
Airlines are often treated as the gold standard of revenue management, and for good reasons. Not only were they the first to do it, but decades of investment have given them a sophisticated toolkit, from demand forecasting and dynamic pricing to network optimization. But rail and bus operators have been building practices of their own, and these deserve more attention from airline revenue management teams than they usually get.
Some of this is quiet innovation, happening below the radar of an industry used to seeing itself as the benchmark. Some of it comes from constraints: high-frequency corridors, thousands of origin-destination pairs, and sparse demand have pushed rail and bus operators to solve problems airlines are only now starting to face at scale.
This kind of cross-industry learning is not theoretical. CAYZN, Wiremind’s end-to-end AI-powered revenue management solution for passenger transport, already runs on both sides of it. Rail operators like SNCF and Trenitalia, bus operators like National Express and Itabus, and airlines like Transavia and Flair Airlines, all optimize on the same core engine.
Working across all three has shown us that airlines can learn a great deal from the methods rail and bus RM teams have developed.
Rail, bus, and airline revenue management are not identical: fare structures, regulatory context, and distribution channels differ in real ways. But all three face the same core challenges:
Rail and bus networks tend to meet these challenges, sometimes at a larger scale, and the solutions they have developed can be applied directly to airlines. Here is what those solutions look like.
A single train can have dozens of origin-destination (OD) pairs, and each one has its own micro-market. Multiply that across a national network, and the number of ODs explodes, while the volume per OD pair stays low.
Classic forecasting methods struggle here. These methods usually project the rest of the booking curve either by adding a fixed pickup estimate (additive) or by scaling current bookings with a historical ratio (multiplicative). Both assume each OD follows a stable pattern. With only a handful of bookings per OD, that pattern isn’t there, and forecasts swing widely from one departure to the next.
The models that hold up are built for sparsity from the start. Instead of treating each OD in isolation, they draw on signal from wherever it exists:
Machine learning methods such as gradient-boosted trees, or temporal models like TSMixer, make it possible to combine all of these at scale. Our own research on deep time-series models goes a step further, training them to forecast how demand responds when prices change.
Airlines face the same problem on thin routes, secondary markets, and off-peak departures, where there are too few bookings for classic methods to produce stable forecasts. The challenge is sharpest on new routes which have no booking history at all: the classic cold start problem.
An airline that borrows rail’s sparsity-ready models gets a forecast it can work with from the first day of sales, instead of waiting months for booking history to build up. The same approach also helps with network optimization, where estimating demand on a flight means modeling many small ODs.
On a train, each leg is shared by many overlapping ODs. Selling a Paris-Lyon ticket can therefore cost the operator a Paris-Marseille passenger, who would have needed that same seat between Paris and Lyon.
Allocating capacity well means thinking at network level rather than segment by segment, using displacement costs and bid prices to measure what each sale takes away from the rest of the network. It also requires an optimization engine that can handle a very large number of nested ODs on each service, combining dynamic programming solvers and heuristics. For optimization under shared capacity, rail is rather an extreme test bed.
Airlines know this problem well. It is exactly the O&D challenge hub-and-spoke carriers face: balancing local and connecting passengers who are competing for seats on the same leg. The logic is the same as on a train: every sale has to be weighed against what it takes from the network. An engine built to handle that many overlapping ODs on every train service is well equipped to arbitrate between local and connecting traffic, which is why it applies directly to airline networks.
The same goes for multi-stop operations. An aircraft flying from A to B to C in a single service creates the same nested ODs as a train calling at several intermediate stations, something rail teams handle as a matter of routine. Airlines running multi-leg routes can draw directly on that experience.
The results are already visible: R&D aimed at cutting optimization time for complex rail clients has significantly sped up optimization for CAYZN’s airline clients too, making the platform smoother to use day to day.
On dense rail and bus corridors, departures can run every 30 minutes, and closely spaced services are effectively competing for the same passengers. Closing a bucket on one train doesn’t necessarily mean losing the sale; it can also push demand toward the next train. The purchase decision tends to be a choice between departures, not a choice made on one departure alone.
Optimization has to be adapted. It starts with modeling how passengers choose between neighboring services and how demand shifts from one to another, while having to estimate price sensitivity and willingness to pay, including buy-up and buy-down behavior. The optimization then builds on these estimates to take the interaction between departures into account, so each one is optimized with its neighbors in mind rather than in isolation.
Rail and bus operators go one step further with battery steering: managing several consecutive departures at once. As Charles Pierre, Wiremind’s CTO and co-founder, explained on the Airline Tech Podcast hosted by Garth Lund, operators on these corridors align pricing across consecutive services, nudging passengers toward shoulder departures when peak ones are filling up.
High-frequency airline markets, such as shuttles and business routes, show the same dynamic. When a fare is closed on one flight, the passenger may take the carrier’s next flight or switch to another airline. The share that stays within the carrier shifts demand between its own flights, and optimizing each flight in isolation ignores that shift.
Airlines already do some of this, but according to Charles, the techniques developed in rail are genuinely more advanced.
Cannibalization modeling helps an airline balance demand across its own flights. It estimates how many passengers will take another of the airline’s own flights when a fare closes, instead of booking one with a competitor. With that estimate, the airline can raise prices on a busy flight without losing those passengers, and keep cheaper fares on the quieter flights around it to win them over.
Rail booking curves are famously steep. A large share of passengers book close to departure, and on high-frequency corridors, bookings can accelerate significantly in the final few days. Walk-up and commuter travel is common, and booking patterns also vary widely from one OD to another, and between weekdays, weekends, and holidays.
Rail revenue management teams have become experts at working that final window, routinely taking several pricing and availability actions per departure in the last days before travel. That takes three things:
All three rest on one prerequisite: live booking data. Prices and availability can only follow demand if every booking reaches the system as it happens. Rail and bus operators depend on this, so their systems are built around it.
