Getting the Most From Your Segmentation Provider

"If you're not thinking segmentation, then you're not thinking," Ted Levitt of the Harvard Business School supposedly quipped. While every marketer would readily agree with him, getting segmentation right can be as challenging as it is necessary - and getting it wrong can be downright disastrous: Working with segmentation schema that are irrelevant to your business is a waste of time and money; working with too few or too broadly defined segments means missed opportunities; and working with too many or too narrowly defined segments means stretching your marketing resources to their breaking point.

Since there are so many options and variables involved in segmenting your audience, it is best to rely on analysts, either external or internal, who understand your business and who understand how to match your needs with the myriad segmenting approaches. To sketch out a useable framework for getting the most valuable results from your segmentation provider, I consulted Ben Ben-Baruch, a Senior Business Intelligence Consultant represented by Aquent who got his first contract assignment with General Motors in 1997 and has been there ever since.

"Whatever segmentation provider you use and whatever methodologies and data they employ," Ben says, "the key is ensuring that you can use the segmentation to meet your business goals. Finding a provider that thinks in terms of your business, presents the data with an emphasis on its proper use, and makes it easy to keep the data fresh, is critical not only to the success of your segmentation process, but to the success of your marketing efforts in general."

Here are his 5 keys for getting the most from your segmentation provider:

1. Start with Defining the Goal of Segmentation

Segmentation schema can serve many different purposes and the business goals should define that purpose. Before approaching your segmentation provider, make sure you have a clear understanding of what it is you want to accomplish with segmentation.

For example, in one case, you might be using segmentation to produce a target list for a direct mail campaign aimed at people who have never bought certain products before. In another case, you might be segmenting your existing customer base for the purposes of ongoing communications.

In the first instance, making sure the right people are getting your message is the key. In the second instance, the important factor may be scheduling communications so that customers receive communications at a frequency which matches their purchase cycles. Either way, the particular method of segmentation, and the specific data necessary to perform it, should determine how you proceed. You need to be able to communicate that to your segmentation provider.

2. Find a Provider Invested in Understanding your Business

Just as you need to be able to communicate your business goals to your provider, it should be clear from the outset that your provider is focused on producing segmentation that meets those goals. One strong indicator is whether your provider asks you probing questions about your business and the needs that segmentation should address.

Another indicator that the provider has your business interests at heart is that they propose a segmentation solution, which may be fairly complex or relatively simple, based on the stated business needs. Be very wary of anyone who seems to be selling a proprietary solution right out of the gate. Good providers will be conversant with a range of segmenting methodologies and will not dogmatically apply the same method to every situation.

3. Don't Treat Segmentation as a "Deliverable"

Segmentation data are supposed to be useful and should be used. This often gets overlooked on both the client and provider side. The mission is often, "Develop Segmentation," and once the segmentation deck is completed, no one is concerned with understanding how to use it or if it is actually being used.

Your provider shouldn't be thinking of segmentation as a deliverable that just gets thrown over the wall. They need to walk you through it, explaining what it means and how it can be used. If they've done a simple categorization, then they should explain why that's all you need and how you should use it to meet your needs. If they have done clustering, they need to walk you through the clusters and describe how they behave across a variety of dimensions that are, again, important to your business -- and explain how to use this particular segmentation scheme. If they have provided you with multiple segmentation schema, they should walk you through each one and explain how they can be used singly and together.

As important as explaining how the segmentation should be used, it is equally important to explain how it should NOT be used. At this point, the client needs to decide how much consultation they require to successfully implement the segmentation for marketing.

4. Clarify the Ongoing Relationship

Segmentation is meant to inform your marketing activities not just at one point in time but over a period of time. Therefore, you need to understand it's longevity (or "shelf-life"). Chances are, it will need to be updated to remain valid and useful. This raises several questions. First, who is going to be responsible for updating and maintaining the data? Is this part of the ongoing relationship with the vendor? Does it get handed over to an internal team? Or do you rely on a flexible group of contractors who update the work as needed?

Secondly, did the provider use data that you can easily update? For example, Census data can be good, but it becomes less and less useful over time. Are you going to have to engage a company that works with interim Census reports to regularly update it (which can be expensive)? Or will you have to buy data from a company that provides interim updates between decennial censuses? Similarly, if the segmentation relies on survey data, are you going to have to re-survey, or can you use other data to refresh what you already have?

Ultimately, the best input data are data that can be easily refreshed. Still, regardless of the type of data or data source, the issue of ongoing maintenance and updating should be clarified with your segmentation provider up front.

5. Educate Yourself

While finding a trusted partner to perform segmentation analysis for you is invaluable, the more you understand about what they are doing the more you'll get out of it. If you know what different approaches can and can't do, you'll know better what to ask for and how to apply what you get.

For instance, there is a difference between actual data, like the exact amount paid by someone for a certain product at a certain time, and inferred data. Inferred data are "best estimates" usually based on models or upon attributing group characteristics to all members of a group. Income and wealth data are typical examples, because, in almost all cases, marketers do not know people's income or wealth. They rely on other data from which wealth and income can be inferred (or buy such data from data providers). Caution needs to be exercised whenever making marketing decisions based on inferred data because the inferences will usually be wrong for some portion of your audience.

By the same token, when using statistical clustering, which groups (or segments) people who look alike across several pre-defined criteria, you cannot assume that everyone in the cluster can be characterized by the description of the "average" member of the cluster. In fact, it is entirely possible that the "average" person in the cluster doesn't really exist at all and is only a statistical artifact. That "average" person is merely an amalgam of the central tendencies across all of the factors considered relevant for the clustering. And this difference between the actual, individual members, and the statistical generated "average" member, will arise, with some variation, whether the data you've used is actual, inferred, or a combination of the two.

Though this may seem a little esoteric, the main point is that business users need to know when they are acting on actual knowledge about customers and when they are acting on statistical knowledge of a group of customers who are not as individuals identical with the general characteristics of the group.

Image Courtesy of mrhayata.