Database : Put Yourself in the Driver's Seat
Expand markets and optimize costs with a new segmentation and modeling strategy
January 2009 By Rick WitsellTo meet the challenge, organizations are developing both descriptive and predictive segments to target marketing messages. The combination gives them the ability to see which populations, offers and channels are likely to be most profitable—as well as the capability to “dial in” the desired level of ROI.
Marketers have the ability to target the ROI they need from each channel. The key is to develop broad-based, descriptive segmentations (i.e., overlay their housefiles with additional data) and build predictive models to zero in on the desired customer behaviors.
Looking Into—and Beyond—the Housefile
Building a segmentation strategy that gives you the flexibility to operate profitably in a multichannel world starts with reviewing your housefile.
Simple segmentations are readily apparent: actives and expires; payers and nonpayers; product interests and demographics. These descriptions are great for doing things like selecting list sources and segments for mailings. They can help you target some likely media sources, like package inserts and magazine ads.
Also, descriptive segments are crucial for customizing creative. For instance, if you are fielding an offer for political biographies, you use your descriptive segmentation to change the copy and featured book for the Republicans versus the Democrats you’re targeting.
But many marketers are learning that with the help of a good statistician and access to the right third-party data sources, they can make their segmentations work smarter.
Successful descriptive segmentations move beyond intuitive target groups by using statistical techniques to uncover correlations and segments that may not be readily apparent. Overlaying third-party data can significantly deepen and broaden the resulting segments. Consider four rules for developing descriptive marketing segmentations.
2. Business rules. Only consider segmenting based on variables that are appropriate and actionable for the entire available universe.
3. Actionable. A good segmentation provides basic decisioning for crafting custom offers and customer relationship management.
4. Statistical data discovery methods. Develop segmentation rules that identify data-driven variations in your populations.
Analyzing your customers through the prism of third-party data sources helps you see far deeper into your customers and gives you a more intensive understanding of their buying interests and behaviors. Consider some of the outside data attributes you can use to rethink your customer segments:
- RFM: Recency, frequency, monetary value.
- Product Type: Continuity, one-shot, online subscriber, magazine, music, video, sweeps.
- Affinity Type: History, health, gardening, sports, cooking, home, crafts, kids.
- Acquisition Channel: Mail, telemarketing, e-mail, Web store, package insert.
- Performance: Fast payer, slow payer, returner, write-off, repeat buyer, renewer.
The data you have on your customers, great as it is, just isn’t a complete picture. Understanding what, when, where and how consumers do business across a large group of marketers gives you insight you can’t get from your housefile. New sources of data can help you create broad-based, descriptive segments that provide opportunity for product development, channel selection and increased lifetime value.
Broad-based segmentations can deliver surprising findings. Suppose you discover that many of your customers have a strong interest in something you don’t sell, like gardening products. You might be able to test into those lists. Or if your customers have a strong affinity for video, you might consider offering a DVD as a premium. If you discover that a large percentage of your customers are responding to e-mail offers, you can head in that direction.
However, there is a danger in overdoing rules-based segmentations. As you develop insight into which groups prefer specific offers via specific channels, you run the risk of making your segments so small that they are no longer truly predictive.
Segments can only show you that, on average, a certain group is likely to perform in a certain way. Just because you’ve identified three one-eyed Albanian gypsies who bought your control offer through a search term, it doesn’t mean the fourth one will. The group is too small to be statistically predictive.
A new set of broad-based segmentations can expand your marketing reach and help you mine new sources of names and new channels. The key is to make sure your segments are statistically significant, big enough to be actionable, and likely to remain stable and consistent while you execute your expanded marketing plans. Chances are you’ll begin to see a pickup in response as you reach out to new prospects in new ways.
Dialing in Customer Performance
As valuable as descriptive groups are, you can’t segment your way to maximum profitability. But you can refine the performance of your descriptive segments with predictive models.
Behavioral models can help you refine campaigns and follow-up marketing by allowing you to predict the likelihood of profit-driving behaviors such as payment, multiple orders and renewal. Model-based, predictive segmentations can be applied in the following ways:
- Refine descriptive segments to increase campaign impact.
- Manage campaign costs by eliminating lower-performing groups.
- Expand market reach by targeting the top-performing groups.
- Segment incoming orders for fulfillment and upsell by profiling for profitable behaviors.
As an example, imagine that your descriptive Segment Y is statistically more likely to respond to a given offer, but the payment rate for the group is only 50 percent. Let’s say your business requires a 60 percent payment rate on new orders to be profitable. A net payment model can help you identify the individuals (or households) who are more likely to respond and pay within the segment. By subsegmenting the group, you can see that targeting only the top 80 percent of the file is likely to achieve your 60 percent cutoff (as shown in the gains chart).
Building a solid model requires that you have results from previous campaigns to that segment and that the modeler has access to a large source of relevant data. The resulting model can dramatically expand the reach of your marketing campaigns and significantly increase ROI by helping you identify and target only the most profitable groups.
A good model also can help you tackle the problem of shrinking mailing list universes. Chances are you’ve tested many lists that were on target for your segment but yielded results that were marginal or slightly under your net profitability criteria. A well-targeted model based on solid data lets you retest those lists and only promote to the groups most likely to pay.
The result? Your overall mail costs go down because you don’t promote to the entire file, but your net response stays more or less the same and payment goes up. You deliver higher ROI. And remember, if your broker can negotiate a net rental arrangement with the list manager, you won’t be paying for names that don’t meet your modeling criteria.
Models also can help you expand list populations by letting you mine segments that you haven’t been able to work with before. If you’ve been continuing on a 60-day hotline select, the model should let you expand to a 90- or 120-day select, providing larger universes.
Finally, your models also can work for you in untargeted media like Web-based lead generation programs, package inserts or other channels. Models can be applied to responders—or in real time on your e-commerce site—to help you make decisions about fulfillment, upsell targeting and future marketing contact.
Take Back the Wheel
If you sometimes feel like you’re suffering from “Bosley’s Lament,” rethinking your segmentation and targeting strategy is a powerful way to realign your marketing spend.
Redeveloping your descriptive segments—and adding third-party data if it’s available—can help you identify new universes and new offers. Building predictive models to target profitable customer behaviors lets you aggressively seek out the populations you need to grow profitably.
In this multichannel, multi-opportunity world, it’s a one-two punch that can put you back in control.
Rick Witsell is vice president of marketing at Alliant. Despite being a co-founder of the company with more than 20 years of experience in direct marketing, he recognizes that this article could not have been written without the patience and insight of many people on the Alliant team such as Serge Bernard, Russell Greenberg, Dan Parzych and Bart Surrick. Witsell can be reached at rwitsell@alliantdata.com.
The Secrets of Direct Marketing Arithmetic
