Events

 

 

Event Sponsored By:
           
      StatWizards           Sawtooth Software       Statistical Innovations

                   

          
Sponsorship and exhibit opportunities are available. Reserve your space today by contacting Lore Torres at 312.542.9033 or ltorres@ama.org.

A. Introduction to R for Marketing Researchers
Chris Chapman, Google and Steven Ellis, Google
Introduction to R for Marketing Researchers is a hands-on introduction to the R statistics platform focused on marketing applications. The course presents the basics of the R environment, R language, statistical procedures, and graphics. The tutorial begins with the basics of data processing in R and then examines common statistical analyses such as regression models (classical and Bayesian versions). It then introduces graphics and plotting in R. The tutorial will work with a dataset that exemplifies common types of marketing analyses (product sales by store given varying promotions, prices, and store-level covariates). The final section discusses applied topics for marketing research, including an overview of R options to explore for clustering and segmentation, choice models, and automated reporting.

B. MaxDiff: Opportunities and Challenges
Bryan Orme, Sawtooth Software and Keith Chrzan, Sawtooth Software

MaxDiff extends the traditional method of paired comparisons.  Most often used for measuring the importance/preference for an array of items, it also applies to any context where the researcher wants to measure the intensity of a perception or belief.  MaxDiff avoids scale use bias (particularly useful for cross-cultural research) and leads to greater discrimination and predictive validity than ratings-based scales.   Other competing methods for scaling will be mentioned and compared, including rating scale, Q-sort, magnitude estimation, constant sum scaling, and an unbounded rating scale.

Some researchers have cautioned that MaxDiff lacks sound error theory that appropriately links the model specification to the way that respondents answer MaxDiff tasks.  In other words, the statistical model does not properly account for the data generation process.  We’ll point out the objections to MaxDiff, the modeling challenges, and then state our position regarding best practices and the practical significance of these issues.

Regarding experimental design, methods available for designing MaxDiff

experiments include catalogue plans (e.g. BIBD), computer-generated plans (Sawtooth Software) and efficiency search plans (SAS).  The pros and cons of such methods will be covered.  Analysis methods include counting, logit, latent class, and HB.  Attendees may use the information contained in this tutorial to perform their own analysis using any commercial or open-source MNL system.

Anchored MaxDiff is an attempt to overcome the ipsative (relative) nature of MaxDiff scaling.  Louviere’s indirect method and the direct binary scale methods are two recent approaches.   We demonstrate how to do these procedures and present evidence of usefulness and validity (as well as pitfalls).

Finally, we focus on the practical uses of MaxDiff in a variety of research contexts.  One that prominently features in practitioners’ toolboxes is market segmentation.  We will also cover simulation via TURF-based methods for portfolio optimization.
 
C. An Introduction to Probability Models for Marketing Research
Peter S. Fader, The Wharton School of the University of Pennsylvania and Bruce G. S. Hardie, London Business School

Central to a complete understanding of today’s “leading-edge” market research techniques is a sound intuitive appreciation of the basic foundations upon which these sophisticated tools are built. For example, both hierarchical Bayes models and latent class models build on simple probability modeling concepts (e.g., zero-order choice process, Poisson counts, conditional expectations, and exponential interpurchase times) — yet how many researchers are comfortable at precisely defining these concepts or explaining the motivation for using them?

This tutorial aims to fill in these gaps by bringing practitioners fully up to speed on the basic methods that may underlie many of their current or future research activities. Our two broad objectives are (1) to review the basic terminology and logic associated with the area of probability models as applied to marketing research problems, and (2) to develop participants’ skills through a set of case studies that demonstrate the model building process in detail. We will illustrate all of the steps required to develop a probability model, estimate its parameters, and interpret the results. Careful and extensive use is made of the Solver tool in Microsoft Excel, which makes it possible to construct all of these models within a familiar spreadsheet environment. By the end of the tutorial, participants should be quite comfortable with all of the aforementioned principles and models and the managerial issues that surround them.

Customer-base analysis seeks to use information on the history of customer purchase patterns to identify which individuals are most likely to be active (or inactive) customers and to predict future purchasing patterns by those customers listed in the firm’s transaction database. Any researcher hoping to make statements about “customer lifetime value” must deal with these issues, but unfortunately the set of commonly available tools is not well-suited for the task. 

