Choice-Based Network Revenue Management under Weak Market Segmentation

We present improved network revenue management methods that assume customers to choose according to the multinomial logit choice model with the specific feature that the sets of considered products of the different customer segments are allowed to overlap. This approach can be used to model markets with weak segmentation: For example, high-yield customer segments can be modelled to also consider low yield products intended for low-yield customers, introducing implicit buy-down behavior into the model. The arising linear programs are solved using column generation and involves NP-hard mixed integer sub problems. However, we propose efficient polynomial-time heuristics that considerably speed-up the solution process. We numerically investigate the effect of varying the intensity of overlap on the respective policies and find that improvements are most pronounced in the case of high overlap, rendering the method highly interesting for weakly segmented market applications.