Why Segmentation Studies Fail And How to Do Them Right!
In order to learn about the state of customer segmentation research, GfK Custom Research North America talked to marketing and research directors in a variety of industries. We uncovered what was working, what was not, and heard a lot of disappointment and frustration with segmentation studies, including the lack of actionability in the results.
Results were not living up to expectations, and segmentation was getting a bad reputation.
In this paper, we share our learning from this effort and identify the root cause for this failure and the contributing elements. We also outline a three-phase process that avoids these pitfalls and results in segmentation that gets used. This process is the basis for our successful Strategic Segmentation product.
Summary
Segmentation is a valuable research tool when used properly. It can provide marketers with focus and direction to grow their business. Our clients tell us—and our own experience substantiates—that historically too much emphasis is placed on how to do the segmenting, and not enough on why segmentation is appropriate or how results will be used to make decisions. GfK believes that it is less important how you segment (many tools are available and work well for this) and that planning and deployment are key parts of successful segmentation work.
With a phased approach, the value of segmentation is greatly enhanced—yielding segmentation schemes that get used to make winning business decisions.
Segmentation Today
This we know for certain: segmentation studies—no matter how they are done, or what their purpose—always seem to:
- Include cute names to describe the segments, to give them life, to make them memorable,
- Involve large, usually six-figure budgets, and
- Generate huge reports in very thick binders.
Because businesses need more than cute names and binders that sit on the shelf, unused, these segmentation studies are often considered failures—failing to accomplish goals, failing to be embraced by the organization, and failing to spark meaningful changes in business strategy or tactics.
Root Cause: Results are Not Actionable
The major reason for this failure is that segmentation results are simply not actionable. Despite solid research design, quality data collection, and careful, labored definition of segments that seem to make sense, results seem to fall flat. They may be captured in binders, but they do not impact the organization or its marketing efforts.
In our research and subsequent hypothesis testing, three contributing causes for this lack of actionability emerged.
1. Inability to Locate Segment Members
Cute segment names don’t help locate and clearly identify segment members in the marketplace. Facts, hard data, including demographics, categorical data, behavioral data, and attitudes are required.
In the not-so-distant past, research considered its job the identification of segment membership within the study sample. Finding segments in the marketplace or among customers was considered outside of marketing research’s realm. Research reports would identify a certain number of segments, describe how they differ from one another, estimate the size of each segment, and not much more. We’ve all seen the cover memo slapped on a research report in a quick hand off to marketing, that in effects says: “Here are your segments.” Occasionally, some demographic elements that corresponded with certain groups (e.g. older loyals, etc.) would be identified. But this was the exception, not the rule.
More recently, attitudinal segmentation has gained popularity. GfK has always espoused this approach because it adds the missing piece of how the consumer thinks. While this is very helpful with developing targeted messages, some marketers still feel that attitudinal segmentation doesn’t help them take specific actions.
Today, savvy marketing researchers utilize database marketing tools and techniques and integrate research data with customer database information and other enhanced data to provide much more comprehensive identification of segment members and even predicting, on an individual consumer basis, likely segment membership.
With predictive modeling and scoring of customer databases, research has entered the world of closed-loop learning systems. Researchers, working with marketing, build test-and-learn loops and refine these models for even more precision in future utilization.
2. Emphasis on a Technical Solution
We also found that in many segmentation proposals and RFP’s there is far too much emphasis on the tools or techniques that are going to be used to actually develop the segments. Researchers often have strongly-held opinions about use of a factor/cluster approach or tree-based segment definitions. And lengthy discussions can result on whether CART or CHAID is really preferable.
The catch here is that the best technical tool really can’t be chosen until the data is examined. This is especially true when customer behavioral data and enhancement data are used for segment definition along with survey data. On numerous occasions, we set out intending to use one technique, and then, after examining all the data available, determine another technique is preferable and provides much better results.
Most sophisticated data analysts have numerous tools at their disposal. At last count, GfK had 28 ways to do cluster analysis, and although we have favorites, we start with understanding the data set to make sure the path chosen makes the most sense for this data.