This tutorial builds upon the basic “platform” provided in our introductory tutorial to provide a set of techniques and models tailored to address these situations properly. We focus on developing the models entirely in Excel and provide attendees with the relevant spreadsheets and notes on how to implement the models “from scratch”. Our goal is to provide the attendee with tools that can be applied immediately (maybe with some slight

modifications) at his/her place of work.

The structure of the tutorial is as follows:

• Introduction to the idea of customer-base analysis
• Overview of the concept of Customer Lifetime Value (CLV) and the presentation of a general framework for its calculation
• Brief review of the probability modeling basics required for model building (e.g., review of binomial, geometric, Poisson, exponential, gamma, and beta distributions; discussion of common mixtures such as the NBD, beta-geometric, and beta-binomial)
• Presentation of probability models that can be used to answer various managerial questions including the calculation of CLV. (The empirical examples come from settings as diverse as e-tailing, the charity sector and media subscriptions)
• Generalizations of the specific models presented in this tutorial making links to the broader modeling literature

In the past decade, massive shifts in how companies interact with their customers have suddenly made field experiments an economically feasible way to learn about a variety of marketing questions such as what types of promotions are most effective, what products should be stocked at a store, how promotions should be designed, how sales staff should be compensated, how advertising should be designed, etc. Many marketers engaged in online retailing, direct-marketing, online advertising, media management, service operations, and many other marketing-related fields are rapidly embracing a “test and learn” philosophy and a number of platforms such as Adobe Test + Target and Google Content Experiments, have been developed to facilitate rigorous field experiments in the online environment. Just as with the quality revolution in manufacturing during the 1980s and 1990s, the rapid rise of the “test and learn” philosophy in marketing has created a huge demand for those who can design, field, and analyze experiments. Through this course, you will get an overview of critical skills for experimentation, from the statistical methods used to design and analyze experiments to the management and strategy required to execute an experiment and act on the results.

Greg will discuss and distribute his book "Seven Summits of Marketing Research: Decision-Based Analytics for Marketing's Toughest Problems."  The summits are 1) Market Definition; 2) Market Segmentation; 3) Customer Satisfaction; 4) Product Analysis; 5) Pricing Analysis; 6) Advertising Analysis and 7) Optimization.  Their session will concentrate on Chapters 4 and 5 of the book, showing how decision-based analytics can be used to estimate the relationship between needs and wants in a product category; and how this relationship can be developed within the context of discrete and volumetric demand modeling.  The book comes with an interactive decision tool that can be downloaded for free.

The power of R to rapidly process large data sets and perform complex statistical analyses goes well beyond the capabilities of other software. However, the command line interface makes R a decidedly unfriendly environment to deliver results to end users who may be unfamiliar with R. Participants in this tutorial will be guided through simple but informative hands-on examples using programming tools – RExcel-VBA, Amazon Elastic Compute Cloud (EC2) and C++ – that can be used to easily deliver complex analyses run in real time in an interface that is decidedly more friendly than R.

 

Many firms are routinely running A/B and multivariate tests to fine-tune their digital advertising, but there are more effective ways for them to “earn while they learn.”  We show how to solve the “exploration-exploitation” dilemma (known as the “multi-armed bandit” problem) in a practical manner. We apply these methods in a large-scale field experiment focusing on customer acquisition via banner ads for a major financial services firm. 

 

The research uses propensity matching scores to evaluate the results of a large scale field experiment conducted by Intercontinental Hotel group (IHG).  The experimental data will be analyzed using both simple regression techniques to understand the final implications of the experiment and a dynamic structural model of behavior to understand the long-term loyalty implications of frequent buyer program dimensions. 

 

Real-Time Measurement of Brand Health with Continuous Tracking Surveys
Rex Yuxing Du, Marvin Hurley Associate Professor of Marketing, University of Houston, Xueming Luo, Eunice & James L. West Distinguished Professor, University of Texas at Arlington
We develop a novel statistical model that makes more efficient use of data commonly collected through continuous brand tracking surveys, a staple in marketing research practitioners’ repertoire. We show how the proposed model can be used to extract a brand health index that can be updated reliably on a daily basis, without incurring additional costs. The resulting brand health monitoring system combines the strengths of online consumer interest tracking (i.e., relatively low cost and real-time measurement) with the strengths of tracking surveys (i.e., superior sampling control). We apply our method to raw continuous brand tracking data, with 12 consumer perception indicators per brand per day, between 2008 and 2011, for 23 automotive, 16 banking and financial services, and 10 oil and gas brands. We demonstrate that, although the raw data are quite noisy and thus unreliable at the daily level, the proposed method can extract brand health indexes that are updated daily and show clear trend lines, whose major movements and turning points coincide nicely with known major historical events. 

