2016 INFORMS Annual Meeting Program

224 / 561

INFORMS Nashville – 2016
224
MD46
209B-MCC
Customer Choice Estimation and
Assortment Optimization
Sponsored: Revenue Management & Pricing
Sponsored Session
Chair: Vineet Goyal, Columbia University, New York, NY, United States,
1 - Rationalizing Empirical Choice Counts
Srikanth Jagabathula, NYU Stern School of Business,
Paat Rusmevichientong
We consider the rank-aggregation problem where the objective is to find the
ranking over n items that minimizes the number of conflicts with the given
choice observations. A ranking has a conflict with a choice observation (i, S) if i is
not the most preferred item in the subset S according to the ranking. This
problem appears in many practical applications. This a known NP-hard problem
with almost no algorithms that provide theoretical guarantees. We propose a
graphical model approach and show that the complexity scales in the tree-width
of the graph, defined on the choice sets. Numerically, our algorithm out-performs
existing heuristics for important sub-classes of problems of interest in operations.
2 - The Impact Of Consumer Search Cost On Assortment Planning
And Pricing
Ozge Sahin, Johns Hopkins University,
Ruxian Wang
Consumers search for product information to resolve valuation uncertainties
before purchase. Under the consider-then-choose policy: a consumer forms her
consideration set by balancing utility uncertainty and search cost, then she
evaluates all products in her consideration set and chooses the one with the
highest net utility. The choice behavior within consideration sets is governed by
the multinomial logit model. The assortment problems are NP-hard. For the joint
assortment planning and pricing problem, we show that the intrinsic-utility-
ordered assortment and the quasi-same-price policy, which charges a same price
for all products except at most one, are optimal for the joint problem.
3 - Waste Reduction Of Perishable Products Using Dynamic Pricing
Arnoud V. den Boer, University of amsterdam, KdVI, science park
904, room F3.33, amsterdam, Netherlands, ,
Jieying Jiao
According to the 2013 UN Food Wastage Footprint study, “approximately one-
third of all food produced for human consumption in the world is lost or wasted”.
The impact of this waste can hardly be overstated. Some Dutch supermarkets try
to mitigate food loss by discounting products sold on their expiry date. This
measure reduces the expected waste, but its effect on profit is an open question.
In this talk we show how to optimize the discount percentage in a Markovian
framework, and we discuss the effect on profit and waste in a diffusion limit of
the inventory process.
MD47
209C-MCC
Pricing, Promotion Planning, and
Revenue Management
Sponsored: Revenue Management & Pricing
Sponsored Session
Chair: N. Bora Keskin, Duke University, Durham, NC, United States,
1 - Multiple Equilibria In Pricing Problems With Network Effects
William L Cooper, University of Minnesota, 111 Church Street
S.E., University of Minnesota, Minneapolis, MN, 55455, United
States, , Chenhao Du, Zizhuo Wang
We consider multi-product price-optimization problems with network effects
wherein the expected utility each individual customer obtains from a product is
increasing in the number of other customers who buy that product. Such
network effects give rise to equilibrium constraints that describe how sales
depend upon prices. In some cases there may be multiple different vectors of sales
quantities that satisfy the equilibrium constraint for a given vector of prices.
Moreover, the seller’s revenue may be quite different at those different equilibria.
In this talk, we compare the seller’s revenue-maximizing prices under differing
assumptions about which of the multiple equilibria will prevail.
2 - Value Of Targeted Promotions: Evidence From A Large
Department Store
Bharadwaj Kadiyala, PhD Candidate, The University of Texas at
Dallas, Richardson, TX, 75080, United States,
Ozalp Ozer, A Serdar Simsek
Gift cards have become a popular vehicle for promotional campaigns run by many
departmental, consumer electronic, and online retail stores. Using a proprietary
data set from a large department store, we investigate the value of targeted
marketing efforts via emails in the context of gift-card promotional campaigns.
We estimate the effects of online gift card promotions on customer purchase
behavior and then discuss how to use these estimates to plan for targeted
promotional events, a step towards one-to-one marketing.
3 - Pricing From Observational Data
Nathan Kallus, Assistant Professor, Cornell University and Cornell
Tech, 111 8th Avenue #302, New York, NY, 10011, United States,
Dimitris Bertsimas
Given price-demand data, pricing is often addressed by a predictive approach: fit a
model to predict demand given price observation, substitute into profit, and
optimize price. Predictive approaches fail to find the optimal price, which is not
generally identifiable from observational data. We bound suboptimality. We
provide identifiability conditions and corresponding methods for pricing and
prove consistency, asymptotic normality, and convergence rates. We develop a
hypothesis test for optimality of pricing from observational data and demonstrate
predictive approaches lose significant profit while our parametric method is
indistinguishable from optimal and recovers 36-70% of losses.
4 - Using Contingent Markdown With Reservation To Deter Strategic
Consumer Behavior
Gustavo Vulcano, NYU/UTDT, New York, NY, United States,
Navaporn Surasvadi, Christopher S Tang
We examine a contingent price markdown (CM) mechanism with guaranteed
reservations under which a retailer sells multiple units to forward-looking
consumers who arrive over time according to a Poisson process. We study the
consumer purchasing behavior in equilibrium, and numercially compare the
performance of our mechanism against two benchmarks: Fixed Price (FP) and
Pre-announced Discount (PD). Using extensive numerics, we identify market
conditions under which CM dominates both FP and PD in terms of the retailer’s
revenue and consumer’s surplus. Finally, through a fluid approximation to the
stochastic model, we analytically show that CM weakly dominates the other two
mechanisms.
MD48
210-MCC
Peeling Back the Onion in Social Media Analysis
Invited: Social Media Analytics
Invited Session
Chair: Chris Smith, Air Force Institute of Technology,
2950 Hobson Way, Wright-Patterson AFB, OH, 45433, United States,
1 - Assessing Bias Correction For Social Media Samples
Christopher Wienberg, USC Institute for Creative Technologies,
While social media has made it possible to quickly and cheaply gather the
opinions and experiences of large numbers of people, it is unclear how well social
media users represent broader real-world populations, introducing the possibility
of serious estimation bias. We investigate the applicability of traditional sample
reweighting techniques for estimating real world population characteristics from a
sample of social media users, with an eye towards using automatic attribute
inference from social media profiles to make predictions about real-world
populations.
2 - Who’s In Charge: Looking At Hierarchy And Heterarchy On
Social Media
Robert Schroeder, Naval Postgraduate School, ,
Sean Everton
Depending on the context of the information being passed on social media,
conversations can seem to be fairly flat with no-one in charge, or highly
centralized with a few main accounts key to the flow of information. With
networks of different size, this paper compares various network measures of
hierarchy in order to better classify social media conversations.
MD46