3. Flawed Process
This focus on technical solutions points out that current research processes for segmentation are flawed. Instead of focusing on techniques, researchers need to concentrate on how the results are going to be used.
To do this, product and marketing managers must play an active part throughout the process. They have ideas about which segments currently exist, which of them may provide the most value, and which might in fact be costing the organization. They also know what they have done and could possibly do to change customer behavior, to improve profitability, and to develop new products to meet the needs of the various segments. It is important that these key stakeholders take a hard look at whether or not their organization is ready to treat different customers differently.
The best results happen when you begin with the end in mind. Shifting the emphasis from how the work is to be done to how the results are going to be used has tremendous impact, resulting in segmentation work that gets used.
“This is the best value for the money we have ever spent [on research]. It was worth every penny.”
—President, Big Box Retailer,
GfK’s Segmentation Work
Building in Success
Once we understand the causes of failure, the next step is developing a superior process to avoid the problems and, in fact, build in success. As a result of this investigation, GfK created a unique product, Strategic Segmentation, specifically designed to overcome barriers to successful segmentation.
GfK Custom Research North America’s Strategic Segmentation involves three essential phases:
Phase 1—Planning: Successful segmentation studies begin with well thought out plans to obtain the right customer information and, most importantly, for how this information will be used to make business decisions. This phase starts with a comprehensive review of existing customer information from all sources, including past research, customer databases, and commercially available market data. Then, planning meetings are held with key stakeholders to flush out business objectives, hypotheses about the current situation and critical business issues, and to start creating a list of possible action plans for deployment of the results.
Phase 2—Research and Analysis: Based on the information obtained in the planning phase, an appropriate research design is implemented. This design may first include qualitative research among consumers to assure that management hypotheses are in-sync with the “Voice of the Customer.” Stakeholders are briefed on the qualitative results and asked to buy into the proposed quantitative research design. They are also involved at key check points throughout data collection, data processing, and analysis. An iterative process with key client contacts is used to pick most appropriate segmentation scheme and names.
Phase 3—Deployment: After research results are summarized and presented, stakeholder work sessions are conducted to set direction for action steps based on the findings. This integration of learning to business decisions is a key part of GfK Strategic Segmentation because it provides rich understanding and context to the segmentation data. Also, researchers that have experience across industries can provide “out of the box” insights that can be quite valuable.
Tools to Aid Deployment
During the deployment phase, we typically use a number of tools or exercises to further enhance the value of the segmentation research and improve actionability of results.
Value Mapping helps determine which segments are most desirable and which are least desirable based on two dimensions:
- Growth potential or worth of the various market segments to the business
- Effort required on investment required to acquire new customers and keep and grow existing customers.
The value mapping process has both qualitative and quantitative components. At deployment work sessions, stakeholders, based on their business knowledge, decide which variables go into the creation of the growth and effort index components. They also agree on the relative weight of each variable. Using the segmentation results, a standardized index is created.
For instance, in a simple case, it may be decided that household income and current home ownership are the components of Growth Potential. Segments that have a richer mix of high income home owners will have higher Growth Potential scores than those with more low income renters.
Value mapping provides a strategic map of opportunities.
Modeling Attitudes can help predict specific attitudes consumers are likely to hold. This modeling is in addition to predictive modeling that identifies likely segment membership for non surveyed customers. For example, it could be quite helpful for financial institutions to know which of their customers have a high-tech orientation, and would be glad to move most of their servicing to the web vs. which would find such an offer an abomination. A model that links high-tech attitudes to database information would make this possible.
Marketing Data Marts and Data Cube are often created to provide the client with easy, user-friendly access to the segmentation data. These tools make the integrated data (from the survey, existing customer databases, and data enhancement) available to marketing management for ongoing use in planning and implementation. Queries can be made and reports delivered in a variety of formats, from web access to paper reports.
For a copy of this article, click on the link next to “Downloads” located at the top of this post.
