one of these things, and we're seeing quite interesting results. It's not just the coding itself, which is speeding up, it's also a lot of the other jobs to be done around the wider concept of engineering, which are speeding up.

So, in particular, lots of benefits in terms of things like writing documentation, writing tests, and also just reading large volumes of code and making sense of them, which is what actually takes a lot of the engineering time. A lot of people think engineers spend all their time just writing code, but actually, there's this whole set of tasks that need to be done, and those tools are helping with a lot of those things.

The other really interesting thing we've done is we've rolled out ChatGPT Enterprise to all of our

Trainline employees, the tech, and the non-tech ones. That's had a very wide adoption internally. We're starting to see these use cases organically show up. It's starting to not just speed us up, but also, in a way, make us smarter, I would say.

A good example of that is we're seeing employees building their own areas of expertise on top of

- you can have custom ChatGPTs. So, an example of that is that some folks have built an experimentation GPT internally. What that means is if you want to go and ask a question or have a conversation with it, you can ask it about any of the experiments we've ever done as a business on any of our product surfaces, and find out the insights that came from that as you're figuring out what to do next.

It's a great example where it's not just about going faster the net result of that sort of thing is it makes everyone as smart as the smartest person, if you see what I mean. So, we're seeing that effect as well, of us just as a high knowledge, high skill business, you're up-levelling people across teams, which is quite an exciting prospect.

Nick Stockman

Yes, 100%. When you think about things that you haven't managed to harness yet, any obvious near-term future potential to unlock?

Mike Hyde

Yes, in addition to the core engineering use case that we continue to iterate on, we're seeing some good benefits in customer service, we're deploying AI tools over there, and that's helping us with faster customer response time, and also multilingual coverage, which is quite powerful without having to scale out a very broad set of native language teams market by market.

Especially as we expand into Europe, and have customers now from all around the world, that ability to have multilingual support without having to explode the size of our team is a real superpower. I don't know, any other examples?

Pete Wood

No, I think those are probably the main ones we've had so far.

Nick Stockman

That's perfect. Yes, and Pete, maybe just a segue from that, how are you thinking about the balance between allowing some of these efficiencies to flow through as operating leverage to the business versus reinvestment back into the group?

Pete Wood

Yes, the customer service was a clear win for us in terms of leverage. There wasn't much discussion to be had on that one. If I think about the tech team, our stance right now is to bank the efficiency, and then redeploy it as moving faster, doing more. We made a rationalisation in our cost base 12 months ago or so. We've got the right size team for what we're chasing after now, and we'll keep that under review, but there are no plans to move the cost down. Instead, we'll take the wins as they emerge.

Nick Stockman

Perfect, and that's maybe a little segue just into a question that I just had in relation to expenditure more broadly, actually, specifically on the overseas business. So, you've obviously seen that the French market opened up in fits and starts. How are you thinking about the balance between EBITDA progression in Europe, with the need to spend more money on marketing when France really comes back with a vengeance?

Pete Wood

Yes, I think about the countries in slightly different ways, and to point at what we call foreign travel, those tourists that come to Europe, that's a profitable segment that continues to grow, and the marketing equation is simpler because you're essentially playing an in-year, lifetime value, CAC kind of ratio.

Then, if I take Spain as a different example, we started four years ago, with very low awareness.

So, we had to inject marketing at pace to drive awareness and consideration, and to get that flywheel started.

If I go back to November, we said that that market was now at a point where we were going to think about how we rebalance some of the growth versus profitability, and that's a test and learn kind of process, but the direction of travel is to pull out some of the brand spend, to pull out some of the performance marketing spend that got that flywheel going, and then to learn what the right balance is, and to reset that equation, if you like.

Then, if I think about France, as another example, a couple of years ago, we pulled brand marketing spend out because the liberalisation, or lack thereof at that point, didn't support widespread national campaigns. Instead, we focused on Paris, Lyon, and subsequently, when

Marseille opened up, we put more brand marketing into that.

So, that's at a point of the cycle where we're investing more heavily, and as different cities open up, we'll put more into capture that moment, maybe to a point where there is a national campaign, once again, that makes sense, because you've got much of the map networked out.

