Database Special Report: Beyond the Black Box
Segmentation Is More Than a Tactic—It’s Also a Strategy.
Using statistical techniques to segment customers is an effective tactic, but how you market to these segments is a strategy. Rather than operating in a vacuum, statisticians and marketers can work together to not only predict behavior, but change it.
For many direct marketers, state-of-the-art segmentation means using the database to determine, with the greatest possible accuracy, which customers will respond to a given offer. They often point with pride to a statistical technique that selects the best 20,000 names from the database.
These modeling techniques typically are employed in a “black box” fashion, which is to say they usually come with no explanation, other than which deciles should be contacted and which should not. In some cases, the names simply are segmented into deciles, and it is left to the marketer to test and discover which can be contacted profitably.
Because the outcomes of these techniques are delivered to marketers without explanation, they rarely affect the offer. Statisticians assume their job is to select the best possible names from a given set, in a quantity usually pre-determined by marketing. The responsibility of creating offers, communication planning and budgeting becomes the responsibility of others.
This mind-set is starting to change.
Data-driven segmentation techniques that solely focus on predicting responses and sales miss the point of direct marketing entirely. Direct marketers are not just trying to predict customer behavior; they are trying to change customer behavior.
The 40/40/20 rule is an old direct marketing axiom. The rule states that 40 percent of a campaign’s success depends on the list, 40 percent depends on the offer, and 20 percent depends on the creative. For old-school direct marketers, the database drives the 40 percent of a campaign’s success that depends on the list. For cutting-edge marketers, the database also supports offer selection and even creative.
Each database segment should tell the marketer at least two things:
1. What makes the segment unique in such a way that the marketer can understand how customers with those characteristics are likely to respond to certain offers.
2. How likely the segment is to respond in such a way that the marketer can decide not only whether or not to contact that segment, but when and how often.
Modeling Is Tactical; Marketing Is Strategic
While there is no doubt that segmentation can be a very effective tactic, it can be more effective when employed as part of an overall strategy.
For example, consider the following two scenarios for a direct marketing organization that has a database with 100,000 names.
Scenario No. 1: The marketing manager has 20,000 mail pieces ready for a September campaign, so she asks the statistician to select the 20,000 best names. He does so using an advanced, neural-networking technique that has been shown to be highly predictive of which customers will respond.
The 20,000 pieces are identical, and each customer will get the same offer. The offer, which is $20 off any order over $100, has been the most effective in the past.
The mail pieces go out, and the campaign is a success. Responses are very close to what was predicted, and all segments within the 20,000 pieces perform above breakeven.
Scenario No. 2: The marketing manager is planning a September campaign. She sits down with the statistician and reviews past campaigns, current customer segments and past offer performance.
They realize September has been a strong month in the past, and they estimate they can reach 40,000 customers profitably. They also find that customers with lower average orders tend not to respond well to the best-performing overall offer, which is $20 off any order over $100, and customers with high average orders often place smaller orders than usual when given that offer.
They create 40,000 mail pieces that can be versioned for three different offers. Customers with low average orders get $10 off any order over $50; typical customers receive $20 off any order over $100; and customers with very large average orders get $100 off any order over $500.
The mail pieces go out, and the campaign is a success. All segments within the 40,000 pieces perform at or above breakeven. Furthermore, the low average order segment had a higher average order than usual, with most spending more than $50. The high average order segment also had a much higher average order than usual. Interestingly, response rates were better than expected in the low average order segment, and about the same in the high average order segment, despite the increase in average order.
Which scenario is a more advanced use of the database? Which is a more effective use of segmentation?
Segmentation by Frequency of Contact
For many direct marketers, business is seasonal. However, the seasonality of repeat purchases may be different than the seasonality of first-time buyers. Some catalogers contact customers year-round, even though they may only prospect in one season or even one month.
For example, a company selling soaps, lotions and candles by catalog mostly contacted customers and prospects around the holidays. It had been in business for about five years, and most customers were one-time buyers.
By reviewing its database, the cataloger discovered that new customers were as likely to buy again, regardless of the time of year, if they received a new catalog. The big difference was recency; customers who didn’t get a catalog for several months were not as likely to make a second purchase as the customers who happened to buy shortly before the next catalog was mailed.
As a result, the cataloger changed its communication strategy. Catalogs now are mailed 10 times a year, even though some mailings are quite small. For some mailings, only buyers from the last 90 days are chosen. For mailings during peak seasons, buyers who haven’t bought for several years may be contacted.
With no other changes, the number of buyers that made three or more purchases doubled in less than a year. Repeat purchases soared, and lifetime value jumped dramatically.
The segmentation analysis did not only change the tactics of who gets mailed when; it changed the cataloger’s strategy. Before, it thought of itself as a holiday/seasonal business, driven mainly by gift giving. Now, it knows its best customers buy again—often for themselves. If the cataloger doesn’t contact its customers often enough, it will lose them. This new strategy is changing the tone of the offers and influencing a change in the creative message.
Segment by Understanding
Marketing is communication, and proper communication requires understanding. Someone once said, “Never tell a story without making a point, and never make a point without telling a story.”
The point of effective segmentation modeling is that you can select, or de-select, names very efficiently. But you have to know the story to develop the right creative and build the best offers. As marketers, you should demand that regardless of how effectively a segmentation technique makes a point, it must also tell a story.
The story is the explanation, and you must understand it to fully implement the lessons it has to offer.
Alan Weber is president of Marketing Analytics Group, a direct marketing consultancy based in Cleveland, Mo. He can be reached at (816) 618-3338, or by e-mail at: firstname.lastname@example.org.