Every retailer is interested in driving repeat purchases from one-time buyers. The common conception is that having that second touchpoint with a customer gives you the ability to forge a relationship and build loyalty.
In other words — using a relationship metaphor that we turn to a lot — every retailer thinks that the second date is at least as important as the first date.
If you’ve been following our series on segmentation best practices, then you already know why behavioral segmentation is more useful than demographic segmentation, you know how to define a “good” segment, and you know how to enrich your segmentation with marketing experiments that enhance your existing data, and common mistakes that’ll screw up your best efforts.
Where to Start
So as a retailer looking at a database filled with a messy mixture of one-time buyers, what’s the first thing you do?
Most retailers facing the slough of unsorted customers will turn to demographic segmenting with a handful of creative assets to distribute evenly among them. But where do you start with demographic sorting?
Do you separate them out by gender? Depending on what you’re selling, that might be a decent idea.
Do you separate them out by age? Again, it depends.
What about by location? Income level? Marital status?
Not so Fast!
Okay, sorry, let’s back up. Just about any email provider will let you upload a gender column that you can use to segment your list for when you need to send something to the male-identified and something to the female-identified.
The problem is that gender segmentation is very, very often irrelevant.
We say “often” because there are absolutely cases of certain beauty products or types of apparel that are only relevant to people identifying as a certain gender or people within a certain age range or income level.
But these cases are exceptions to the rule that people as customers aren’t most accurately described by demographic attributes.
The other option when facing the morass of unsorted one-time buyers is to start segmenting by the first product purchased.
Let’s say you’re a men’s apparel retailer with a diverse range of products. You’ve got long-sleeve dress shirts, casual short-sleeve button-downs, slacks, jeans, shorts, swim trunks, socks, belts, etc., etc.
As this fictitious men’s apparel retailer, you look at what your customers bought on their very first visit, and you can use that information to push them more of the same.
This simple starting point says, “If you bought this once, you’ve demonstrated that you’re interested in buying this type of product.”
Purchase behavior tells you a lot about a person. In fact, a purchase is the strongest indicator of interest and therefore the best data you have for segmentation.
Not only was this customer looking for this type of product — they came into your store; they found it; they put it in their cart; they gave you their credit card number; they received the product, and they used the product. That’s a lot of information considering that only a very small percentage of people who pass through actually convert and purchase.
More than clickstream behavior, more than email opens and clicks, more than social engagement, purchases are the most powerful predictor of future behavior. Of course, the other data is incredibly valuable. But customers tell you a lot when they make a purchase.
The con to this might already be popping up in the back of your mind.
We keep saying, “Push them more of the same.” So should you actually push more of the same? Should you try cross-selling? If so, what products should you try in your cross-selling strategy?
This is where things get interesting and we start learning segmentation by experimentation. You can start segmenting based on the predicted responses, and here’s an example:
You look at the first purchases of your one-time buyers and notice that the vast majority of them buy swim trunks and never come back.
The marketing team would really have nothing beyond intuition to support them in finding the best way to follow up with these customers. Should you send more swim trunk promotions? Maybe those customers are in the market for short-sleeve button-downs that would seem at home in a tiki bar? Or t-shirts for hanging out poolside?
Or, wait, now it’s October and they bought their swim trunks when we ran that promotion in June, so maybe they’re in the market for long-sleeve dress shirts?
The truth is that you probably don’t know enough just from that first purchase.
Systematically Going from the First Date to the Second
Consider this. Every single month, every single week, you’re gathering people whose first purchase is in the swim trunks category. There are a couple of things you can do with this proliferation of data.
The first thing you can do is run a series of controlled experiments.
Here’s a simple example.
Of the people who first bought swim trunks, you’ll split them into, say, three pockets (this is a reasonable size for your creative team to handle). These three pockets are your “push more of the same” group, your cross-sell #1, and your cross-sell #2. (You’ll also want to include a holdout control — a slice of this segment that receives no email — to see how they behave in the wild.)
So, for instance:
- Group 1: More swim trunks
- Group 2: Dress shirts
- Group 3: T-shirts
- Group 4: No email
Now you’ve got this specific segment of one-time buyers whose first (and only) purchases included swim trunks, and you’ve divided them into three experimental groups to test on plus the control group.
You want to run this test until the results reach statistical significance, but once you’ve collected and validated the data, you have some powerful customer intelligence in hand.
If it turns out that people who first bought swim trunks respond especially well to an email promoting dress shirts, then you’ve figured out the right way to reach out to a segment of your customers and you have a revenue-driving insight you can build upon.
Another thing you can do is look at your segment of high-CLV customers, say your top 10%. Within that, find the segment whose first purchase included swim trunks. Then see if there’s a pattern in their second purchases. This is an oversimplified example of how predictive insights work so that the kinds of insights gathered from the experiments mentioned above are culled from your existing first-party datasets.
With predictive modeling, things can get much more excited and complex, with the model able to correlate page views, email opens and clicks, in-store visits versus online purchases, and more to build truly sophisticated segments. But on a basic level, it’s doing the same kind of pattern recognition that you would in that three-segment experiment.
If you want to learn more about segmentation, check out our previous articles on the topic:
- How to Slice the Pie, Part 1: Demographic Versus Behavioral Segmentation
- How to Slice the Pie, Part 2: Design Data-Rich Customer Segments
- The 4 Layers of Effective Buyer Personas
- 3 Common Segmentation Mistakes (How NOT to Slice the Pie)
And if you want to learn more about how to convert one-time buyers to repeat customers, click here to download our beautiful omnibus, One and (Not) Done.