Calendar

The A/R/T Forum continues to bring leading researchers from academia and industry together to exchange ideas, discuss new methodologies, and to provide an overview of standard practices. The ongoing mission of the A/R/T Forum is to create and maintain a strong link between technical excellence and methods that is actionable in business practice.

Since its inception in 1990, the A/R/T Forum has brought together leading researchers from academia and industry to exchange ideas, discuss new methods and review standard practices.

Over the course of two and one-half days of programming, attendees are encouraged to share in a lively and open discussion and critique of the various methods. Additional learning opportunities are provided through optional tutorials on a variety of topics:

Call for Papers:  Coming Mid August 2009

SAMPLE PROGRAM -  2008 Program

Sunday, June 15, 2008
Pre-Conference Tutorials

7:30 a.m. - 1:30 p.m. Tutorial Registration
8:15 a.m. - 12:15 p.m. Concurrent Tutorials A, B*, C, D*, and E   (* Tutorial repeats)

8:15 a.m. – 12:15 p.m.

A. Intro to Data Mining
Dan Steinberg, Ph.D. President & Founder, Salford Systems 

 Models in Marketing Research: Introduction
Peter S. Fader, Frances and Pei-Yuan Chia Professor of Marketing, The Wharton School of the University of Pennsylvania
Bruce G.S. Hardie, Professor of Marketing, London Business School

C. Making the Most of the ART Forum: An Exploration of Topics and Methods
Jeff D. Brazell, Ph.D., President and CEO, The Modellers, LLC
Cindy R. Ford, Ph.D., Executive Vice President, The Modellers, LLC

D. Agent Based Modeling and Simulations
Rosanna Garcia, Assistant Professor of Marketing, Northeastern University

E: An Introduction to Partial Least Squares Analysis
Wynne Chin, Professor, University of Houston

1:00 p.m. - 5:00 p.m. Concurrent Tutorials F, G, H, I and J  

F. Introduction to Discrete Choice Modeling
Jon Pinnell, President, MarketVision Research

G. An Introduction to Partial Least Squares Analysis (*)
Wynne Chin, Professor, University of Houston

H. Market Segmentation: Conceptual and Methodological Foundations
Wagner A. Kamakura, Duke University

I. Probability Models for Customer-Base Analysis)
Peter S. Fader, Professor of Marketing, The Wharton School of the University of Pennsylvania
Bruce G.S. Hardie, Professor of Marketing, London Business School

J. An Introduction to Bayesian Statistics and Marketing
Greg Allenby, Helen C. Kurtz Chair in Marketing, The Ohio State University

1:30 p.m. - 7:00 p.m. Conference Only Registration

5:00 p.m. - 5:30 p.m. Welcome Orientation
5:30 p.m. - 7:00 p.m. Welcome Reception (Poster authors available for disccusion)

Monday, June 16, 2007
Conference Papers

8:10 a.m. – 8:15 a.m.         Introductions by Session Chair

8:15 a.m. – 8:45 a.m.         The Effect of Exposure to Fine vs. Broad Survey Scales on Subsequent Decision Making
Vicki Morwitz, Stern School of Business, New York University
Gülden Ülkümen, University of Southern California
Amitav Chakravarti, New York University,

We find that exposure to broad (i.e., few scale points) versus fine (i.e., many scale points) survey scales, alters consumers’ information processing styles, which in turn alters their categorizing and purchasing behavior in a variety of subsequent and unrelated tasks, from basic cognitive behaviors (e.g., grouping), and consumer decisions (e.g., new product adoptions), to more general decision making strategies (e.g., susceptibility to heuristics). Consumers previously exposed to broad survey questions adopt a more lenient, and tolerant processing orientation. In contrast, consumers previously exposed to fine survey questions adopt a careful and critical processing orientation. Consequently, prior exposure to broad questions leads to greater susceptibility to many context effects and decision heuristics, whereas prior exposure to fine questions leads to more careful and considered decisions, invariable across subsequent

                               
8:45 a.m. – 9:15 a.m.         Quantifying and Maximizing Aesthetic Preference
Peter Boatwright, Carnegie Mellon University
Jonathan Cagan, Carnegie Mellon University
Seth Orsborn, Missouri University of Science and Technology

One of the greatest challenges in new product development is the creating of a product form that is attractive to an intended market audience. Just as choice-based conjoint has been successfully utilized to explore product features, we have developed methods that enable and support conjoint analysis to explore consumer preferences within a continuous parametric range of visual aesthetics (physical product forms). We apply our work to vehicle design, where our application can facilitate vehicle design by providing a time and cost efficient method to obtain market research on aesthetic preferences for vehicle design. In general, this methodology will allow product developers to incorporate rich preference feedback from the market about product form, where market preferences can be collected extremely early in the development of the product concept.

