Segmentation is one of the few marketing jobs that gets worse the more effort you put into it. The plan starts with four sensible groups, grows into fourteen, then needs data nobody collects, and six weeks later everyone is still sending one message to everybody. Growth Bully, a Malta performance marketing agency, inherits that half built state often, and the fix is almost always to delete most of the plan.
A segment is only real if it changes what you send. If two groups receive the same message, they were never two groups, however different they look in a report.
What makes a segment worth building?
One test: it changes the message. A segment earns its place when you would write a genuinely different subject line, offer or next step for it. If the copy would be identical, the split is a reporting filter, not a segment, and it costs you build time and ongoing maintenance for nothing.
Apply that test before you touch the tooling and most plans collapse by half, which is the point. Every segment you keep has to be defined, maintained, and fed its own content for as long as it exists. Fewer segments, each with a message that could not be sent to anyone else, beats a tidy structure nobody writes for.
Which segments should a small list start with?
Four, based on what the contact has already done with you. Never enquired, enquired and went quiet, bought once, and bought more than once. These four cover almost every list, need no extra data collection, and each one has an obvious next message, which is why they survive contact with a busy week.
- Never enquired. They know the brand but have not raised a hand. The job is one specific proof point and one easy ask, repeated patiently.
- Enquired and went quiet. The highest value group on most lists and the most neglected. They already told you they had the problem, so the message addresses whatever usually stalls a decision. This is the group a reactivation campaign exists for.
- Bought once. The cheapest revenue available, because trust is already established. The message is the natural next purchase, not a reintroduction to the business.
- Bought repeatedly. Ask them for referrals, reviews and early feedback, and stop selling them things they already buy.
Four segments, four messages. That is a programme a two person team can genuinely run every month, and running it beats a better plan that stays in a document.
Should segments be based on who people are or what they did?
What they did, nearly always. Behaviour is recorded accurately, updates itself, and predicts the next action far better than job title, company size or age. Descriptive attributes are useful for reporting and for ad targeting, but they rarely tell you what to send next, which is the only question segmentation answers.
There is one honest exception. When your offer itself differs by who the buyer is, for example an industry where the problem and the language genuinely change, then the attribute decides the message and deserves a segment. Everywhere else, behaviour wins, and the behaviour that matters most is recency: what someone did last and how long ago.
How many segments is too many?
More than you can write for every month. Divide your list by the number of segments and look at the smallest one. If a segment is too small for its results to mean anything, or too small to justify writing a dedicated message, it should be merged back into its parent until the list grows.
The warning signs are consistent across the accounts we audit:
- A segment that has not received its own message in the last quarter. It is a filter, not a segment.
- Rules so detailed that nobody can say which segment a given contact is in without checking.
- Contacts sitting in two segments and receiving both messages in the same week.
- Segments defined on fields that are usually empty, which quietly means the segment is mostly missing people.
- A structure that needs a rebuild every time someone adds a form.
Each of those is a maintenance cost that pays nothing. Cut it and put the time into the message instead.
What do you actually send each segment?
One message per segment, built on the single thing that group needs to hear next. Not a newsletter with four variants, and not the same offer with a different greeting. The point of the split is that a person who went quiet six months ago and a person who bought last week deserve genuinely different sentences.
Keep the sending rules boring: one message, one ask, one next step, and a clear rule about who gets it. That is also what keeps the automation honest, because follow up that runs itself only works when the entry condition is something the system can actually see. Segments defined in behaviour do that cleanly, segments defined in intuition do not.
Where does segmentation break?
At the data layer, almost every time. A segment is only as good as the field it reads, so a list where purchases are not recorded, enquiry dates are missing or the same person exists three times cannot be segmented reliably by any tool. Fix the record first, then split it.
Two habits prevent most of it. Record the enquiry date and the purchase date on every contact, because those two fields alone power all four starting segments. And deduplicate on one identifier so one person is one record, which is also what makes scoring and reporting trustworthy. Once that holds, the same structure supports revenue from email flows and the follow up discipline inside our LeadLock system and every lead generation programme we run.
If your list is big enough to matter and still gets one message, four segments will out earn fourteen. Send us your current setup and we will show you which four your data already supports.