inicioir al DIIbuscar
customization in assortment planning with demand learning

Motivated by the availability of real-time data on consumers we study a series assortment planning problems under model uncertainty in the context of online advertisement and online retailing. First, we study the problem of assortment customization based on consumer characteristics when there is limited initial information about their preferences. Our results identify the critical role assortment experimentation plays in balancing the exploration vs. exploitation trade-off arising from initial model uncertainty. Then, extend our results to a broad class of problems under initial model uncertainty. By establishing fundamental limits of performance our main result answers in a precise manner three fundamental questions: what, when and how to explore.  Finally, we address the 'big-data' issue that arises in practical settings, where not all information contained on a consumer profile is relevant. We show that policies that redefine clusters dynamically consistently over perform relevant benchmark.