 

Does Sample Size Still Matter?
David G. Bakken, Chief Insight Officer, KJT Group, Inc., Megan K. Bond, Vice President, KJT Group, Inc.
In this paper we explore the impact of inferences from very [DGB] small samples on the outcome of management decisions.  In many cases management has some prior belief about the states of nature.  We explore the potential advantage of incorporating Bayesian inference to improve the confidence in managerial decisions based on small samples..  Traditionally, survey researchers reconcile differences between survey results and prior beliefs by citing the uncertainty reflected in the sampling error or looking for other explanatory factors (such as possible survey measurement error).  The Bayesian approach integrates the different sources of information (i.e., prior belief and observed survey results) to arrive at the most probable estimate.  In full realization, a Bayesian approach considers not just the probability that “truth” lies outside some range of values but seeks to estimate the probability of each of many possible hypotheses, given the data was that obtained.

    
11:50 a.m. - 1:20 p.m. Lunch - Parlin Award Presentation and Recipient Remarks
 

1:20 - 4:30 p.m. Monday Afternoon Sessions 
 
Session 3: Issues in Advanced Conjoint Design (Mike Mulhern, Session Chair)

Respondent Heterogeneity, Version Effects or Scale?
Keith Chrzan, SVP Sawtooth Analytics, Sawtooth Software, Aaron Hill, VP, Client Services, Sawtooth Software
Utilities in HB analysis of a discrete choice experiment differ from respondent to respondent.  Preference heterogeneity is that portion of the heterogeneity that doesn’t owe to other explanations – two salient ones being version effects and differences in respondent reliability (scale).   Scale effects may be easier to remove from DCE data but if version effects are strong and a design uses a large number of versions, respondent heterogeneity may contain an excessive amount of preference-irrelevant version-related heterogeneity.  This presentation attempts to separate the different sources of respondent heterogeneity to identify how much may be explained by preference-irrelevant factors like scale and version effects.  If version effects explain a greater proportion of variance, design strategies using large numbers of versions would be less attractive for studies in which segmentation is an important objective.

         
Dynamic Menu Choice – A New Tool for an Old Problem
Paul Markowitz, Senior Manager, Global CIG, Bain & Company, Inc.
As practitioners, task complexity is something we face every day.  We routinely make trade-offs around consumer decisions we want to simulate, attributes we want to test and drivers of choice we want to uncover.  In menu choice tasks, complexity is an even larger issue – tasks that appear simple are actually complex and those that appear complex are beyond what we should reasonably ask consumers to complete.  This paper presents a new method for individually customizing menu-based discrete choice tasks. The task is customized to eliminate undesirable options and focus attention on meaningful trade-offs.  The resulting tasks are cognitively easier for consumers to answer and yield better data. In addition, model estimation is simpler and more straight-forward. 

 
Session 4: Conjoint Analysis and Product Design (Jane Tang, Session Chair) 

Generating Perfect Package Designs by using Dynamic Optimization Methods
Carlo Borghi, Eline van der Gaast, Virginie Jesionka and Gerard Loosschilder (SKIM)
Following our successful presentation at the 2012 ART conference, we further explore the field of visualization research. This time, we will take on a different aspect of visual communication: advertisement and package design. We will create an optimization algorithm that can efficiently and effectively navigate an extremely large space of possible designs to identify the optimal design consisting of visual, graphic, and textual elements. By making use of all information collected up to a point (a first subset of all surveys to be completed), we can identify areas of the solution space with high potential during the data collection process (so in real time). We then use this information to steer the rest of the data collection process towards studying the high potential areas of the solution space in greater detail. We believe that this will result in a higher likelihood that we find the “real” optimal solution that has higher internal and external validity. In this paper, we demonstrate our method based on an actual study.

