skip to main content


Title: A Statistical Learning Approach to Personalization in Revenue Management
We consider a logit model-based framework for modeling joint pricing and assortment decisions that take into account customer features. This model provides a significant advantage when one has insufficient data for any one customer and wishes to generalize learning about one customer’s preferences to the population. Under this model, we study the statistical learning task of model fitting from a static store of precollected customer data. This setting, in contrast to the popular learning and earning paradigm, represents the situation many business teams encounter in which their data collection abilities have outstripped their data analysis capabilities. In this learning setting, we establish finite-sample convergence guarantees on the model parameters. The parameter convergence guarantees are then extended to out-of-sample performance guarantees in terms of revenue, in the form of a high-probability bound on the gap between the expected revenue of the best action taken under the estimated parameters and the revenue generated by a decision maker with full knowledge of the choice model. We further discuss practical implications of these bounds. We demonstrate the personalization approach using ticket purchase data from an airline carrier. This paper was accepted by J. George Shanthikumar, special issue on data-driven prescriptive analytics  more » « less
Award ID(s):
1845444
NSF-PAR ID:
10301867
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
Management Science
ISSN:
0025-1909
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. We consider dynamic assortment problems with reusable products, in which each arriving customer chooses a product within an offered assortment, uses the product for a random duration of time, and returns the product back to the firm to be used by other customers. The goal is to find a policy for deciding on the assortment to offer to each customer so that the total expected revenue over a finite selling horizon is maximized. The dynamic-programming formulation of this problem requires a high-dimensional state variable that keeps track of the on-hand product inventories, as well as the products that are currently in use. We present a tractable approach to compute a policy that is guaranteed to obtain at least 50% of the optimal total expected revenue. This policy is based on constructing linear approximations to the optimal value functions. When the usage duration is infinite or follows a negative binomial distribution, we also show how to efficiently perform rollout on a simple static policy. Performing rollout corresponds to using separable and nonlinear value function approximations. The resulting policy is also guaranteed to obtain at least 50% of the optimal total expected revenue. The special case of our model with infinite usage durations captures the case where the customers purchase the products outright without returning them at all. Under infinite usage durations, we give a variant of our rollout approach whose total expected revenue differs from the optimal by a factor that approaches 1 with rate cubic-root of Cmin, where Cmin is the smallest inventory of a product. We provide computational experiments based on simulated data for dynamic assortment management, as well as real parking transaction data for the city of Seattle. Our computational experiments demonstrate that the practical performance of our policies is substantially better than their performance guarantees and that performing rollout yields noticeable improvements. 
    more » « less
  2. Customer cancellations and no‐shows have an important impact on the performance of an airline's revenue management system. In this study, we study airline network revenue management problems with customer cancellations and no‐shows, considering two types of demand models: independent‐demand and choice‐based‐demand models. In independent‐demand models, the booking request rates for different products are independent of the assortment of products offered by the airline (exogenous), whereas in choice‐based‐demand models they depend on the airline's assortment (endogenous). The stochastic problems of optimizing the booking policies for the two demand models are not tractable because of the high‐dimensional state space. Therefore, for each model, we use the corresponding deterministic, continuous‐time, and continuous‐state model, known as the fluid model. A booking policy based on the fluid solution can be implemented for the original stochastic, discrete model. This fluid policy is shown to be asymptotically optimal when the arrival rates become high and the seat capacities become large. In the independent‐demand setting, an optimal fluid solution can be computed by solving a convex optimization problem for which efficient algorithms exist, whereas for some choice‐based‐demand models, the fluid control problems are known to be intractable. We also develop mixed fluid policies for the independent‐demand setting, taking advantage of the real‐time booking state, to improve the performance over the fluid policy. In a factorial experiment of 20 small‐sized single‐leg problems, the revenue loss under one of the mixed fluid policies is demonstrated to be less than 1% of the optimal value for all problems.

     
    more » « less
  3. We provide algorithms that learn simple auctions whose revenue is approximately optimal in multi-item multi-bidder settings, for a wide range of valuations including unit-demand, additive, constrained additive, XOS, and subadditive. We obtain our learning results in two settings. The first is the commonly studied setting where sample access to the bidders' distributions over valuations is given, for both regular distributions and arbitrary distributions with bounded support. Our algorithms require polynomially many samples in the number of items and bidders. The second is a more general max-min learning setting that we introduce, where we are given "approximate distributions," and we seek to compute an auction whose revenue is approximately optimal simultaneously for all "true distributions" that are close to the given ones. These results are more general in that they imply the sample-based results, and are also applicable in settings where we have no sample access to the underlying distributions but have estimated them indirectly via market research or by observation of previously run, potentially non-truthful auctions. Our results hold for valuation distributions satisfying the standard (and necessary) independence-across-items property. They also generalize and improve upon recent works, which have provided algorithms that learn approximately optimal auctions in more restricted settings with additive, subadditive and unit-demand valuations using sample access to distributions. We generalize these results to the complete unit-demand, additive, and XOS setting, to i.i.d. subadditive bidders, and to the max-min setting. Our results are enabled by new uniform convergence bounds for hypotheses classes under product measures. Our bounds result in exponential savings in sample complexity compared to bounds derived by bounding the VC dimension, and are of independent interest. 
    more » « less
  4. null (Ed.)
    We study the power of selling opaque products, that is, products where a feature (such as color) is hidden from the customer until after purchase. Opaque products, which are sold with a price discount, have emerged as a powerful vehicle to increase revenue for many online retailers and service providers that offer horizontally differentiated items. In the opaque selling models we consider, all of the items are sold at a single common price alongside opaque products that may correspond to various subsets of the items. We consider two types of customers, risk-neutral ones, who assume they will receive a truly random item of the opaque product, and pessimistic ones, who assume they will receive their least favorite item of the opaque product. We benchmark opaque selling against two common selling strategies: discriminatory pricing, where one explicitly charges different prices for each item, and single pricing, where a single price is charged for all the items. We give a sharp characterization of when opaque selling outperforms discriminatory pricing; namely, this result holds for situations where all customers are pessimistic or the item valuations are supported on two points. In the latter case, we also show that opaque selling with just one opaque product guarantees at least 71.9% of the revenue from discriminatory pricing. We then provide upper bounds on the potential revenue increase from opaque selling strategies over single pricing and describe cases where the increase can be significantly more than that of discriminatory pricing. Finally, we provide pricing algorithms and conduct an extensive numerical study to assess the power of opaque selling for a variety valuation distributions and model extensions. This paper was accepted by Gabriel Weintraub, revenue management and market analytics. 
    more » « less
  5. Recent advancements in physics-informed machine learning have contributed to solving partial differential equations through means of a neural network. Following this, several physics-informed neural network works have followed to solve inverse problems arising in structural health monitoring. Other works involving physics-informed neural networks solve the wave equation with partial data and modeling wavefield data generator for efficient sound data generation. While a lot of work has been done to show that partial differential equations can be solved and identified using a neural network, little work has been done the same with more basic machine learning (ML) models. The advantage with basic ML models is that the parameters learned in a simpler model are both more interpretable and extensible. For applications such as ultrasonic nondestructive evaluation, this interpretability is essential for trustworthiness of the methods and characterization of the material system under test. In this work, we show an interpretable, physics-informed representation learning framework that can analyze data across multiple dimensions (e.g., two dimensions of space and one dimension of time). The algorithm comes with convergence guarantees. In addition, our algorithm provides interpretability of the learned model as the parameters correspond to the individual solutions extracted from data. We demonstrate how this algorithm functions with wavefield videos. 
    more » « less