INFORMS Nashville – 2016
130
4 - Research Opportunities In Project Scheduling
Rainer Kolisch, Technical University of Munich,
rainer.kolisch@tum.deOperations Research has been applied in project scheduling for more than half a
century. This talk summarizes achievements and outlines research opportunities.
MA26
110B-MCC
Dynamic Matching
Invited: Auctions
Invited Session
Chair: John Dickerson, Carnegie Mellon University, 9219 Gates-
Hillman Center, Pittsburgh, PA, 15213, United States,
dickerson@cs.cmu.edu1 - Dynamics Matching With Departures
Maximilien Burq, Massachusetts Institute of Technology,
Cambridge, MA, United States,
mburq@mit.eduVahideh Manshadi, Itai Ashlagi, Patrick Jaillet
We study dynamic matching in an infinite-horizon market with stochastic arrivals
and departures, in which some agents are a priori more difficult to match than
others. We analyze the effect of batching for policies that match agents through
cycles of length 2 or 3. We show that if only cycles of length 2 are allowed, the
benefit of batching is not significant. However for 3-cycles, batching can result in
a considerable gain over greedy. Furthermore, using data from the National
Kidney Registry, we provide simulations that confirm our theoretical results.
2 - Dynamic Matching In Over-the-counter Markets
Yu An, Stanford, Stanford, CA, United States,
yua@stanford.eduZeyu Zheng
We model the dynamics of liquidity premium in an OTC market with
heterogeneous assets. A monopolistic dealer matches supply and demand flows in
order to maximize his profits. Inventory building by the dealer increases the
average waiting time for those customers who rejected immediacy offers, and
therefore helps the dealer extract rents via liquidity premium. The dealer’s dual
role of liquidity provision and matchmaking creates inefficient monopoly, and in
equilibrium, he holds too much inventory compared to the first best. Our result
helps explain the recent growth in all-to-all trading platforms in the corporate
bonds market, as they circumvent these inefficiencies.
3 - Toward A Credit-based Mechanism For Dynamic
Kidney Exchange
John Dickerson, Carnegie Mellon University,
dickerson@cs.cmu.edu, John Dickerson, University of Maryland,
College Park, MD, 20742, United States,
dickerson@cs.cmu.eduWe discuss progress toward creating a credit-based matching mechanism for
dynamic barter markets—-and kidney exchange in particular—-that is both
strategy proof and efficient, that is, it guarantees truthful disclosure of donor-
patient pairs from the transplant centers and results in the maximum global
matching. We show that no such mechanism that supports cycles and chains of
any length can be both long-term individually rational and economically efficient;
we then give light assumptions under which such a mechanism can exist.
MA27
201A-MCC
Empirical Research in Operations II
Sponsored: Manufacturing & Service Oper Mgmt
Sponsored Session
Chair: Nils Rudi, INSEAD, 1 Ayer Rajah Avenue, Singapore, 138676,
Singapore,
nils.rudi@insead.edu1 - Fitting, Clustering And Forecasting Product Life Cycles:
Model And Empirical Validation
Jan A Van Mieghem, Harold Stuart Professor, Northwestern
University, 1, Evanston, IL, 60209-2001, United States,
vanmieghem@kellogg.northwestern.eduKejia Hu, Jason Acimovic, Douglas Thomas
We present an approach to fit product life cycle (PLC) curves from historical
demand data and use them to predict/forecast demands of ready-to-launch new
products. We propose three types of models to fit PLC: the BASS diffusion model,
the polynomial model and the piecewise-linear model and compare their
goodness-of-fit and complexity for fitting different categories of products. Using
time-series clustering techniques, we cluster the fitted PLC curves into several
representative patterns. Finally, we validate out-of-sample forecast accuracy using
actual demand data of a computer company.
2 - Managing Multichannel Delivery Of Healthcare Services:
Case Of Telemedicine In Rural India
Kraig Delana, London Business School, PhD Program Office,
London, NW1 4SA, United Kingdom,
kdelana@london.edu,Kamalini Ramdas, Sarang Deo
Telemedicine is a potent intervention to improve healthcare access for difficult-to-
reach populations. We investigate the impact of the introduction of rural
telemedicine facilities on access to eye care for patients in rural India using more
than 4 million patient visit observations from the largest eye care system in the
world. In particular, we exploit growth in the network of telemedicine centers
over time and space to identify changes in where and how early patients seek
care using a difference-in-differences methodology. Our results have implications
for effective multichannel delivery of complex services such as healthcare.
3 - Forecasting Demand For New Products: Combining Subjective
Rankings With Historical Data
Marat Salikhov, INSEAD, Boulevard de Constance, Fontainebleau,
77305, France,
marat.salikhov@insead.eduNils Rudi
We combine subjective ranking inputs with historical data for new product
demand forecasting. The methods yields good fit with data, both for order
statistics of proportions of total demand and for predicting the actual demand.
MA28
201B-MCC
Online Retailing
Sponsored: Manufacturing & Service Oper Mgmt
Sponsored Session
Chair: Dorothee Honhon, University of Texas at Dallas, Richardson, TX,
United States,
dorothee.honhon@utdallas.eduCo-Chair: Xiajun Amy Pan, University of Florida, Gainesville, FL,
United States,
amy.pan@warrington.ufl.edu1 - Probabilistic Selling For Vertically Differentiated Products:
The Role Of Salience
Quan Ben Zheng, University of Florida,
quan.zheng@warrington.ufl.edu,Xiajun Amy Pan,
Janice E Carrillo
This paper studies probabilistic selling for vertically differentiated products,
whereby consumers do not know the exact identity of a product until after
making the purchase. Our work discovers the crucial role of consumers’ salient
thinking behavior: consumers focus on and overweight the salient attribute of a
product in their perception. We show that probabilistic selling can improve the
seller’s profit with salient thinkers even when this strategy does not emerge with
rational consumers. Consumers’ salient thinking behavior enables the seller to
utilize the probabilistic product to transform the consumers’ choice context and
direct their attention to quality.
2 - Maximizing Profitability In Online Retail Through Free
Shipping Threshold
Jiaqi Xu, Carnegie Mellon University, 5000 Forbes Avenue,
Pittsburgh, PA, 15217, United States,
jiaqixu@wharton.upenn.edu,
Gerard P Cachon, Santiago Gallino
We present a data-driven model to analyze the profit implication of an online
retailer’s free shipping threshold decision. A key component of our model derives
from the empirical observation that customers often increase their basket size at
checkout to qualify for free shipping (order padding). We find that a free shipping
threshold policy is effective when the extra sales from order padding do not
substantially reduce the total amount of future purchases, the retailer charges
only a small portion of the fulfillment cost for orders that do not qualify for free
shipping, and product handling costs for returns are low.
3 - Learning From Clickstream Data In Online Retail
Bharadwaj Kadiyala, PhD Candidate, The University of Texas at
Dallas, Richardson, TX, United States,
bharadwaj.kadiyala@utdallas.edu,Dorothee Honhon, Canan Ulu
We study the problem of an e-tailer who learns about consumer preferences from
observing sales or clickstream data on his website in a Bayesian fashion. We use a
ranking-based model to represent consumer choice for two types of products:
basic products for which consumers have well-defined preferences and fashion
products for which consumers discover their preferences via browsing. We prove
that, when the e-tailer learns from clickstream data, it may be optimal to show
products on the search page, but display them as unavailable later on their
product information page. We also numerically estimate the value of learning
from clickstream data versus learning from sales data only.
MA26