By Kelly Kwon, Data Analyst, Polaris Marketing Research

Segmentation (Cluster Analysis)

Segmentation analysis is a statistical technique that researchers use to classify customers who have similar needs or characteristics and who are likely to exhibit similar purchasing or usage behavior. Advance statistical analysis software allows marketing researchers to compare respondents’ answers to a questionnaire, and put them into groups, where members share properties in common. These “clusters” contain actual or potential customers who can be expected to respond in a similar way to a product or a service offering. With the information gained from cluster analysis, companies can customize sales tactics and modify product offering to better address the underlying value propositions that will most likely appeal to a given customer set.

At Polaris, we often use two advance analysis techniques to segment our client’s customers – Maximum Difference Analysis (usually referred to as MaxDiff) and Conjoint. Although MaxDiff shares common techniques with conjoint analysis, it’s easier to use (for respondent, researchers, and end client) and applicable to a broader variety of research situations. However, MaxDiff is not a substitute for a conjoint, as conjoint offers unique benefits for studying products or services made up of complex features added together.

MaxDiff

The goal of MaxDiff analysis is to assess which attributes are the most important for respondents. The advantage of Max-Diff analysis is the forced nature of the question. For example, we ask respondents to pick one attribute that’s most important and one that’s least important from a group of four attributes.

Due to variety of different possible combinations, this question will be repeated nine times with random four choices from a list of 12. With advance statistical software, such as Sawtooth, we can categorize an individual based on what is important to them. With the above MaxDiff question for example, we can categorize a respondent in one of two segments – those who are “knowledge” focused and those who are “convenience” focused.

Also, by labeling the database as segment 1 or 2, we can easily run crosstabulation and explore how differently or similarly the segments answered other questions in the survey.

Although easy to answer, MaxDiff questions can be very tedious for respondents to complete, resulting in early break-off and incompletes. Thus, it is important to design the questionnaire in a way that helps respondents understand what to expect. Additionally, the fewer the attributes, the easier it is for respondents to complete the series of tasks.

Conjoint

Conjoint evaluates how customers make tradeoffs between various product features and outputs an assessment of the relative importance of attributes on a product set. It is based on the premise that the relative importance of product attributes are more accurately measured when evaluated together rather than individually. The procedure uses manageable, random subsets of all the possible combinations of attributes being tested to determine the relative importance of each. Please see example below:

In this example, the question wording was “Which option would you choose to buy?” The goal of this analysis was to test the attractiveness of various features, including brand, cost, design, and warranty.

With analytical software, conjoint analysis estimated the value consumers placed on each attribute, determining the degree to which each affected overall preference. From the example above, in addition to price, warranty and brand was important system feature to consumers, and concept design was the least important consideration.

At Polaris, we create simulation models, or “simulators” with conjoint data that allow testing of various levels of features to see how changes in each feature impacts choice. In the simulator, you can change the warranty year in one concept and change the price in other concept and see what the demand is for product 1 vs. product 2.

This has been a brief overview of segmentation, MaxDiff, and conjoint analysis. If you’d like more detailed understanding of the analysis in this newsletter, please follow the links below:

Segmentation/Cluster Analysis

MaxDiff

Conjoint

__________________________________

Kelly Kwon is a Data Analyst in the analytics department at Polaris Marketing Research Inc., where she handles a variety of data manipulation tasks involved in survey research. She has a bachelor's degree from the University of California, Berkeley in Molecular and Cell Biology and in Marketing from Georgia State University.