Airlines see a similar but milder version of this, with late business bookings and last-minute leisure surges. Their approach to that window, however, is often different. As Charles also pointed out on the podcast, airlines tend to default to yield protection in the close-in period, rather than actively stimulating volume the way rail teams do. For him, that leaves a real opportunity for airlines to be more dynamic.
Forecasting and optimization calibrated for rail’s sharper, less predictable booking curves can support that shift: if they hold up under rail’s constraints, they will hold up under an airline’s as well.
The live booking data foundation carries over too. Airlines that feed their revenue management with live bookings, as rail and bus operators do, can use all three capabilities: finer forecasts close to departure, a better estimate of true demand once fares close, and re-optimization that follows demand as it builds.
In a rail network, complexity doesn't add up; it multiplies. Every origin-destination pair, every departure, and every price point is another factor, not just another line item. The result is a massive volume of calculations that has to be processed continuously and close to departure.
Handling that takes two things:
Analysts can then focus on the decisions that truly require human judgment instead of monitoring every market.
Airline networks face the same multiplication of origin-destination pairs, departures, and price points, with an added factor of their own: flights usually go on sale up to a year before departure. This means that a single analyst can be responsible for thousands of flights open for sale at once, and no team can monitor that by hand. Rail needs automation because of the sheer number of ODs, while airlines need it because of how long and how wide their sales window is. An engine built to carry rail’s load handles airlines’ without strain. Airline teams also get the same automation from the start: re-optimization that runs on its own, and alerts only on the markets that need a decision. Their analysts spend their time on the calls that need human judgment, not on watching every market.
The interface follows the same logic. Making rail’s complexity manageable required a streamlined, uncluttered design, and airline teams get that clarity too.
A rail operator rarely competes only with other rail operators. On most routes, passengers compare options: trains, buses, planes, and sometimes also carpooling. They then choose based on price, total travel time, and convenience. Someone traveling from Paris to Lyon isn’t asking which train to take; they’re asking how to get to Lyon.
The opening of European rail markets to competition has added another layer. On some corridors, several rail operators now run on the same line, alongside buses and airlines. Which of these alternatives matters most also changes from one route to the next. On short trips, a bus or a shared car may be the main rival. On longer ones, it is often the plane.
Rail and bus revenue management teams have learned to look at the whole market, not just their own mode. In practice, that means:
Airlines, by contrast, tend to benchmark themselves only against other airlines. Yet travelers increasingly look at alternatives to flying, and price weighs heavily in that choice.
The Madrid-Barcelona line is a perfect example of what happens when the rail alternative becomes competitive. Liberalization began in May 2021, when Ouigo started running on a corridor that Renfe’s AVE had served alone for more than ten years. Iryo followed too. With these new operators, ticket prices fell by up to 40%. Rail’s share of the air-rail market grew from 63.3% in 2019 to 83.1% in 2025*. In March 2025, Vueling withdrew from the route, pointing to how difficult it had become to compete with high-speed rail for point-to-point passengers.
The effect also works the other way: when rail disruption hit the corridor in early 2026, demand shifted back to air, with Iberia capping economy fares at €99 each way, and Vueling temporarily returning to the route.
For airlines, widening the competitive set matters most on short-haul routes, where a train or a bus is an option worth considering. Monitoring these alternatives gives a more accurate picture of what passengers will actually pay. It helps avoid pushing price-sensitive travelers onto the train, and it shows when there is more room on price, for example, when the rail alternative is sold out, expensive, or disrupted.
*Sources:
https://www.railvolution.net/news/madrid-to-barcelona-rail-wins-over-plane
https://www.railwaypro.com/wp/spains-high-speed-rail-sees-passenger-numbers-double-as-fares-fall-by-40/
Even the most accurate forecast and the best optimization only matter if the system behind them can put them into action. The revenue management engine sets prices and availability, but the reservation system decides how finely those decisions can be expressed and how quickly they reach the market.
Many rail operators now work with modern technology stacks that allow them to:
Airlines working within legacy reservation architectures often lack that flexibility: fares sit on a fixed ladder, OD groupings are set in advance, and booking signals don’t always reach the revenue management system in real time. As Charles noted on the podcast, this puts a ceiling on what even the most sophisticated models can deliver. No matter how accurate the forecast is, the reservation system decides how much of that accuracy actually reaches the market.
The rail and bus experience shows what a more flexible setup makes possible. For airlines, the lesson is that better forecasting and optimization only pay off fully when the reservation system can keep up with them.
Rail, bus, and airline revenue management are not competing disciplines. They work on the same core problems at different intensities, and rail and bus networks tend to push those problems further: more origin-destination pairs, thinner demand on each, closer departures, and steeper booking curves.
Each lesson in this post follows the same pattern. A constraint specific to rail or bus forced a more robust solution, whether that meant forecasting demand where bookings are scarce, optimizing across overlapping ODs, accounting for departures that compete with each other, reacting to late bookings, predicting no-shows to set overbooking, or running all of it at scale with minimal manual monitoring. Airlines face the same problems, often at a smaller scale, so those solutions apply directly.
None of this takes anything away from what airlines have built. Their toolkit is still a reference for the industry. The point is that a toolkit gets stronger when it draws on methods developed under tougher conditions elsewhere.
That is the idea CAYZN was built on: a single engine, tested against rail’s complexity and airlines’ demands, rather than two separate products for what is ultimately the same problem. Rail, bus, and airline teams already optimize on it today.
Discover how you can take advantage of our end-to-end AI-powered Revenue Management System, CAYZN, today. Reach out to us for a demo here.