Fares with fewer restrictions need a forecast that expects travellers to buy the cheapest open fare

Classic revenue management forecasts each booking class as if it had its own travellers. That held while advance purchase, minimum stays and no refunds kept business travellers out of the cheap fares. Fare brands and simpler fares remove many of those rules, and then most travellers buy the cheapest fare that is open. A forecast that does not expect this protects too few seats for the travellers who would pay more.

A forecast by class reads buy-down as lost demand

Suppose class M is open on a flight that a business traveller would have booked in H. The traveller buys M. The system records one booking in M and none in H. Next season it forecasts less demand for H, protects fewer seats for it and opens M earlier, so more travellers who would have paid H buy M.

Cooper, Homem-de-Mello and Kleywegt describe this feedback as the spiral-down effect. They show the conditions under which it happens: when booking controls rest on the wrong model of how travellers choose, high-fare sales and revenue fall step by step.

Three published ways to forecast for travellers who choose

Boyd and Kallesen describe how the models change when passengers buy the lowest available fare. They call demand for a particular product yieldable, and demand that buys the cheapest open fare priceable, and forecast each in its own way.

Fiig, Isler, Hopperstad and Belobaba convert the fares and demand of a model in which travellers choose into an equivalent model with independent demand, so the optimiser of a traditional revenue management system can keep working. In practice the optimiser then values a cheaper class at its fare less what opening it costs in buy-down.

Fiig, Weatherford and Wittman propose an error measure that compares the forecast with the bookings made, given the classes open at the time. In their simulations, reducing that error raises revenue whatever the fare structure or the way travellers choose.

Fare brands in one class add a second choice

When Basic, Standard and Flex share a booking class, the class opens and closes all three together. The traveller then makes a second choice at the same seat: whether a bag or free changes is worth the step up. The forecast for that class needs the share of travellers who choose each brand, or the class is valued at the Basic price alone. This is our judgement from how brands are filed; the simulator on our Method page lets each traveller choose a brand in this way.

What to check in your own system

• Which forecasting method runs in each market. Vendors use names such as hybrid forecasting and fare adjustment, and a system that has them may still run the classic method on some routes.
• Whether the forecast corrects for closed classes, which hide the travellers they turned away.
• Whether the share of bookings in the lowest open class is rising while the forecast for the higher classes falls from one season to the next.
• How forecast error is measured, and whether it accounts for the classes that were open.
• What each change did to revenue, measured against flights that did not change. Seven questions show how to read the result.