 

Designing Health Plans for an Health Insurance Exchange - Conjoint and Shopping Simulation
Scott "Rocky" Shook Ph.D., Manager, Market Research & Insights, Blue Cross Blue Shield of Minnesota, Kathryn Martin, Strategic Planning Consultant, Blue Cross Blue Shield of Minnesota, Michael J. Walsdorf, Director of Research, Stonegate Advisors, Marc E. Pierce, President & Founder, Stonegate Advisors, Alok Verma, Director, Stonegate Advisors, Nico Peruzzi, Ph.D., Outsource Research Consulting
Health care reform will make health insurance available to millions who are currently without health insurance through health insurance exchanges.  Health insurance exchanges will provide an online shopping experience with products regulated by state departments of insurance and other regulators.  Blue Cross Blue Shield of Minnesota and Stonegate Advisors worked together on a two-part study: an Adaptive Choice Based Conjoint (ACBC) design was used to identify product attributes that drive choice; a simulated online shopping experience helped us identify health plans that people will choose, while at the same time providing deep insight into the shopping habits people will use when shopping on an online health insurance exchange.  

 

Session 5: Advances in Conjoint Analysis (Jeff Dotson, Session Chair)
 
Evaluating Choice Models of Price Promotion
John R. Howell, Ph.D. Candidate, Ohio State University, Sang Hak Lee, Assistant Professor, University of Iowa, Greg M. Allenby, Helen C. Kurtz Chair in Marketing, Ohio State University
Many firms use volume based price promotions to drive customer behavior.  Volume based pricing is not well suited to existing discrete choice models that assume customers only purchase one unit.  We demonstrate a model that properly accounts for different volume based pricing schemes such as volume discounts, coupons, and buy-one/get-one deals.  Using this model and an online conjoint experiment, we test these different pricing schemes to demonstrate how this model can properly account for the different observed behaviors these pricing structures induce. 

Estimating Individual-level Choice Models
Bart D. Frischknecht, Senior Research Fellow, Centre for the Study of Choice (CenSoC), UTS Business School, University of Technology Sydney, Christine Eckert, Senior Lecturer, Marketing Discipline Group & CenSoC, UTS, John Geweke, Distinguished Research Professor, Economics Discipline Group & CenSoC, UTS, Jordan J. Louviere, Distinguished Research Professor, Marketing Discipline Group & CenSoC, UTS
This paper demonstrates a method for estimating logit choice models from discrete choice experiments for small sample data, including single individuals, that is computationally simpler and relies on weaker prior distributional assumptions compared to hierarchical Bayes estimation. We show how this method is particularly well suited for estimating values of choice model parameters from small sample choice data thus opening this area to the application of choice modeling. For larger sample sizes of around 100-200 respondents, preference distribution recovery is similar to hierarchical Bayes estimation of mixed logit models for the examples we demonstrate. The method should be of interest to research practitioners and academics because the models are simple to understand and implement, the method allows the study of individual decision makers in a way not previously possible, and the models are quick to estimate and are natural for parallel computing. We demonstrate model implementation in Excel, Matlab, and SAS using non-proprietary code available from the authors.  

 

Session 6: Leveraging Social Network Data (Jeff Dotson, Session Chair) 

Identifying High Value Customers in a Network:  Individual Characteristics Versus Social Influence
Qin Zhang, Assistant Professor of Marketing, University of Iowa, Sang-Uk Jung, Lecture of Marketing, University of Auckland, New Zealand, Gary J. Russell, Henry B. Tippie Research Professor of Marketing, University of Iowa
Firms are interested in identifying customers who generate the highest revenues.  Traditionally, customers are regarded as isolated individuals whose buying behavior depends solely on their own characteristics.  In a social network setting, however, customer interactions can play an important role in purchase behavior.  This study proposes a spatial autoregressive model that explicitly shows how individual characteristics and network effects interact in generating firm revenue.  Using model output, we develop a method of identifying individuals whose purchase behavior most impacts the total revenues in the network.  An empirical study using a user-level online gaming dataset demonstrates that the proposed model outperforms benchmark models in predicting revenues.  Moreover, the proposed value measure outperforms a variety of benchmark measures in identifying the most valuable customers.  

