Customer segmentation model

A customer segmentation model is the rule set that divides your customer base into groups that behave alike. It names the variables, the number of groups, and the cut points. It describes groups and the differences between them, never one individual buyer.

That last line is the boundary. A segmentation model sorts the whole base at once; account scoring watches one customer at a time and moves as that customer moves.

It is also not a data platform plan — the platform decides which fields exist and whether you may use them, while the model decides what you conclude from the fields you already hold.

Key takeaways

• A segmentation model is about groups, their variables and their cut points, never one account.
• Pick variables that predict behaviour you can act on, not variables that are simply easy to collect.
• Four to eight segments is the usual working range, because sales teams cannot remember twenty.
• Rebuild on a fixed cadence, since the base shifts underneath a model that is left alone.

How it works

You choose a basis, pick the variables, run the cut, then test whether the groups actually behave differently. A good model produces segments that are big enough to serve, distinct enough to treat differently, and stable for at least a year.

Consumer teams usually start with audience segmentation, because media buying needs groups that map onto what a platform can actually target. It is the cheapest basis to build and the weakest at predicting churn.

Size matters as much as logic. A segment too small to staff is a research finding, not an operating unit — four to eight groups is the range most commercial teams can actually run.

Variables that are easy to collect are not always variables that predict. Postcode is cheap and weak. Months since the last purchase is cheap and strong, which is why it survives in almost every model.

Behavioral segmentation earns its keep faster. What people did last quarter predicts next quarter better than who they are, and the data already sits in your transaction records.

A customer value segment adds the commercial lens. Margin and cost to serve decide which groups deserve a named owner and which should be served through self service channels.

In business markets the model usually feeds coverage design. Account based selling is what the top segment gets, and the model’s real job is deciding which accounts belong in that tier.

Segments also drive commercial terms. Value based pricing outsourcing arrangements work only when both sides agree which customer groups the price is being set against.

Name the segments in plain language. A label like Cluster 3 never survives contact with a sales floor, while weekly bulk buyers travels on its own.

Decide the refresh rule when you build, not later. Write down which variables get recalculated monthly, which get recalculated yearly, and who signs off a change to the cut points.

Rebuild annually, and test the result. If two segments respond identically to the same offer, you have one segment wearing two names — collapse them and save the overhead.

Examples

A customer segmentation model looks different in each sector, because the variable that predicts behaviour differs. Retailers lean on recency and basket, telecoms lean on usage, and business software leans on seats and adoption.

Retail. Grocers and general merchandisers cut by recency, frequency and basket value, then test offers by group. The US Census Bureau’s retail programme publishes the most comprehensive data available on US retail activity as an outside benchmark.

Telecommunications. Operators segment by usage pattern and contract tenure, because a heavy data user on an old tariff is a very different risk from a light user who just renewed. Dates matter here; tenure bands shift every quarter.

Business software. Vendors cut by seats, adoption depth and industry, then set coverage by segment. HubSpot’s service guidance defines segmentation as grouping customers with shared characteristics such as industry, the same cut these vendors make.

Outsourced delivery. Offshore analytics teams commonly build and refresh these models for mid market clients, which keeps the cadence honest without a permanent in house data science payroll.

One rule holds across all four — the model only earns its cost when something downstream changes, whether that is a price, a contact plan or the name against an account.

Related terms

• Revenue per user: a per head measure often used to sanity check whether segments really differ commercially.
• Cross-sell matrix: the grid of product pairs that a finished segmentation model usually feeds.
• Audience segmentation: the media facing cut, built for targeting rather than for service design.
• Behavioral segmentation: grouping by what customers did, which predicts better than who they are.
• Customer value segment: the commercial cut by margin and cost to serve.

FAQ

How many segments should a model have?

Four to eight for most businesses. Fewer than four rarely changes any decision, and more than eight stops being usable by the people who apply it. Count the segments your frontline can name from memory.

What is the difference between segmentation and personalisation?

Segmentation groups customers and treats each group the same. Personalisation varies the treatment for each individual, usually driven by a model that runs on one record at a time.

Which variables work best?

Behavioural and value variables beat demographics in almost every test. They are drawn from what customers actually did with you, so they are harder to fake and easier to act on. Demographics still help in media buying, where platforms target what they can see.

How often should the model be rebuilt?

Annually in stable markets and every six months in fast moving ones. Also rebuild after any event that changes the base, such as an acquisition or a pricing overhaul. Diary the rebuild, because models nobody remembers never get rebuilt.

Can a segmentation model be outsourced?

Yes, and the build often is. Keep the choice of variables in house, because that choice encodes your commercial strategy.

Explore Outsource Accelerator’s outsourcing resources when you want delivery help building or refreshing a segmentation model.