10:10 a.m. – 10:40 a.m.     Modeling Customer Decisions as Bayesian Networks
Pieter Sheth-Voss, Research Director, Eidetics

By modeling customer decisions as Bayesian networks, market researchers can uncover and make sense of the complex tangle of customer, product, setting factors that influence current choices and new product opportunity. Bayesian networks encompass many familiar market research models, and can faithfully capture the relationships among dozens or hundreds of variables. As a result, they offer important practical advances in market research design, analysis, and reporting.

10:40 a.m. – 11:10 a.m.     Analysis of Ordinal Models: Issues in Data Collection and Analysis
Thomas Murphy, Qualitative Consultant, MarketTools Inc.
Michael Conklin, Chief Methodologist, MarketTools Inc.
Robert Stephan, VP Advanced Analytics, MarketTools Inc

In this session we explore and compare various methods of data collection for various comparative models such as Bradley-Terry, Plackett-Luce, and MaxDiff. With simulated data we show how some popular methods of reducing respondent burden produce biased results when the data is fit with these comparative models. Taking the methods that produce unbiased results in simulation we show that real respondents cannot actually perform some of these tasks with accuracy. We conclude with some recommended procedures for data collection and model fitting.

1:05 p.m. – 1:35 p.m.         Adaptive Choice-Based Conjoint
Bryan Orme, Sawtooth Software
Rich Johnson, Sawtooth Software
Thomas Otter, Goethe University

Current CBC questionnaires have weaknesses: they are tedious, not very focused on respondent requirements, and they assume compensatory behavior. The authors introduce a new Adaptive CBC method that employs three stages: BYO, Screening Tasks, and Choice Tasks. The approach mimics the purchase process of formulating a consideration set using non-compensatory heuristics (such as “must have” or “must avoid” features), followed by a more careful tradeoff of alternatives within the consideration set using compensatory rules. The logic for constructing adaptive CBC questionnaires is transparent, and the data may be analyzed using standard MNL software.

1:35 p.m. – 2:05 p.m.         Multiplay Conjoint in the Telecom Market
Marco Hoogerbrugge, Senior Methodologist, SKIM
Jürgen Warnecke, Senior Market Analyst, KPN

Telecom providers nowadays are operating in several submarkets simultaneously. For instance they are active in fixed line telephony and mobile telephony, in internet and in the television market. These markets are clearly converging. There is a technical convergence because all submarkets are gradually moving towards IP-based service. There is also a commercial convergence, for instance several suppliers offer consumers a discount when they choose this supplier for several services simultaneously. A ‘multiplay conjoint' model will be presented in which choices in several submarkets are modeled simultaneously.

2:05 p.m. – 2:25 p.m.         Discussion

2:25 p.m. – 2:55 p.m.         Refreshment Break

2:55 p.m. – 3:00 p.m.         Introductions by Session Chair

3:00 p.m. – 3:30 p.m.         Random Regret Minimization: A New Discrete Choice Model
Caspar Chorus, Eindhoven University of Technology
Theo Arentze, Eindhoven University of Technology
Harry Timmermans, Eindhoven University of Technology

This paper presents a new model for riskless and risky discrete-choice analysis. The model is rooted in Regret Theory and is designed to combine the formal and econometrical tractability of Random Utility Maximization Models with the behavioral realism of regret-based semi-compensatory choice models. The model is rigorously tested using data on multiattribute risky choice-behavior in a stated travel mode- and route-choice context.

3:30 p.m. – 4:00 p.m.         Market Share Constraints and the Loss Function in Choice Based Conjoint Analysis
Timothy J. Gilbride, University of Notre Dame
Peter J. Lenk, University of Michigan
Jeff D. Brazell, President/CEO, The Modellers, LLC

Choice based conjoint analysis is a popular marketing research technique for learning about consumers' preferences and making market share forecasts under various scenarios for product offerings. Managers expect these forecasts to be "realistic" in terms of being able to replicate market shares at some pre-specified or "base case" scenario. Frequently, there is a discrepancy between the recovered and base case market share. The authors develop a Bayesian decision theoretic approach to incorporate base case market shares into conjoint analysis via the loss function.

Tuesday, June 17, 2007
Papers

8:15 a.m. – 8:45 a.m.         An Empirical Test of Pricing Techniques
Beau Martin, Senior Marketing Scientist, Market Strategies International
Bob Rayner, VP Marketing Sciences, Market Strategies International

What is the best way to measure pricing? We compare seven different techniques simultaneously using a split design to understand the differences in recommendations when looking at a single product. We uncover biases and best practices from among the following techniques: stated willingness to pay, monadic response (with and without competitive choices), CBC, Van Westendorp, Gabor Granger, and alternative-specific discrete choice. 