Using Hidden Markov Models to Identify Job Seekers from Social Network Data
Peter Ebbes,  Associate Professor , HEC Paris, Oded Netzer, Associate Professor, Columbia Business School
The majority of LinkedIn’s revenue comes from providing job search solutions. However, one of the key challenges for LinkedIn is to identify job seekers, as most job seekers will not publicly announce they are searching for a job.š In this research we build a hidden Markov model (HMM) to detect job seekers by tracking members’ activity on the LinkedIn website.š We demonstrate the use of a HMM to uncover latent behavior from observed large-scale data, solving one of the primary business questions of a major social network company.š Methodologically, we demonstrate how HMMs can be used to naturally fuse longitudinal panel data with one time survey data. 

 

 

Improving Latent Class Conjoint Predictions with Nearly Optimal Cluster Ensembles
Kevin Lattery, VP Marketing Sciences, Maritz Research
Cluster ensembles often improve over simple segmentation.  In this paper, we take a similar approach to Latent Class analysis of conjoint data.  We create multiple segmentation solutions by relaxing convergence criteria of Latent Class analysis to one of near optimality.  The multiple Latent Class segmentations provide a posterior distribution of the predicted probabilities for each respondent.  We show how synthesizing these predictions significantly enhances respondent level heterogeneity and improves respondent level predictions.  

 
Combining Latent-Class Discrete Choice Models with Non-Linear Optimization Methods
George Boomer, Managing Member, StatWizards LLC, Richard Greenburg, Manager, Product/Project-Customer Energy Efficiency and Solar Dept., Southern California Edison, Anne Dougherty, Manager of Social and Behavioral Research, Opinion Dynamics Corp.
The California Public Utilities Commission and the utilities it regulates take conservation very seriously.  If conservation programs run by utilities can demonstrate a reduction in energy use beyond what individuals do by themselves, they can claim earnings; if they can’t, the programs die.  At stake are millions of dollars in earnings and thousands of tons of atmospheric carbon.  For programs that include price incentives such as LED subsidies, demonstrating effectiveness is particularly difficult, because bulbs differ in energy efficiency, type, use, price, and energy savings; and users differ in preferences for LED products.  To calculate optimally effective incentives by bulb, we built a non-linear programming model linking an Excel-based latent-class discrete-choice model to Southern California Edison’s program-planning workbook.

Session 8: Applications of MaxDiff (Bruce Hardie, Session Chair)
 

Models of Sequential Evaluation in Best-Worst Choice Tasks
Tatiana Dyachenko, Ph.D Candidate in Marketing, Ohio State University, Rebecca Walker Naylor, Assistant Professor of Marketing, Ohio State University, Greg M. Allenby, Professor of Marketing, Ohio State University
We examine the nature of best-worst data for modeling consumer preferences and show that the best and worst responses do not originate from the same data-generating process. We propose a sequential evaluation model and apply it to data from a national survey investigating concerns associated with hair care.  Our analysis finds support for order and elicitation effects as well as evidence of global inference retrieval in choice tasks, which can be represented by the central limit theorem and normally distributed errors as opposed to extreme value errors. 

Anchored Adaptive MaxDiff: Application in Continuous Concept Test
Jane Tang, SVP, Advanced Analytics, Vision Critical, Rosanna Mau, Director, Advanced Analytics, Vision Critical, LeAnn Helmrich, Senior Vice President, Consumer Insights, Vision Critical, Maggie Cournoyer, Senior Research Manager, Consumer Insights, Vision Critical
Many firms have a continuous concept test program based on monadic or sequential monadic ratings.  MaxDiff is superior to ratings, but does not lend itself easily to tracking across the many waves of a continuous program.  The anchoring transforms the relative preferences of the concepts into an absolute scale.  We look into how an anchored adaptive MaxDiff can be set up in this environment so that all the concepts tested are comparable across the different testing periods. 

