Behavioral segmentation: choose the action behind the customer group

Behavioral segmentation groups customers by actions such as how often they buy, how recently they ordered or how they use a product. For a coffee subscription, a long gap between orders identifies accounts to investigate before choosing a reminder, discount or delivery change.

In this illustrative coffee example, your team wants to encourage the next order. A customer who has not bought for 90 days might still have coffee, need a different delivery schedule or have disliked the roast. Those situations call for different actions.

Different customer situations call for different actions

Before building a complicated set of segments, identify the decisions your team can actually make. For the coffee subscription, you could change the quantity, timing, product or price of the next offer.

A discount is useful when price is the obstacle. It does little for someone who already has enough coffee. That is why an order pattern should start an investigation, rather than stand in for an explanation.

Define the groups so every account has one place

Take a snapshot of the customer accounts eligible for the campaign and their completed orders on a stated date. Set the snapshot time, timezone and included date bounds before assigning groups. Use completed purchases consistently: exclude cancellations, decide how refunds count and check whether guest orders can be linked to the same customer.

For this example, assign each account using the rules below, in order. Once an account matches a rule, stop; do not put it in another group.

1. Unknown: you cannot reliably connect the account with its purchase history. Check the records before treating the customer as new or lapsed.
2. New: the history is reliable and the first-ever completed order was within the past 30 days. Focus your questions on the first delivery and what the customer expects next.
3. Repeat: among the remaining accounts, there were at least three completed orders in the past 90 days. Find out whether the customer values a predictable delivery plan, flexibility or savings.
4. Lapsed: among the remaining accounts, there was a completed purchase, but none in the past 90 days. Ask what changed after the last order.
5. Other: the history is reliable, but the account meets none of those rules. Keep it here while you decide whether a different action is useful.

For example, a customer whose first order was 20 days ago and who has already placed three orders is New. New comes before Repeat, so the account is counted once. A customer with one completed order 45 days ago is Other: they are beyond the New window, below the Repeat threshold and have ordered too recently to be Lapsed.

The order is Unknown → New → Repeat → Lapsed → Other. Unknown comes first because missing purchase records can make a regular buyer look new or lapsed.

Use the buying cycle to choose your time windows. Ninety days may be a long absence for a monthly coffee order and perfectly normal for a product bought twice a year. The 30-day and 90-day cutoffs here are example rules to adapt to your category.

The last order is the starting episode

Once you have identified the group, ask about the latest episode before presenting an offer:

• “Think about your last order. What happened afterward?”
• “What, if anything, would make another order useful?”

For a lapsed coffee customer, follow up on how much coffee remains, when another delivery would become useful and what they have chosen instead. Leave room for a reason your team has not considered. Someone may have moved, changed their routine or stopped drinking coffee.

Ask these questions before showing a discount or a new bundle. Otherwise, you risk getting reactions to your proposed solution without understanding the problem it needs to solve. Our segmentation guide explains how to connect customer needs with the action each group should receive.

The explanation determines the offer

Suppose the customer still has plenty of coffee. Compare a smaller pack or a later refill. If their routine has changed, compare scheduling flexibility. If the roast disappointed them, investigate the experience before asking them to buy more of it.

Keep the commercial terms visible when you compare offers. A smaller pack with a different unit price and a delayed delivery with the same quantity solve different problems. Your study should tell the team which tradeoff matters to the customer and why.

The purchase-history group remains useful: it tells you whom to contact and what episode to ask about. The explanation tells you which offer to develop. You can learn that several lapsed customers have different needs without inventing a new segment for every individual answer.

Start with the few distinctions that change the action. Each extra group creates more campaign versions, rules and operational work. Compare the targeted approach with a simpler shared offer before committing to that complexity.

Record quality determines the assignment

Before activation, check a few accounts near each cutoff and a few with missing or duplicated records. These checks answer practical questions:

• Does a customer with three orders since their first purchase 20 days ago appear only in New?
• Does a known customer with one order 45 days ago remain in Other?
• Does a customer with no completed order in the past 90 days enter Lapsed only after their earlier purchase history has been confirmed?
• Do unmatched guest orders or duplicated accounts create a false lapse?

Count Unknown and Other alongside the named customer groups. Unknown tells you how much history needs repair. Other tells you how much of the eligible customer base sits outside your chosen rules. Neither should disappear from the campaign report.

Then check whether you can reach the selected customers and deliver the proposed offer. A group is only worth targeting if your systems can find it and your team can act on it.

Judge the action by what happens next

Save the original group assignment, the date and the eligible customer list before sending the offer. Keep those assignments fixed during the evaluation, even if customers later move into another group. Evaluate completed orders in a new window after that date. Do not count the past orders that qualified someone for Repeat as evidence that the new campaign worked.

For the coffee subscription, your report should show which offer each group received, how many eligible customers were included and how many completed another order during the follow-up period. Keep the purchase definition and follow-up period consistent across the groups you compare.

To assess the targeted offer’s added value, randomly assign eligible accounts within each saved group to the targeted offer or a declared shared-offer/control condition. Compare completed orders over the same follow-up window. Later orders alone do not establish what the offer caused.

Use the result to decide whether the extra segmentation is worth maintaining. If a simpler offer serves customers just as well, keep the simpler system. If scheduling flexibility solves a clear problem for one group, that difference gives you a reason to retain it.

From a lapsed coffee account to an offer comparison

A Vensia study can compare modeled responses to a smaller pack, a later refill and scheduling flexibility, then explore the reasons across your declared coffee-buyer groups. Supply complete prices, quantities and delivery terms with the group rules and category buying cycle.

Keep simulated responses distinct from transaction records. Your customers' later completed orders remain the reference for actual repeat purchasing.

Use the coffee offer brief with Vensia to decide which stock, timing, roast or price explanations the study must distinguish. For background on usage and buying occasion, see OpenStax’s segmentation chapter.