Then, there'll be a period of time where we leave that there before feathering back again.

So, hopefully that gives a bit of colour about how we approach it. I have some investors who question whether we've got too much, and others that wonder if we've got enough in. That helps me calibrate that I think we're in the right place at the moment.

Nick Stockman

That's super helpful. Then, just maybe, to come back to the questions that came in around pay- as-you-go. So, you can just remind people around the scale of pay-as-you-go trials that are going on in the market.

I think you've talked publicly about what impact that could potentially have, but can you just maybe then dig into the expansion from a London pay-as-you-go perspective and, yes, what potential impact that has on net revenue, and then perhaps, what timeline we're on for award and pay-as-you-go contracts as well, given I know there's trials going on in that regard?

Pete Wood

Yes, certainly. So, let me take the digital pay-as-you-go trials that are unfolding in the Midlands.

This is nothing to do with London. It is a technology that doesn't require gate lines, and therefore, is attractive for the government in expanding a London-like experience without all the

CapEx of gate lines.

A quick reminder, there are four trials that are stood up. Trainline was first to market, and we're pleased with the progress we've made. We've had feedback, we've had the ability to fine-tune bits of this through the trial to date.

Looking forwards, it's still not clear exactly what the government will do next. The objective of the trial was to learn. We'd love to see it extend, but that's a TBC at this point, and they may choose a different route. I think with that in mind, they are excited and there's an energy to launch this, but there's not really a clearly laid out plan.

So, in P&L terms, I don't think this comes for a few years yet, and we aren't baking that in.

Nonetheless, an exciting opportunity, and great that we've got out the gates on it.

Then, in terms of the expansion of Oval, that started in earnest at the beginning of last year, and we've now got the best part of a year for those first 50 or so stations that went live. There's a decay curve, if you like, on day 1, not many people have taken it up, and then a few people do over time, and that is a headwind for us, as we've said in the past.

There's a £150 million [net ticket sales] headwind net from the whole proposed expansion in broad-brush terms. Think of that being £50 million this [financial] year, maybe £100 million next year, just to give you some altitude on it.

There was a second tranche of stations that went live, let's say another 50 or so in late

December, a bit early at this point with a disruption of Christmas to really get a read on that, but we'll be looking at how that unfolds, and then slated for summer is 100 plus more stations to go, and we'll have to see whether that stays on track and how that unfolds at that point.

Nick Stockman

Great. I think we're largely there, just one other question, maybe just to tie in a couple of the last rounds, just in relation to tech obviously being the dominant theme, and then also the European piece, because we haven't touched on it too much for Mike and Nina.

Just anything that you can say in terms of how the product proposition and tech proposition of the business you think stacks up against the main competition in Europe? Which, I guess, SNCF would be the obvious one to think about it in relation to.

Nina de Souza

Yes, so I think the incumbent operators like SNCF, they play a real important role in their home markets. SNCF in particular, has a strong experience. How does Trainline stack up against that, and aggregation is really the reason to use Trainline.

So, being the trusted platform that is bringing together all of your choice around timing, pricing, et cetera, promotions in one place is the core reason to choose, and to believe, and that is super important that we provide that consistent, agnostic experience for our customers, and has been something that's made us really successful as we've entered Spain and Spain's liberalised.

As we talked about, the differentiation features that we're building in the UK, those would be quite differentiated in the Europe market as well. So, you can think of how do we stack up versus

SNCF is, hey, aggregation is a completely different reason to choose us. Lots of customers will, and there are things that we can do over time that will further differentiate beyond aggregation as those markets scale.

In other countries, like in Spain, there's usability issues, and the customer feedback and sentiment on them is less good than versus an SNCF. So, it is kind of a mixed bag of other operator apps, but we believe, regardless of where you are on that spectrum, that the Trainline proposition currently does, and will continue to stack up as we build into the future.

Nick Stockman

Super. I think that's great. We've covered a lot of ground. I think we're just come up - literally, it has just ticked round to the hour. So, that's perfect. So, yes, Mike, Nina, Pete, thank you very much indeed, all of you, for your time, and indeed to Andrew and Marco for setting this up.

Thanks, everyone, for joining. Yes, we'll leave it there.

[END OF TRANSCRIPT]