4:00 - 4:45 p.m. Speaker Roundtables

Wednesday, June 12 

Registration 8:00 a.m. - 1:00 p.m. 
8:00 - 8:30 a.m. Breakfast 
8:30 a.m. - 12:00 p.m.  Wednesday Morning Sessions 

Session 9: Big Data (David Schweidel, Session Chair)
 
Online Advertising Response Models: Incorporating Multiple Creatives and Impression Histories
Wendy Moe, Associate Professor of Marketing, University of Maryland, Michael Braun, Assistant Professor, MIT Sloan
Online advertising campaigns often consist of multiple ads, each with different creative content. We propose a model that evaluates the effectiveness of each creative in a campaign given the targeted individual’s ad impression history, as characterized by the timing and mix of previously seen ad creatives.  We examine the impact that each ad impression has on both visitation and conversion behavior at the advertised brand’s website. Our model is constructed at the individual level and takes into account correlations among the rates of ad impressions, website visits and conversions.  We also allow for the accumulation and decay of advertising effects, as well as ad wear-out and restoration effects. Our results highlight the importance of accommodating both the existence of multiple ad creatives in an ad campaign as well as the impact of an individual’s ad impression history.  We demonstrate with a simulation how this modeling approach can be used for online ad targeting.  Specifically, our results suggest that, using our model, online advertisers can increase the number of website visits and conversions by varying the creative content shown to an individual according to that individual’s history of previous ad impressions.  For our data, we show a 12.7% increase in the expected number of visits and a 13.8% increase in the expected number of conversions. 

Little Data Embraces Big Data
Paul Neto, Research Director, YuMe
The Introduction of Big Data and Predictive Analytics with Online Survey Based Research - This session explores the methodology, opportunity and challenges for leveraging online survey research, big data and predictive analytics to power marketing insights and drive ROI in online advertising. 

 
Agile Analytics and the use in Data Mining and Predictive Analytics
Kevin Dang, Senior Research Manager - Advanced Analytics, Vision Critical
Over the years, companies have accumulated a significant amount of data on their panelists. For many analytics practitioners it has always been a challenge to access this data in an agile analytic environment outside the traditional enterprise data warehouse and business intelligence (BI) systems.  In this case study we will look at the process of creating an analytics sandbox, the mid-step between ad hoc reporting and a full BI solution, and how we can leverage the environment for data mining and predictive analytics. The presentation will cover the challenges and lessons learned with the development and use of a ‘sandbox’ environment in data mining and predictive analytics for panel management. 

Session 10:  Paul E. Green Award (Bruce Hardie, Session Chair) 
This award recognizes the best article in the Journal of Marketing Research that demonstrates the greatest potential to contribute significantly to the practice of marketing research.  It honors Paul E. Green, Professor Emeritus of Marketing, Wharton School and S.S. Kresge Professor Emeritus of Marketing, University of Pennsylvania.  

H. Introduction to Text Mining and Sentiment Analysis
Matt Taddy, University of Chicago
The internet is awash with rich text information for market researchers, including product reviews, blogs, and social media. Yet this free-form linguistic content is both unstructured and high-dimensional, requiring new data analysis tools and concepts. This tutorial presents basic concepts of text mining and extraction, and will show how to analyze text data and model relationships with consumer sentiment and other drivers of product experience. Specific topics include working with text data sources, parsing linguistic information, data exploration and dimension reduction, and modeling relationships between text elements and consumer sentiment. The tutorial will be highly hands-on, and will lead participants through exercises with real data sets. Attendees should bring a laptop and install the R statistics environment in advance. (Detailed R experience is not required.)

 

I. Advanced Applications of Hierarchical Bayes Choice Models
Jeff Dotson, Vanderbilt University and Elea McDonnell Feit, Wharton Customer Analytics Initiative 
The popularity of software packages like CBC/HB and bayesm have made hierarchical Bayes (HB) a standard tool for many market researchers, however many of those who have learned HB through software have never been exposed to the basic theory behind Bayesian inference. This tutorial will give students grounding in that theory, but with a strong emphasis on how that theory applies to a number of practical issues that arise in the advanced application of choice models. Tutorial participants will gain a deeper understanding of how HB choice "works" as well as some concrete “tricks” that can be applied on their next choice modeling project.