8:45 a.m. – 9:15 a.m.         When is a Picture Worth a Thousand Words? A Comparative Analysis of Five Methods for Deriving Coordinates for Market Bi-plot Perceptual Maps
Frank Wyman, Director, Advanced Analytics. MARC
Kanako Oshimura, Senior Analyst, Advanced Analytics, MARC

Discover which multivariate technique provides the best coordinates for bi-plot maps, locating brand points and attribute vectors in two-dimensional space such that the raw brand-by-attribute ratings data are most accurately reflected. First, the basic steps of five competing analytic methods (each based on either correspondence analysis, discriminant analysis, or principal component analysis) will be outlined. Second, the presentation will discuss the design and results of research conducted using data from over 30 real studies to ascertain the relative goodness of the five competing methods in terms of best supporting the classic interpretations of bi-plot maps.

9:15 a.m. – 9:35 a.m.         Discussion

9:35 a.m. – 10:05 a.m.       Refreshment Break

10:05 a.m. – 11:30 a.m.     Session IV (Segmentation)

10:05 a.m. – 10:10 a.m.     Introductions by Session Chair

10:10 a.m. – 10:40 a.m.     Cluster Ensemble Analysis, beyond K-Means
Joseph Retzer, Director- Marketing Sciences, Maritz Research 
Kurt Pflughoeft, Sr VP R&D, Market Probe 

Cluster ensemble or consensus clustering analysis is a relatively new advance in unsupervised learning analysis. It has been suggested as a generic approach for improving the accuracy and stability of “base” clustering algorithm results. Cluster ensembles begin by generating multiple cluster solutions using a “base learner” algorithm(s) (e.g. k-means) and deriving a “consensus” solution more robust and of higher quality than any of the individual ensemble methods used to create it. Empirical results on synthetic data suggest it is capable of detecting non-spherical cluster groupings. Evaluation of consensus clusters must then be based on measures which reward cluster quality but which also do not assume spherical cluster shape. This paper will introduce cluster ensembles and provide evidence of their performance improvement using both synthetic and actual market data. Lastly, it will compare and contrast various measures of cluster quality in order to find a measure well suited to evaluating consensus clusters.

10:40 a.m. – 11:10 a.m.     Segmenting Consumers in a Category Characterized by Frequent, Occasion Based Decisions
John Patrick Pemberton, VP, Sr Director, Research Development, InsightExpress
Erica DesRoches, VP, Director Marketing Research, InsightExpress
Melissa Ludack, Senior Marketing Science Analyst, InsightExpress

Market research segmentation is typically based on attitudes towards a category. For categories associated with frequent purchase decisions marked by variety seeking behavior, attitudes may not be the best basis for segmentation. Segmenting occasions based on need states associated with the occasions is useful for understanding the occasions but lacks direct link back to the consumers who originate the needs. This discussion will identify a consumer level segmentation methodology that better captures occasion based needs that are likely to result in variety seeking product selection.

1:05 p.m. – 1:35 p.m.         A Disaggregate Model of Advertising and Brand Consideration
Greg Allenby, Ohio State University
Nobuhiko Terui, Tohoku University
Masataka Ban, Tohoku University 

The informative and persuasive effects of media advertising are investigated using a scanner panel dataset of detergent purchases and media exposure information. Advertising's informative role corresponds to its influence on the set of brands included for consideration, while its persuasive role allows for a direct effect on the marginal utility. Media exposure is shown to primarily affect brand consideration, not marginal utility, supporting its informative role in decision making. Implications of the need for repeated media exposures, and the effective duration of advertising effects, are explored.

1:35 p.m. – 2:05 p.m.         A Promotion-mix Model with Inter-Brand Competition and Interactions
Arun Bhattacharya, Senior Manager, Organon USA.

Monopolistic competition is gradually setting in the pharmaceutical market where the impact of competition from other firms needs to be taken into account in optimizing promotion-mix. Managers need to know the joint effectiveness of marketing activities of all other competitive brands to be able to determine their own optimal promotion mix plans. Managers should be able to anticipate how other competing brands are likely to react. The impact of such interactions is central to the promotion-mix concept, which this paper intends to address.

2:05 p.m. – 2:25 p.m.         Discussion

2:25 p.m. – 2:55 p.m.         Break

2:25 p.m. – 4:20 p.m.         Session VI (More Methods)

2:55 p.m. – 3:00 p.m.         Introductions by Session Chair

3:00 p.m. – 3:30 p.m.         Methods to Correct for Cross-cultural Differences in Scale use Heterogeneity
Joan Fredericks, Senior Vice President, Harris Interactive
John Bremer, Vice President, Harris Interactive
Randall Thomas, Director, Harris Interactive

There are a number of ways that have been proposed that would allow us to more adequately compare data collected from respondents in different cultures or countries. We review 3 different approaches, and compare the efficacy of each.                             

