Can you reach the customers your segmentation recommends?

A segmentation has done its first job. The team agrees which customers matter most, how many there are and what they need. Then the media planner asks how to find them, the CRM team asks which records to tag, and sales asks how to spot one in a conversation.

You can reach the customers a segmentation recommends when its segments can be recognized from information your teams and channels actually have, as precisely as the intended use requires. Describing the people who matter and finding them again outside the survey are different tasks, and they can need different evidence.

Where the description meets the data

Inside the survey, placing people in segments is simple, because everyone answered the questions that define them.

Segments built on needs, motives or attitudes are often the most useful for strategy, since they explain why people choose. Those are seldom fields a customer database records. The database knows what people bought, when and through which channel. A salesperson knows what can be learned in a few minutes of conversation.

Each of those views is narrower than the survey’s, so any rule that places people from them will place some wrongly. The useful questions are how often, for which segments, and what a wrong placement costs.

Bought audiences add a further layer, because the labels data brokers sell for ad targeting are classifications too, made by someone else. In field tests of more than 90 third-party audiences from 19 data brokers, the authors of a peer-reviewed paper in Marketing Science concluded that “audience segments vary greatly in quality and are often inaccurate across leading data brokers.”

Three uses, three standards

Some segmentations never need to find an individual. They guide strategy: which customers a portfolio or a brand should be built for.

That use needs segments that are meaningful, soundly sized and stable enough to plan around. Nobody has to identify a single customer by name.

Planning asks for more. A media plan or a distribution choice needs to know where a priority segment is concentrated: the channels, programs or regions where it makes up a larger share of the audience than it does of the market. Survey questions about media and shopping can often answer that without placing each person.

Individual recognition asks the most. Tagging records in a CRM, routing a sales lead or tailoring an email needs a rule that places each person, and an estimate of how often it will be wrong. A misdirected email costs little. A sales team spending its week on the wrong accounts costs a good deal more.

A segmentation built to be found

At a sports venue, account managers, membership services, sales, operations and marketing all served the same members, each team with its own picture of them.

The brief set one condition above the rest: every member had to be placed in a segment, accurately, from data the venue already had, so any team could look a member up.

So we built the segments twice: once from a member survey that grouped members by their needs, and once, with the venue’s data team, from fields kept for every member. Each rebuild was judged by how closely it reproduced the survey’s segments. The allocation tool was kept simple on purpose, for staff who would need to use it quickly.

The client changed its email schedule by segment, and reports that the segments informed its brand positioning and its planning of events. Being recognizable was a condition of that brief from the start, which is when it is easiest to design for.

What a usable recommendation says

A usable recommendation is clear about coverage: which customers can be classified at all.

Prospects who are not yet in the database cannot be tagged there, and customers with thin records may be placed on very little. A segment that looks large in the survey can be small in the file.

It is equally clear about error: how often the rule disagrees with the survey’s own placement, segment by segment, and what a wrong placement would cost in the intended use. A rule that is accurate on average can still struggle with the segment that matters most.

And it is candid about the trade-off. Segments defined only by what a database records are easy to find and can be hard to explain. Segments defined by needs explain choice and can be hard to find. How much meaning to trade for recognizability depends on the use.

Someone also has to own the rule once the research ends: applying it to new customers, noticing when records change or agreement with the survey drifts, and deciding when it needs rebuilding. Without that owner, the segments stay in the report while the database goes on as before.

Finding them is part of choosing them

A segment no one can find may still guide a portfolio or a brand.

It should not anchor a targeting plan until someone has shown how its members will be recognized, in which channel, and how often the recognition will be wrong. That question belongs in the segmentation brief, beside which customers matter most, because the answer can change how the segments should be built.