MarketingFramework / modelModelAccessible

Segmentation and personalised marketing

How can segmentation and personalised marketing support strategic choice or positioning?

AccessibleStrategicIndividual2 min read

Contents

Segmentation is the process of slicing up the ‘mass market’ for a particular product or service into a number of different segments, each one consisting of consumers with slightly different needs.

Segmentation divides a market into groups whose differences justify a different marketing response. Personalised marketing takes that logic towards the individual, selecting content, offers or experiences from customer-level data. Both methods aim to improve relevance; both can create complexity, unfairness or intrusion when applied without discipline.

When to use it

Match offers and experiences to materially different customer needs.

Identify groups whose needs are underserved.

Focus acquisition, service and product investment.

Personalise when individual-level variation creates enough value to justify data, operational and governance cost.

Origins

Economic ideas about differentiated demand developed in the 1930s, including Edward Chamberlin’s work on product differentiation. General Motors, under Alfred P. Sloan, had already demonstrated a portfolio strategy—“a car for every purse and purpose”—in contrast with a more standardised Ford model.

Wendell Smith formalised modern market segmentation in 1956. Wind and Cardozo later defined an industrial segment in 1974 through shared characteristics relevant to response. Personalised and one-to-one marketing grew in the 1990s as customer databases, digital channels and configurable operations made individual treatment more feasible.

What it is

Segmentation identifies groups that are distinct, measurable, reachable, actionable and economically worthwhile. The goal is not to find the maximum number of clusters; it is to choose where a different offer, channel, price or message improves customer value and organisational performance.

Variables may include geography, demography, firmographics, needs, attitudes, occasions and observed behaviour. Demographic proxies are convenient but often weak explanations of choice and can encode sensitive characteristics. Behavioural data can be more direct yet still reflect unequal access, historical bias or one channel rather than durable need.

Personalisation operates at different depths: selecting a relevant message, ranking recommendations, configuring a product or setting service. Amazon recommendations, a configurable Dell computer and a custom portal illustrate different mechanisms. Individual treatment should not be presented as perfectly knowing the person.

How to use it

Define the market, customer job and decision the segmentation must change.

Gather data from customers and non-customers using lawful, proportionate methods. Combine qualitative understanding with transactions, journeys and survey evidence. Avoid collecting attributes simply because they are available.

Create candidate groups through judgement, rules or clustering. Validate them on held-out or later data for distinct needs, size, reach, stability and response. Examine whether protected or vulnerable groups are unfairly excluded, targeted or priced.

Name segments neutrally and develop a proposition, channel, economics and measurement plan for each. Pilot differentiated treatment and compare incremental outcome with the same treatment for everyone.

For personalisation, define the benefit, permissible data, consent or legal basis, model logic, frequency cap and user controls. Measure recommendation quality, diversity, customer welfare and long-term effect—not only immediate click or sale. Allow people to correct preferences and opt out where required.

Tesco’s Clubcard became a well-known loyalty-data example: purchase histories informed targeted offers through analytical work associated with Dunnhumby. The lesson is not that more tracking is always better. Loyalty exchange, transparency, security and data minimisation determine whether the practice remains acceptable.

Top practical tip

Compete on the segmentation logic, not just the labels. Test a different basis—occasion, unmet job, constraint or desired outcome—when every competitor uses the same category map. The SUV example became a major segment around 20 years before the original account, but current innovation still requires new evidence rather than recycling that boundary.

Top pitfall

Do not over-segment, infer sensitive traits unnecessarily or personalise in ways customers cannot understand or contest. Tiny groups may be unstable, operationally expensive and discriminatory; new markets also require experimentation because stated intentions may not predict behaviour.