Insights / Guide / Consumer behaviour analysis
GuideSegment your customers from the data you already have
How to build useful customer segments from transaction data you already hold, add the why through a few conversations, and turn segments into action.
Most businesses talk to all their customers in roughly the same way: the same newsletter, the same offers, the same follow-up. It is simple to run, but it treats a loyal customer of ten years and a first-time bargain hunter as if they were the same person. Customers notice.
In a 2021 McKinsey survey of US consumers, 71% said they expect companies to deliver personalised interactions, and 76% said they get frustrated when this does not happen. Even if the figures differ in your markets, the direction is clear: customers notice when they are treated as a crowd.
The good news is that useful segmentation rarely needs a large research budget or a new platform. Most of what you need is already sitting in your sales and customer systems. This guide shows how to use it.
Start with what you already know
Most businesses hold more customer information than they use. A typical order or till system records who bought what, when, how much they spent, through which channel and whether they used a promotion. Over a year or two, that history describes behaviour far more accurately than any survey of intentions.
Three measures that do most of the work
A simple and well-tested starting point looks at three measures for each customer: how recently they last bought, how often they buy and how much they spend. Each customer can be scored on each measure, for example from one to five, and grouped by their combined pattern.
- High-value regulars: bought recently, buy often and spend well. They deserve recognition and early access, not discounts.
- Occasional bargain buyers: appear mainly during promotions and buy little at full price. They respond to offers but rarely become loyal.
- Promising newcomers: bought once or twice recently. The next ninety days decide whether they become regulars.
- One-time visitors: bought once, some time ago, and never returned.
- Lapsed customers: once bought regularly but have not returned for longer than their usual interval.
Add what and how
Then look at what each group buys, in which categories, through which channels and at what times. You may find that high-value regulars buy across many categories while newcomers start with a single entry product, or that lapsed customers drifted away after a particular price change or range decision. Each pattern suggests a different action.
Business-to-business firms can use the same approach with a few adjustments. Replace individual shoppers with accounts, and add measures such as the number of product lines bought, the share of orders placed online and the number of contacts who order. An account that buys from one product line through one contact is both at risk and an obvious opportunity.
A worked example
A typical regional homeware retailer might find that a small group of regulars produces a large share of full-price sales, that most newcomers first buy a single item in one category, and that customers who make a second purchase within two months are far more likely to stay. The obvious action is to focus on that second purchase.
Add the why
Numbers show what people do, not why they do it. A lapsed customer may have moved house, found a cheaper competitor, had a poor delivery experience or simply stopped needing your product. Each reason calls for a different response, and the data alone cannot tell them apart.
Speak to a handful of customers in each segment, ideally by phone or video for twenty minutes. Ask open questions and listen more than you talk.
- What were you trying to achieve when you first bought from us?
- What else did you consider, and why did you choose us, or not?
- What nearly stopped you from buying?
- What would make you buy more often, or come back?
- How do you prefer to hear from us, if at all?
A dozen conversations often reveal motives that no dashboard shows. You may discover that your most valuable customers buy for a professional use you never marketed to, or that newcomers are confused by your delivery options. These insights are cheap to gather and often change the plan entirely.
Choose the people to call deliberately. Include customers who fit the segment well and a couple who sit at its edges, and make sure lapsed customers are represented, even though they are harder to reach. Write up each call in a few lines straight afterwards, and look for the reasons that come up three or more times. Those recurring reasons are the ones worth acting on.
Make segments usable
A segment is useful only if you can reach it and act differently towards it. Elaborate personas with names, hobbies and stock photos tend to be admired in a workshop and forgotten a month later. Practical segments are short, specific and tied to action.
For each segment, write four lines:
- A one-line description based on behaviour, such as “bought once in the last 90 days, single category”.
- The main need or motive, from the conversations.
- The best channel to reach them: email, post, in store, sales representative or social media.
- One action to try this quarter, with a measure of success.
Keep the number of segments small, typically four to six, so that teams can remember them and act on them. A sales team that knows six segments by heart will use them. One that receives a twenty-segment model will ignore it.
Make sure the segments can be applied in the systems people use every day. If the sales team works in a CRM, add the segment as a field on each account. If marketing sends email from a separate tool, sync the segment there automatically. A segment that lives only in an analyst’s spreadsheet will not change how anyone treats a customer.
Test before you scale
Treat each segment action as a test. Send the second-purchase offer to half the newcomers and nothing to the other half, then compare repeat rates after two months. Invite half the high-value regulars to an early preview and compare their spending with the rest. Simple comparisons like these show which actions actually change behaviour and which merely reward people for what they would have done anyway.
Refresh the segments every quarter. Customers move between groups, and the movement itself is informative. A rising number of lapsed customers is an early warning that deserves attention well before it shows up in revenue.
Respect privacy
Collect and use customer data in line with local laws and the consent customers have given, and be clear with customers about how their data is used. Segmenting on purchase behaviour within your own business is generally straightforward. Combining your data with third-party sources, or using sensitive characteristics, needs much more care and usually legal advice.
Personalisation that feels helpful builds trust. Personalisation that reveals more than customers expect you to know damages it.
A first step this month
Export twelve to twenty-four months of transactions, score customers on recency, frequency and spend, and group them into four to six segments. Call three customers in each. Then choose one action for one segment, run it as a test and measure the result. That single cycle will teach your team more about its customers than another year of uniform campaigns.
Source: McKinsey, Next in Personalization 2021 (US consumer survey).
