The operators who outperform do not predict better. They detect faster and react sooner.
Every revenue management conversation eventually arrives at the same place. Someone says: we need better forecasts. Better data. Better models. Better accuracy. And the assumption underneath is always the same: if we could just predict demand correctly, everything else would follow.
It sounds reasonable. It is also the wrong place to start.
The accuracy illusion
The car rental industry operates in one of the most volatile demand environments of any service business. Fleet deliveries shift by weeks or months with little warning. Airline schedules change and reroute inbound demand overnight. A local event gets cancelled. A heatwave drives coastal bookings through the roof. A geopolitical crisis reshapes travel flows across entire regions.
In this environment, how accurate can a forecast realistically be?
Benoit Rottembourg, a mathematician who spent three years leading pricing transformation at Maersk, the world's largest container shipping company, faced this question head on. His team's demand forecasts, three weeks ahead of vessel departure, had less than 70% accuracy. In maritime, where a single vessel carries 20,000 containers and demand is shaped by commodity seasons, weather patterns, port disruptions, and geopolitical events, the volatility made traditional forecasting almost impossible.
His response was not to build better models. It was to push back on his own data science team. His position was clear: you will never have a good forecast in a volatile world. Stop designing for accuracy. Design for robustness.
What robustness means in practice
Designing for accuracy means investing in better prediction models, longer data history, more variables, more computation. The goal is to get the number right before the season starts.
Designing for robustness means accepting that the number will be wrong and building a system that detects the gap between forecast and reality fast enough to act on it. The goal is not to predict perfectly. The goal is to recover quickly.
The distinction matters because it changes what the operator prioritizes. An accuracy-first approach leads to heavier upfront planning and higher confidence in the plan. A robustness-first approach leads to lighter planning, continuous monitoring, and faster decision cycles.
At Maersk, despite the poor forecast accuracy, the pricing system delivered results because it could react in minutes rather than weeks. When the old process required two to three weeks to adjust prices through Excel spreadsheets, the new platform allowed pricing managers to see booking flow in near real time and correct course within hours. The forecast was bad. The response time made it irrelevant.
The car rental parallel
Car rental operators face the same structural challenge. WeYield's own research confirms that relying too heavily on last year's data introduces inaccuracies and cannot predict new situations well. Post-Covid travel patterns broke most historical baselines. Past trends do not always predict future demands, and reliance solely on historical data can sometimes mislead, especially after periods when travel patterns have undergone significant reformation.
Yet the reflex remains the same. When demand does not match expectations, the first question is always: why was the forecast wrong? It is almost never: how fast did we detect the gap and what did we do about it?
Consider what a typical operator faces in the 30 days before a peak week. Fleet has been purchased or leased based on a plan made months earlier. Pricing was set based on assumptions about demand, competitor behavior, and channel mix. Then reality begins to diverge. Bookings come in slower than expected on one station and faster on another. A competitor drops rates on a specific car group. A broker cancels and rebooks a block of reservations at a lower price.
The operator who designed for accuracy is now stuck defending a plan. The operator who designed for robustness is already adjusting.
Three things that matter more than forecast accuracy
- Speed of detection. The most important metric is not how good the forecast was. It is how quickly the operator sees the difference between where they are and where they should be. A Performance Hub that shows booking pace by departure date, compared to the same period last year, gives the operator this visibility in real time. Not at the monthly review. Not at the end of the season. Every day.
- Granularity of the signal. An aggregate forecast for "summer" is useless for daily decisions. What matters is the utilization outlook for a specific departure date, at a specific station, for a specific car group. When the signal is granular enough, the operator does not need a perfect forecast. They need to know: is this particular date running ahead or behind? That is a question a booking pace curve can answer without any forecasting model at all.
- Asymmetry of the response. Not every gap requires the same action. Being slightly behind on a shoulder date is recoverable with a small rate adjustment or a broker activation. Being significantly oversold on a peak date is a different problem entirely, one that cannot be solved by cutting prices but requires protecting margin and keeping capacity available for high-value late bookings. Pricing Insights is designed to detect these velocity shifts and signal when the pace of bookings is accelerating faster than the utilization curve justifies, before the operator gives away margin unnecessarily.
The forecast trap
There is a deeper risk in the obsession with forecast accuracy. It creates a false sense of control. When the forecast says demand will be strong, operators price high and stop monitoring. When the forecast says demand will be weak, operators price low preemptively and leave money on the table if reality turns out better than expected.
Both errors come from the same source: treating the forecast as a decision rather than as a starting hypothesis to be tested against incoming data every day.
Rottembourg saw this at Maersk. Some data scientists on his team insisted that a better forecast model would solve everything. His answer was always the same: what can we do with a poor forecast? Because that is the only forecast you will ever have in a world where wars reroute shipping lanes, rain delays beef production in Australia, and a port decides to skip a stop because the vessel is running late.
In car rental, the disruptions are different but the volatility is the same. A new airline route opens and changes demand at a coastal destination. A competitor exits a market. A fleet delivery arrives three weeks late. A heat wave in Northern Europe sends bookings to Scandinavia instead of the Mediterranean.
The operators who win are not the ones with the best crystal ball. They are the ones who assume the crystal ball is cracked and build their process around fast detection and disciplined reaction.
From planning to piloting
The shift is not about abandoning forecasts. Every operator needs a plan. The shift is about changing the relationship with that plan.
A forecast is an input, not an instruction. It tells you where to start looking, not where to stop thinking. The real work begins after the plan meets reality, which it does every single day.
The question is not: was the forecast right? The question is: how quickly did I see that it was wrong, and what did I do in the next 24 hours?
That is the difference between planning and piloting. Planning happens once. Piloting happens continuously. And in a volatile market, the pilot always outperforms the planner.
We help car rental operators move from seasonal planning to daily piloting, with booking pace curves, velocity alerts, and station-level utilization tracking that make every gap visible before it becomes a problem. Book a meeting and let's look at your numbers together.
