Targeting based on behavioral profiles may look like this:
- High-Value Customers, based on dollar amount (even for one purchase). The top 10% to 15% of the customers in terms of spending level, for example.
- Study Frequent Shoppers and/or Visitors (who are not necessarily the same as high-value customers), and mimic their profile among one-time buyers.
- New-to-the-Brand Customers. Don’t treat all one-time buyers the same way, as they may not have had enough time to show their full potential yet.
- Multichannel Customers. Identify differences in profiles between online and offline customers, and also follow after profiles of multichannel users.
- Customers Who Bought Multiple Product Types or Multiple Brands (they are the best targets for cross-sell/up-sell models). Model after them and sort out high scores among one-time buyers.
- Profile Customers With a Specific Product Purchase (e.g., strategic “entry” product buyers, targeting by distinct product lines, etc.), and treat look-alikes of them with different offers.
- Separate Full-Price Purchasers bargain seekers (using % of transactions in various discount banding, including 0% discount), and create a probability model on the “cheapness” spectrum (i.e., “likely to be a bargain-seeker” moving toward “likely to be full-price purchaser”).
- Seasonal Buyers (peak vs. off-season). Do not ignore offseason shoppers — whenever that is for you — as they may open up new opportunities. In fact, omitting Christmas-only buyers from analytics may reveal interesting patterns.
- And of course, analyze one-time buyers in depth, to define profiles of “non-target.”
I’m sure many marketers have been playing with this type of target for years. Now let’s add the timeline view:
- Actives — Recent buyers or what some call “hot line names.”
- In-Market — (e.g., 60 days since the last purchase, depending on average number of days between transactions for the base)
- Faders — Need to be gently nudged.
- At-Risk — Bribe them if you have to, before they go dark.
- Inactive — Dormant customers. Treat them as if they are prospects, but with some “dated” transaction data (which is still better than nothing).
Marketers who still “batch and blast” indiscriminately should really look at life stages of the customers more carefully. Are you bombarding the recent buyers with two emails a day, and not doing enough to wake up the dormant customers? Even if one is not building some optimization model for every stage of the game, just by treating these basic segments (based on moving timeline) differently will improve the situation.
Yes, you may have to bribe old customers to come back. But the important takeaway is never to treat everyone the same way. And if you use this two-dimensional approach, you will be able to create enough varieties in targeting and messaging without investing heavily on deeper analytics.
Now, going back to that one-time buyer problem, marketers should not treat all of them the same way. Some may be just new, and haven’t even had a chance to come back to you yet. Some may be aging really rapidly, and you must act on them now. Some may have gone dormant, but in that one purchase, she may have spent a lot of money. And depending on what they have bought through what channel, you need to treat them differently. If they took a fat discount and never came back, well, you may have to curb the urge to send more discount offers and just let them go.
When there is a big fat problem in front of us, it always helps to break it down into multiple categories of problem areas. You will win some and lose some in the beginning. But if you really apply that time-tested “closed-loop marketing” principle, things will get better, one metric at a time.
And the one that you really have to watch out for? “Number of transactions per customer,” as it is proven that multi-buyers bring the most revenue to retailers. That one metric that is not up for debate.