In the hospitality industry, an allotment is a block of pre-negociated “rooms” which have been bought by a tour operator. In the context of campsites, allotments represent a significant share of the mobile homes sales. Once the tour operator has contracted the allotement its role is to resell the mobile home weeks to final customers.
For the campsite owner, dealing with tour operator allotment requests is a poisoned chalice. On one hand these pre-booked sales are more or less a guarantee of selling a good share of its inventory. On the other hand, the discount level is so high that selling the whole inventory through allotments could potentially ruin the business. Hence, a balance must be found between allotment contracts and estimated direct sales to final customers (at full price, or lightly discounted price).
The purpose of our presentation is to show that the stochastic optimisation problem with recourse at stake is highly combinatorial and that algorithmic approaches relying on continuous relaxations of the demand behave poorly. For multi-site allotment optimisation with service level requirements from tour operators, we developped a Lagrange decomposition technique based on local Markov Decision Process solvers that outperforms classic “fluid displacement” approaches. We will provide experimental results on instances with 200 campsites 20000 mobile homes and 15 tour operators inspired from a leading european actor of the campsite industry.
This subject will be presented at the occasion of the 27th European Conference on Operational Research (EURO 2015) in Glasgow.