INFORMS Marketing Science Conference 2018: Observations from the Conference
by John Colias, Ph.D.
INFORMS is one of the world’s largest analytical organizations, with over 12,000 members worldwide. The initials stand for Institute for Operations Research and the Management Sciences. I recently attended INFORMS Marketing Science Conference at Temple University in Pennsylvania. Here are my main takeaways, particularly focusing on data sources and quantitative methods.
Artificial Intelligence (AI), Machine Learning, and Deep Learning were the focus of much attention and emphasis. Over 60 presentations (1/10 of the total number of presentations and roughly 1/3 of the marketing models presented) included some type of AI, machine learning, or deep learning. As new tools of marketing science, it seems that wherever machine learning or deep learning can be applied, they are being applied.
Text Analytics is still a hot topic. Textual data is rapidly multiplying in quantity, and organizations are seeking economical ways of extracting useful information from that data. With the ability to scrape textual data off the web, text analytics is growing in importance. For example, text analytics is being used to answer questions such as:
- What is the most effective way for a company to respond to bad online reviews?
- How do different ways of responding to bad reviews impact customer loyalty?
- How can one leverage social online networks to reach the most targeted customers?
All the work in text analytics is intended to address the issue of how to learn from textual data. Progress is evident, but human language is extremely complex. Text analytics is still in its infancy.
Choice modeling topics were still present, but were greatly overshadowed by all the “buzz” and excitement around AI, machine learning, and deep learning. Choice modeling presentations focused on new Bayesian methods that incorporate new theoretical behavioral models (for example, extending the random utility model) or on methods that use scanner data to explore correlation patterns across many product categories. Some methods incorporated either online clickstream data (to predict purchases), or consideration-set modeling (measured as clicking on products, whether or not a purchase is made).
While the term “big data” is no longer big news, large data sets have become the norm, so marketing scientists were focusing on how to make sense of huge data sets. For example, how does one analyze data with millions of observations and thousands of products? New methods of grouping products are being developed. One presenter reviewed how he adapted Latent Dirichlet Allocation (LDA), a method in Natural Language Processing that groups words into topics, in order to group product SKUs into product categories.
Mobile and digital data applications continue to grow; for example, researchers are using mobile location data to understand social networks, and they are also using personal driving data to optimize auto insurance offerings.
Survey methodology was hardly mentioned, even though the conference was built around the theme of marketing science. Survey data (despite its purity and power) was seldom discussed or incorporated into analyses.
While leading academics in the field of Marketing Science have embraced new data sources and modeling methods, they also maintained that marketing theory, developing testable hypotheses, and understanding behavior are still of central importance.
In the final analysis, machine learning, deep learning, text analytics, and surveys are simply tools within the marketing science toolkit. It will be exciting to see how the new data and new tools will contribute to marketing!
About the Author
John Colias (email@example.com) is Senior Vice President and Director of Advanced Analytics at Decision Analyst. He may be reached at 1-800-262-5974 or 1-817-640-6166.
Copyright © 2018 by Decision Analyst, Inc.
This posting may not be copied, published, or used in any way without written permission of Decision Analyst.