Airline Overbooking

Recently the topic of overbooking becomes a highlight as there was an incident in one of airlines which dragged its own customer who was selected randomly to go out of the plane when the overbooking prediction was not correct.

There are several angles from which we can look at the issue.

First, how do we define the goals. If the objective is profit, all companies need to make profit. So what is the reason other companies have lower probability of the same issue happening. Perhaps we need to ask how do we define profit. Is it short term or long term? Or how many period of profits are we talking about. Also, how do we define customer satisfaction and how to incorporate it in the profit calculation as bad customer experience can translate into additional costs and may reduce future income.

Second, if the goals are clear, how do we reach the goals? On what basis do we make decision to maximize the probability of reaching the goals? Is it based on gut feelings or models or both? If model, is it based on a single model or multiple models? What are our considerations to select the model? Given the same model, how many literature reviews or references do we have to support the idea that the selected model works.

Third, in selecting the models to use, we need to refer to some sets of data. How long is the validity of the data? Or when do we need to consider collecting new sets of data which may result in different conclusions?

Fourth, how do we handle variations? We know that there may be different kinds of trends on certain category, such as day of the week, hour, season, geographical location, etc. How do we segmentize the market or the trend? What are the simplifications we make? Also, do we assume that the involved parties are not able to react? Or what kind of agents flexiblity inside the model?

Fifth, how do we determine bad and good decision process? We need to remember that good decision may not result in good outcome. In the same way, good outcome does not necessarily mean good decision was made. What we need to preserve is the good decision as good decision has the probability of having better outcome in the long run.