Years ago, a writer by the name of Richard Carlson made a lot of money by telling people “Don’t Sweat The Small Stuff.” That advice, while good for managing the anxiety of day-to-day life, is bad advice for marketers looking to manage day-to-day interactions with their customers. The “small stuff” may, in fact, be the key to unlocking what your customers want.
Most marketers today equate personalization with segmentation. They believe that, if they can divide their customers into the right customer segments, they can deliver personalized experiences at scale. But personalized in this case isn’t individualized. To illustrate this point, I’ll use the example of a retailer that wants to sell a new sweater using a segment-based campaign.
A North American retailer has created a new line of cashmere sweaters for the Fall season. In order to promote the new line, their marketing team looks to create a segment of customers who are most likely to buy a cashmere sweater from them. They start by looking at who has bought a similar item in the last twelve to eighteen months. From there, they identify the shared characteristics of those consumers—90 percent female, 75 percent with incomes over $100,000, 70 percent above the age of 40—to create the ideal segment. Next, they go through their database to find lookalike consumers—i.e., those with the same three characteristics—and combine the two pools of customers to create their target segment for the campaign.
On the surface, the marketer is doing everything right: they’ve identified the right people, those most likely to buy, and have the foundation to create a buyer persona for their campaign messaging. The trouble is, they’re using the wrong math. The idea that “1 + 1 + 1 = 3” and that “3” equals their ‘perfect’ targeted customer is inherently flawed. What happens if one of those 1’s is missing? Maybe the database field for “age” is empty, or the income is unknown. These customers are discarded from the group because their attributes don’t “add up,” and so the opportunity is lost. But what’s really lost is the opportunity to personalize the campaign beyond a few shared attributes.
Recommendation vs. Segmentation
By looking at a wider group of attributes and data points known as content affinities, marketers can go beyond broad segmentation to target the individual. The companies that do personalization best, like Amazon and Netflix, don’t even do segmentation anymore. They do recommendations. They can recommend products or movies specifically for you because they not only track what you’ve done in the past, but they’ve been building up insight into why you do it. They know whether you bought a cashmere sweater because of the color, the brand or because you love the feel of cashmere. They know if you watched Deadpool 2 because you like superhero movies or dark comedies. When Amazon or Netflix presents you with an offer, it’s not because you fit a certain segmented stereotype, but because it’s exactly the type of thing that usually piques your interest.
The challenge with sweating the small stuff, of course, is that there are a lot of little things to consider when you have thousands (or even millions!) of customers. Moving from a solely segmentation-based model to a recommendation-based model involves more complex decisioning techniques that include segmentation, content recommendations, defining the next best experience, lookalike modeling and identifying the appropriate triggers. Marketers don’t settle for segmentation because they believe it’s the best way to do personalized marketing, but rather because they believe it’s the best they can do under the circumstances. Through technologies like customer data platforms and machine learning, however, marketers can find deeper connections in their data, so that decisions are based, not on several shared customer attributes, but on several hundred unique content affinities, and they can do it at scale in real-time. That’s important because your customers aren’t the sum of a few shared attributes or a drop in a segmented bucket of similar customers; they’re individuals with very personal ideas of what the next best experience with your brand looks like.
As daunting as it seems, delivering one-to-one marketing at scale isn’t rocket science. It’s data science. Fortunately, you don’t need data scientists to get those insights, as I explain in my next blog, The Mathematics of Marketing, Part Two: Machine Learning, Modeling, and Moneyball.
