Try a MaxDiff Exercise Below
Try a MaxDiff Exercise Below
Needs-based customer segmentation is one of the first steps in a disciplined product development. But what is needs-based segmentation? Why is it a smart strategy to start with it? And most importantly, how to go about doing it?
Products and services are often a combination of features, each feature may be valued differently by customers. It is especially true for Software-as-a-Service products in which often hundreds of capabilities are bundled together and offered as “the product.”
Often, the product is then sold at “the price,” which is usually the reflection of the value it creates. The problem is that customers value the differentiating features very differently, due to their preferences, individual needs and ability to pay.
Imagine that there are five customer segments, each with unique needs and economic value.
Source: The Strategy and Tactics of Pricing by Th. Nagle and G. Müller
In the illustrative image above, the company set its (one) price based on which segment’s maximum willingness-to-pay maximizes contribution margin. In this example, segment C is the most profitable, and so price will be set at $10.
The single-price strategy leaves money on the table. Segments A and B are willing to pay quite a bit more, and by paying only $10, they’re enjoying a ‘consumer surplus.’ Another problem is that a price-point of $10 misses out on nearly half of the market. Segments D and E aren’t interested in the product at $10.
Source: The Strategy and Tactics of Pricing by Th. Nagle and G. Müller
It is significantly more profitable to create a structure of prices that aligns with the differences in economic value and customer needs. Notice on the images, above, that by providing solutions to each segment, the company – in this example – would nearly double its profit contribution.
Financial success then depends on the company’s ability to identify distinct customer needs and then create solutions for those needs. Ideally, each solution will be set at a price that near-maximizes the segment’s economic value.
Finding customer groups that are unique based on distinct needs is certainly a challenging task. The researcher may be familiar with various statistical techniques, yet applying the correct inputs to these methodologies is just as critical as picking the most appropriate clustering method.
When searching for unique needs, the most common approaches involve survey-based approaches. Survey-based clustering can quickly become over-complicated and complex. The more survey questions (dimensions) are included in the segmentation the less clear and often more confusing the segments become.
One of the increasingly popular segmentation methods uses a technique called MaxDiff.
A MaxDiff is a survey-based method in which survey takers pick their favorite and least favorite item from a subset of a list of items.
Survey-takers, during the MaxDiff exercise, typically go through eight to ten different subsets of items and make repeated choices of best/worst or most preferred/least preferred.
The result of a MaxDiff exercise is a prioritization and scoring of all the items. However, the item scores are likely different by each survey-taker. The MaxDiff choices can be analyzed using a technique with a fancy name called Latent Class logistic regression, which looks for patterns in the choices.
The image, above, shows a MaxDiff survey question. In this example, the survey-taker is asked to evaluate a CRM system based on feature desirability. The four items, shown here, are a subset of 20 product features. Using the Latent Class logistic regression technique, the researcher is able to find unique clusters of customers whose members have similar feature-needs within the segment but have different needs across the groups.
It’s hard to separate respondents well into needs-based segments if respondents are giving a lot of top-box (straightlining) answers that lack discrimination. Scale use bias further complicates matters and makes it hard to segment people properly whose different backgrounds / education / cultures lead them to use rating scales differently.
MaxDiff escapes the scale use bias problem as MaxDiff questions don’t involve a rating scale. Respondents just pick best and worst items from a list. MaxDiff questions also lead to much greater signal relative to noise than traditional rating scales. If you’ve struggled with these issues before when conducting segmentation, you’ll be very pleased with how well MaxDiff works for needs-based segmentation.
If MaxDiff analysis is used for segmentation, it is important that the items in the exercises are part of the same concept.
Finding the distinct segments and their unique preferences and needs is only step one in the process. These findings will allow product managers and marketers to structure and price the offering according to the segments’ needs. The research should include exercises to measure customers’ willingness to pay for the various features to assist in setting up the pricing structure for the unique offerings.
Hopefully the segmentation will yield valuable insights on other aspects around the customer segments that will allow marketers to target the customers. For example, the size of the segment, the business size, location, etc… can prove to be invaluable for a marketer.
Using MaxDiff with a Latent Class logistic regression segmentation technique has been increasingly popular in the market research community. In fact, we would like to invite you to experience needs-based segmentation (using MaxDiff and Latent Class) in an exercise.
We thought you may want to learn a bit about yourself: what you see as some of the biggest challenges facing the world are? And how many people think like you do? In a way, we can call it ‘concern-based segmentation,’ instead of needs-based segmentation.
On the top right of this page, you can enter a MaxDiff exercise on “Trends in the world that concern you.” We invite you to participate - it should not take longer than 3-5 minutes.
Once everyone has completed the MaxDiff, we will run a Latent Class logistic regression analysis on the choices of all the respondents and email you the segments with their descriptions – and which segment you belong in. This is only for some fun, of course – and no information will be shared in any publication. I hope you’ll participate! Enjoy!