Segmentation grounded in behaviour, not personas.

The segments in your marketing deck were probably drawn by hand three years ago. The segments in your data are different, less tidy, and considerably more useful.

01 / The problem

Where this usually goes wrong.

Churn models are the clearest example of a general problem in this area. A model that predicts churn accurately is often predicting the inevitable — it identifies customers who have already decided, at a point where no intervention will change the outcome. It scores well and saves nobody.

The useful question is not who will churn, but who will churn *and* can be retained *and* is worth retaining. That is three different problems, and only the first one is what most vendors deliver.

02 / What we build

The parts of the system.

A unified customer record: Identity resolution across your transactional, support and digital systems, with explicit rules for how conflicts are settled. Almost everything else in this category depends on getting this right first.
Behavioural segmentation: Segments derived from what customers do — frequency, recency, basket composition, channel, support contact — reviewed with the commercial team so they map to actions somebody can take.
Value and churn modelling with a horizon: Predicted lifetime value and churn risk at a specified horizon, with the intervention window built in rather than discovered later.
Measured intervention: Retention and growth actions run against a holdout group, so the effect of the action is separated from the effect of having selected likely-loyal customers.

03 / How it is measured

What we agree to be judged on.

Set before the build starts, against a measured baseline, and reported honestly afterwards — including where the numbers are disappointing.

Incremental effect, not accuracy: The measure is how many customers were retained who would otherwise have left — which requires a control group and cannot be inferred without one.
Stability: Segments that reshuffle every month are noise. Membership stability over time is checked before anything is built on top of them.
Actionability: Every segment must have an owner and an action. A segment nobody does anything differently for is a description, not an insight.
Value concentration: How much revenue and margin each segment carries, so effort goes where it pays rather than where the population is largest.

04 / Honest limits

What this will not do.

Behavioural data explains what customers did, not why. Where the reason matters — a service failure, a competitor's offer, a change in circumstance — it needs research alongside the modelling, and we will say when that is the gap.

There is also a line between personalisation and intrusion, and it is closer than most models assume. Just because an inference is possible does not mean acting on it visibly is wise; we build this constraint in explicitly.

05 / When it applies

You probably need this if:

Customer data lives in systems that disagree with each other
Retention spend is spread evenly regardless of customer value
Segments are defined by intuition and have not been revisited
Churn is noticed after the customer has already gone

Recognising two or more of these is a reasonable trigger for a diagnostic. Recognising none of them is a reasonable trigger for not spending money here yet.

06 / Related

Often scoped together.

These capabilities share data, infrastructure or evaluation approach with customer insights, and are frequently part of the same engagement.

Next step

Is customer insights the right instrument for your problem?

The diagnostic exists to answer exactly that, including the possibility that the answer is no. It is time-boxed, fixed-fee, and the written assessment is yours.

Request a diagnostic All capabilities