9:00 am to 9:30am — Registration and Continental Breakfast

9:30am to 12:00pm — Markov Modeling in CRM: Improve Customer LTV by Optimizing Customer Migrations

This session will introduce the concept and methodology behind the powerful technique of Markov Modeling. This technique has been widely used in diverse fields such as engineering, biomedical science, and supply chain management. Its applications in direct marketing are relatively new, but this versatile analytical tool allows marketers to perform sophisticated and practical marketing optimizations.

To illustrate the potential and power of these toolsets, we will provide a range of relevant and real-life case studies that show how professional marketers can take advantage of this powerful approach. We will focus on homogeneous Markov model, Semi-Markov model, Hidden Markov Chain, and Markov Decision Processes. This session will make Markov Modeling accessible to general Direct Marketing professionals.

This session will cover:

• Strategic-level marketing resource optimization
• Customer purchase cycle, driver analysis and forecasting
• Customer brand preference and product category preference migration
• State-machine based life time value estimations using Markov chain models
• Detect customer relationship pattern changes using hidden Markov chain
• Optimize direct marketing longitudinal contact strategy using Markov decision processes

Speakers:
Hongjie Wang, VP Customer Analytics, Fulcrum Analytics, manages the analytical team at Fulcrum. He has developed statistical and optimization models, as well as segmentation solutions for a diverse range of clients to support their marketing and management operations. He has also been the chief designer of several of Fulcrum’s analytical solutions, including Incremental Value Modeling, Customer Equity Models, and Cross-sell Optimization, which have been used to improve business results for Fulcrum’s retail clients. Mr. Wang co-authored several papers in analytical and trade journals. He was a speaker at NCDM, DMA seminars and guest-lectured at Carnagie Mellon University, Tepper Scholl of Business. Mr. Wang earned his B.S. in Mathematics and Machine Learning and M.S. in Operations Research from Virginia Tech.

David King, CEO, Fulcrum Analytics, has served in his current role since December 2001. David has been working with clients in the areas of CRM, marketing, privacy, and security for over 18 years. Before coming to Fulcrum Analytics, David was a senior executive at KnowledgeBase Marketing developing data warehouses, implementing statistical models, and CRM architecture. Earlier in his career, he worked as a modeling analyst, applying neural networks, self-organizing maps, and genetic algorithms to marketing problems. David received his education from Wake Forest University and the University of North Carolina at Chapel Hill.

12:00pm to 1:00pm - Lunch

1:00pm to 2:00pm – Contrasting Likert Scales and Discrete Choice Methodologies

Likert scales (or scalar methodologies) have been the standard for capturing attitudes and opinions for brands, products, and other marketing concepts. Currently, non-scalar methodologies have been increasingly used in market research as the technological advancements of the internet allow for improved delivery of discrete choices.

In this seminar, our expert speaker, Rus Kehoe will compare and contrast scalar and non-scalar methodologies, while explaining how to generate utility scores and classifications. You will learn the process to capture data for both Likert scale (scalar) and Discrete Choice (non-scalar) methodologies. To display the differences between the two methodologies, we will review the results of a case study for Cheesecake flavor choices. This study will include both utility scores for each flavor to demonstrate the popularity of each flavor, and allow for a comparative ranking. The results will show the classifications with groups segmented by their similar flavor preferences.

Key takeaways from this session will be:

• How non-scalar is a powerful methodology to capture consumers preferences and overcome many of scalar methodologies’ biases
• When both scalar and non-scalar methodologies are capturing preferences, the results that demonstrate wider differences in utility scores lead to more confidence in making actionable strategies
• The seminar will be accessible and equip researchers with more tools to capture consumer preference

Speaker:
Rus Kehoe, Market Research Manager, Insight Express, is involved in questionnaire design, data processing, data analysis, and quantitative methodologies for InsightExpress, an online market research firm in Stamford, CT. As a member of the Marketing Sciences team for IX he has supported pricing studies, segmentation, and Idea Gauge studies for CPG firms and financial service institutions. Additionally, he has provided analyses for the Online Media Measure Group which measures the effectiveness of online ads. He has been a market researcher for the past 14 years with IRI, NYC Informatix, Philip Morris, The NPD Group, United Way NYC, and The Edison Project. Rus holds a M.S. in Social Research from Hunter College, and B.S. in Economics and Finance from Bentley College.

2:00pm - Break

2:15 to 4:30 – Classification and Data Mining

Binary Classification is the activity par excellence in Data Mining. The usual modeling methodologies such as logistic regression produce probability estimates for each case and the practitioner typically decides on a cutoff point. We will review the state of the art in this area (Roc, AUROC, sensitivity, specificity, etc) with added emphasis on the aspect of prediction as opposed to misclassification error. In this context, we will contrast the applicability of these measures both in terms of model and cutoff selections. While there is no necessary “winner”, the audience will benefit from a clear exposition of these topics. For ease of presentation, we will emphasize graphical aspects of the topic.

When you leave this session you will...

• Understand issues of classification and category prediction.
• Understand cutoff determination.
• Understand present tools of model comparison in classification.

Leonardo Auslender, SAS Institute, is a statistician and economist with more than 25 years of business experience and SAS expertise, at present in the Data Mining Research and Development group of SAS Institute. His area of expertise is in the area of Giga-Data Analysis and Methods, and has written papers and given lectures on Missing Value Imputation, Classification Trees, Support Vector Machines, Market-Basket Analysis, Variable Selection in Giga-Bases, Database Marketing, CRM, GDP and (Relative Price) Inflation studies, Expectation Formations, Productivity and Technology effects in the economy, and most recently on Colinearity and malaise in linear modeling. He was a lecturer of Finance and Macroeconomics at Rutgers University. His present interests are in the area of variable selection, Bayesian networks, and Bayesian and Tree methods.

 

© Direct Marketing Association | Privacy Statement