 

Topics covered include:
• Understanding the uncertainty in my estimates (posterior draws)
• Dealing with small sample sizes (informative priors)
• Combining data from multiple sources (error scale)
• How to include information about respondents in my model (covariates in the upper level model)
• Building my own simulator (posterior draws)
• Making accurate predictions of substitution and cannibalization (Independence of Irrelative Alternatives and ways to relax that assumption)
• What's really going on when I 'run' a model and why doesn't it always work? (Metropolis-Hastings algorithms, with specific focus on discrete choice and response models)
• What questions should I ask in my conjoint study? (design of choice experiments)
 

             

J. MaxDiff: Opportunities and Challenges
Bryan Orme, Sawtooth Software and Keith Chrzan, Sawtooth Software 
MaxDiff extends the traditional method of paired comparisons.  Most often used for measuring the importance/preference for an array of items, it also applies to any context where the researcher wants to measure the intensity of a perception or belief.  MaxDiff avoids scale use bias (particularly useful for cross-cultural research) and leads to greater discrimination and predictive validity than ratings-based scales.   Other competing methods for scaling will be mentioned and compared, including rating scale, Q-sort, magnitude estimation, constant sum scaling, and an unbounded rating scale.

Some researchers have cautioned that MaxDiff lacks sound error theory that appropriately links the model specification to the way that respondents answer MaxDiff tasks.  In other words, the statistical model does not properly account for the data generation process.  We’ll point out the objections to MaxDiff, the modeling challenges, and then state our position regarding best practices and the practical significance of these issues.

Regarding experimental design, methods available for designing MaxDiff experiments include catalogue plans (e.g. BIBD), computer-generated plans (Sawtooth Software) and efficiency search plans (SAS).  The pros and cons of such methods will be covered.  Analysis methods include counting, logit, latent class, and HB.  Attendees may use the information contained in this tutorial to perform their own analysis using any commercial or open-source MNL system.

Anchored MaxDiff is an attempt to overcome the ipsative (relative) nature of MaxDiff scaling.  Louviere’s indirect method and the direct binary scale methods are two recent approaches.   We demonstrate how to do these procedures and present evidence of usefulness and validity (as well as pitfalls).

Finally, we focus on the practical uses of MaxDiff in a variety of research contexts.  One that prominently features in practitioners’ toolboxes is market segmentation.  We will also cover simulation via TURF-based methods for portfolio optimization.

In Search of the Holy Grail of Marketing ROI:  Developing a Channel Assessment Process to Identify the Most Impactful Channels and Tactics with Physicians
Suzanne Ursu, Group Manager, Marketing Strategy - Boston Scientific Corporation,  Michael Fix, Manager, Market Research & Intelligence - Upsher-Smith Laboratories

Structural Equation Modeling: The Bridge Between Attitudes And Sales
Noah Ganter, Ph.D,  Vice President, Advanced Solutions, Nielsen, Cherie Kientoff, Ph.D, Research Analyst, Nielsen

Addressing Estimation Challenges in Discrete Choice Modeling
Kurt A. Pflughoeft,  Director of Marketing Sciences, Maritz Research, Joseph J. Retzer, Vice President of Advanced Analytics, Metrix Lab

Optimal Pricing Under Consumer Uncertainty:  Individual Reservation Prices and Strategic Firm Behavior
Michael S. Morgan, President / CEO, Morgan Analytics, Inc.

Marginal Structural Models to Evaluate Promotional Campaigns
Yong Cai, Director, Advanced Analytics, IMS Health, Yilian Yuan, Vice Presient, Statistics & Advanced Analytics, IMS and Health, Anindita Basu, Principle & COE, Commercial Effectiveness Services, IMS Health

Markov Choice Network
Jingsong Cui, Director, Nielsen BASES, Scott Appel, Director, Nielsen BASES

Integrating Multiple Choice Models in Forecasting
Jing Jin, Associate Manager, Nielsen BASES, Julie DiPopolo, Associate Manager, Nielsen BASES

Identifying Surrogate Geographic Research Regions with Advanced Exact Test Statistics
Steven Ellis, Quantitative User Experience Researcher, Google, Inc.

Conjoint Analysis in R … Now with Individual-Level Utilities and Survey Mockups
Chris Chapman, Senior Quantitative Experience Researcher, Google, Inc., Steven Ellis, Quantitative User Experience Researcher, Google, Inc.

Does the Analysis of MaxDiff Data Require Separate Scaling Factors?
Jack Horne, Vice President, Marketing Sciences, Market Strategies International, Bob Rayner, Senior Vice President, Marketing Sciences, Market Strategies International, Ray Reno, Senior Vice President, Marketing Sciences, Market Strategies International