Applying Machine Learning in Market Research

Adaptive Algorithm imageWebinar Thursday 20 June, 2019

Machine learning and automation are more than just hype in market research and promise efficiency gains. Learn how Factworks and Dynata have investigated how a self-learning algorithm can achieve results more efficiently for naming and concept tests. Based on an AB naming test study in the UK and Germany they will show that the intelligent algorithm can determine name winners faster and more reliably than conventional monadic designs.

Session moderator, Ray Poynter

The Presenters

Nadja BöhmeNadja Böhme, CEO at Factworks
Nadja is CEO at Factworks, a full-service research agency based in Berlin and San Mateo, California, USA. During the past ten years, Nadja has enabled clients from tech and finance to make confident business decisions by leveraging custom designs and advanced quantitative analytics. In her former role as head of marketing she actively drove the development of innovative research approaches at Factworks.


Graham WilliamsGraham Williams, Research Director at Dynata
Graham has over 20 years of experience in market research and in that time he has helped a variety of clients ranging from Time Inc., Global Radio, Coca-Cola, ClearChannel, Universal Music and American Express. He also is employed as an expert witness for law firms in IP disputes. Graham is proud of winning the 2008 MRS Applications of Research Award for a research project commissioned by Channel Four. Graham Joined Dynata in November 2018 from GfK.


Ray PoynterRay Poynter, Founder, NewMR
Ray is the founder of, a long-standing member of MRS and ESOMAR, and a member of the ESOMAR Council. Ray is the author of The Handbook of Mobile Market Research, The Handbook of Online and Social Media Research and the #IPASOCIALWORKS Guide to Measuring Not Counting, editor of the ESOMAR book Answers to Contemporary Market Research Questions.


Factworks and Dynata logos


Finding the best name quickly – an intelligent algorithm for naming and concept tests
The Market Research industry is ever changing, and recently the focus has been on how our industry can leverage Automated Research. Automation, when done right, signals a brighter tomorrow: smarter research, completed more quickly and cost-effectively.

For concept and name testing, monadic designs are still the gold standard as they confront respondents with only one concept – making it closest to what they would encounter in reality. However, monadic designs are usually very time and cost consuming especially when many options are being tested.

Dynata and Factworks asked themselves if there is a more efficient way for concept and name tests, leveraging automation and machine learning to save money and time while building on the advantages of monadic designs. They developed an intelligent algorithm which detects the most promising concepts during field and thus optimizes sample distribution in real-time, without need for manual adjustments.

In a joint research in the UK and Germany, Factworks and Dynata tested the newly developed algorithm in an AB naming study that compared the conventional monadic and the adaptive approach.

The results showed: Up to 20% gain in efficiency. Using the algorithm, winners can be detected faster and more reliably. If you’re dealing with a hard-to-reach target group (e.g., B2B sample) or want to test a large number of concepts or names, the adaptive algorithm is a promising new way to go.