3:30 p.m. – 4:00 p.m.         Estimating Common Utility Origins and Scales in Discrete-Choice Conjoint with Auxiliary Data
Peter Lenk, University of Michigan
Lynd Bacon, YouGovAmerica/Polimetrix

A limitation of discrete-choice is the loss of inter-subject comparability because the resulting utility structure is invariant to linear transformations. This deficit restricts its application to settings where only within-subject comparisons are meaningful, such as choice share simulation for product or service optimization. This paper proposes using auxiliary data to recover the origin and scale for discrete-choice conjoint to enable between-subject comparisons. The joint model moves the identification constraints from the discrete-choice parameters to the ratings model, thus recovering a common origin and scale.

4:00 p.m. – 4:20 p.m.         Discussion

4:30 p.m. – 5:30 p.m.         Roundtable Discussions

5:30 p.m. – 7:00 p.m.         Reception (Poster Authors available for discussion)

Wednesday, June 18, 2007
Papers

8:15 a.m. - 8:45 a.m.          Measuring the Causal Impact of Social Networks on Marketing Response: The Role of Opinion Leaders
Puneet Manchanda, University of Michigan
Harikesh Nair, Stanford University
Tulikaa Bhatia, Rutgers University

While it is evident that social networks play a very important role in consumer behavior, demonstrating that interaction within the network causes a given agent’s behavior to change is often a very challenging task. This is especially true in the context of revealed preference data. This presentation will identify and explain these challenges. Solutions to overcome these challenges will be discussed and illustrated in the context of physician prescription decisions. The role of key opinion leaders will also be highlighted.               

8:45 a.m. - 9:15 a.m.          Identifying Growth Potentials with Internal Benchmarking Across Product Categories and Markets
Rex Du, University of Georgia
Wagner Kamakura, Duke University

A common question facing many multi-category, multi-market retailers is that, “How well are we doing in each market we serve, for every product category we sell?” Answers to this question are important because the gap between realized sales and potential sales provides an indicator as to how much growth opportunity a retailer could have in each market for each product category. To address this challenge we propose a modeling framework that allows the retailer to benchmark its sales in one market and category against its own sales in all the other markets and categories, factoring in observed as well as unobserved variables that might influence sales. What our approach produces is a measure of relative sales efficiency, reflecting the gap between observed category sales per customer in a market and an estimated "sales frontier," which is largely shaped by observed sales per customer in other categories and markets. Such an internal benchmarking approach is valuable as well as practical, because it allows retailers to identify short-term growth opportunities by product category and market, using readily available data from their own customer databases. 

9:15 a.m. - 9:45 a.m.          Predicting Product Purchase from Inferred Customer Similarity: An Autologistic Model Approach
Sangkil Moon, North Carolina State University, Raleigh
Gary J. Russell, University of Iowa

Product recommendation models are key tools in customer relationship management (CRM). This study develops a product recommendation model based upon the principle that customer preference similarity stemming from prior purchase behavior is a key element in predicting current product purchase. The proposed recommendation model is dependent upon two complementary methodologies: joint space mapping (placing customers and products on the same psychological map) and spatial choice modeling (allowing observed choices to be correlated across customers). Using a joint space map based upon past purchase behavior, a predictive model is calibrated in which the probability of product purchase depends upon the customer’s relative distance to other customers on the map. An empirical study demonstrates that the proposed approach provides excellent forecasts relative to benchmark models for a customer database provided by an insurance firm

9:45 a.m. - 10:05 a.m.       Discussion

10:05 a.m. - 10:25 a.m.     Break

10:25a.m. - 10:30 a.m.      IIntroduction of Paul Green Award Winner

10:10 a.m. - 10:40 a.m.     Panel Discussion

11:15 a.m. - 11:30 a.m.     Announcement of Best Presentation Award (Closing Remarks by 2009 Chair )

11:30 a.m.                        Conference Adjourns

12:30 p.m. – 4:30 p.m.         4-hr Tutorials K, L, M, N and O*        (*Repeat Tutorials)

K. Advanced Data Mining for Marketers
Dan Steinberg, Ph.D. - President & Founder, Salford Systems

L. Agent Based Modeling and Simulations
Rosanna Garcia, Assistant Professor of Marketing, Northeastern University

M. Advanced Bayesian Statistics and Marketing
Greg Allenby, Helen C. Kurtz Chair in Marketing, Fisher College of Business, The Ohio State University

(NEW)
N. Advanced Market and Customer Segmentation
Wagner A. Kamakura, Duke University

(REPEAT)
O. Advanced Topics in Discrete Choice Modeling
Jon Pinnell, President, MarketVision Research
Bryan Orme, President, Sawtooth Software