Individual Travel Request Generation


One of the core features of the ASTD: Instead of distributing the whole demand of an airport at once and allocating it two 3 fixed service classes, the Request Generator generates individual travel requests based on a whole range of parameters like whether the request is for business or leisure, how many people are traveling, how much luggage the travellers intend to bring, how important different service criteria are to them, how much budget they have, how high their chance for a cancellation or no-show is etc. All of these parameters are constrained by general passenger types, for example “cost-sensitive business travellers” that tend to book business class but that have a higher chance to fly economy if they don’t see the value in picking business.

In the ASTD, request generation would also include origin and destination, travel date and their return date. In current-gen AS, this is a dedicated feature.


A core issue of current-gen AS is that there are really only 3 types of passengers: Eco, Business and First. This means all airlines compete for the same passengers, meaning all products need to be more or less the same. It only depends on the parameters we give those passenger types, whether a service-oriented or a price-based model works best in AS…and that’s what everybody follows. The new model would bring a lot more diversity and provide the foundation for actually enabling different business models in AS (like low-cost vs. network carriers).


In its simplest form, this could be grafted onto the current ORS: Instead of just allocating among the three static service classes after all potential connections have been generated, AS could generate a set of individual requests and match those with the found connections instead.

A core issue here are performance (the ORS generates tens of thousands of viable connections per destination, and matching those to an arbitrary amount of booking requests is an exponential problem). Another is transparency…without more in-depth statistics, it might become tricky to understand why a flight performs the way it does.

That said, there isn’t a general technical limitation that prevents implementation at this point.