Ecommerce Customer Segmentation: How to Find the Customers That Matter Most
A segment is only useful if it changes what you do.
“Women aged 25–34” may be a perfectly valid demographic segment.
But if every person in that segment receives the same product, buys at the same frequency, spends differently and sits at a different point in the customer relationship, the segment may tell you very little about the next commercial decision.
For Revenue Recapture, I care more about questions like:
- How many times have they bought?
- What did they buy?
- When did they last buy?
- What do they normally buy next?
- How valuable have they been?
- Has their behaviour changed?
- Are they approaching a purchase window?
That is customer segmentation with commercial intent.
Segmentation is not personalization
They are related.
Segmentation groups customers who share something meaningful.
Personalization decides what experience those customers should receive.
McKinsey research found that 71% of consumers expect personalized interactions and 76% become frustrated when those expectations are not met. It also found personalized communications increased likelihood of repurchase among surveyed consumers. (mckinsey.com)
But personalization without useful segmentation can become superficial.
“Hi Sarah, we miss you.”
is technically personalized.
It may still be completely irrelevant.
Start with the business question
Do not begin:
What segments can our platform create?
Begin:
What decision are we trying to improve?
Examples:
Improve second purchase
Segment first-time buyers by product and purchase age.
Recover former VIPs
Segment historically high-value customers whose recency has deteriorated.
Increase margin
Separate full-price customers from promotion-dependent customers.
Improve replenishment
Segment by product, quantity and expected depletion.
The question determines the segment.
Purchase behaviour usually matters most
Shopify’s current customer-segmentation guidance recommends using customer data to create segments around repeat customers, lapsed customers, churn-risk buyers, high-frequency low-average-order-value customers, one-time purchasers and discount-dependent buyers. (shopify.com)
That is much closer to how an ecommerce operator thinks.
The segment describes behaviour.
The major segmentation dimensions
Recency
How long since purchase?
Frequency
How many purchases?
Monetary value
How much value?
Product
What did they buy?
Timing
When should they buy again?
Lifecycle
First-time, repeat, loyal, at risk, lapsed?
Channel
Where did they come from?
Discount behaviour
Do they purchase full price?
Returns
Do they generate strong gross sales but weak net value?
Each adds another dimension to the relationship.
Carhartt is a useful example
Bluecore describes how Carhartt uses different strategies for inactive customers based on factors including:
- what they previously purchased,
- colour affinity,
- discount affinity,
- browsing behaviour,
- how long ago they lapsed. (bluecore.com)
That is much richer than:
inactive = 180 days.
It creates meaningful reasons to show different customers different things.
Segmentation should reduce waste
Imagine:
100,000 inactive customers.
Segment by previous relationship.
Now:
10,000 repeat buyers
2,000 former VIPs
35,000 one-time purchasers
53,000 very old low-value customers
You probably do not want to allocate equal:
- discount,
- send volume,
- creative effort
to all four.
Segmentation is an allocation tool.
Segment by product journey
RetentionX allows segments to be compared by:
- customer value,
- product affinity,
- purchase behaviour,
- retention,
- churn. (retentionx.com)
That allows questions like:
Which customers start with Product A and become high value?
Now the segment can inform:
- acquisition,
- merchandising,
- retention.
Segment by cohort too
Shopify’s cohort analysis allows merchants to identify groups acquired in the same period and compare retention, sales, average order value and marketing-source behaviour. (help.shopify.com)
This lets you identify:
Customers acquired during this campaign behaved very differently.
That is more actionable than a store average.
Avoid creating 200 segments nobody uses
More segmentation is not automatically better.
Every segment creates:
- operational complexity,
- smaller sample sizes,
- more creative work.
The useful test is:
Would we actually make a different decision for this group?
If no:
combine it.
The best segmentation is dynamic
A customer moves.
New → active → loyal → at risk → lapsed.
Shopify’s recency, frequency and monetary value groups update customers as behaviour changes. (help.shopify.com)
That is important.
Customer segmentation should describe a relationship now, not permanently label a person.
The Revenue Recapture lens
For us, the most important segments often become:
- second-purchase overdue,
- recent repeat lapse,
- former VIP,
- replenishment overdue,
- cancelled subscriber,
- deep lapse.
Each tells a different commercial story.
And each should lead to a different recovery hypothesis.
The rule
A useful segment answers three questions:
Who are these customers?
What is different about their behaviour?
What will we do differently because of it?
If you cannot answer the third question, the segment may be interesting.
It is not yet useful.
Related reading
Recency, Frequency and Monetary Value Segmentation Explained
Behavioral Customer Segmentation