Segmenting contacts: fewer groups, better results

Segmentation is almost always drawn too finely. Twenty groups means twenty versions of every message — and because nobody sustains that, everyone ends up receiving the same one. Four segments that genuinely get addressed differently do considerably more.

A scattered field resolves into exactly four clearly separated glowing clusters

In short

  • A segment is only a segment if it receives a different message. Anything else is a filter.
  • Three characteristics are worth separating on: buying stage, fit, and how recent the last activity was.
  • Four to six segments are the upper limit of what a small company can sustain.
  • The most important segment is the one that receives nothing any more — inactive contacts cost money and lower the delivery rate.

The usual segmentation gets built backwards: groups are created because the system offers them, and only afterwards does anyone consider what to write to them. That reliably produces many groups and one message.

The test is simple: if two groups receive the same message, they are not two groups.

The three characteristics worth separating on

1. Buying stage

Where is this person? Problem still unclear, looking for an approach, comparing providers, customer. The most important characteristic — it decides which kind of content fits at all.

2. Fit

Does the person belong to the audience? Industry, size, role. It does not separate the message but the question of whether any effort is warranted.

3. How recent the last activity was

When was the last reaction? The most underestimated characteristic — it decides deliverability and cost.

Not useful as segments in their own right: gender, location with no connection to the offering, sign-up date, or which form someone came from. Those details can tailor content — but they do not justify a message of their own.

The four segments that are enough for most

SegmentWhoWhat they getHow often
Newless than 30 days in the databasewelcome sequence5 messages over 18 days
Active and fittingreacted in the last 90 days, in the audiencewritten pieces, occasionally an offer1–2× a month
Customerexisting customer relationshipusage notes, changes, repeat orders1× a month
Dormantno reaction for 180 daysone question, then nothing1× a quarter

Two or three further segments come in if you have very different offerings or work in several languages. More than six is, in experience, not sustainable in a small company.

Worth knowing

The dormant segment is the most valuable — not for what it receives, but for what it no longer receives.

Contacts who have opened nothing for six months lower the delivery rate for everyone else: mailbox providers factor in how many recipients actually engage with your messages. Continuing to write to them also costs money on tiered pricing. A maintained list of 800 active contacts is worth more than an unmaintained one of 3,000.

Setting it up in four steps

  1. Look at the database before making rules. How many contacts have reacted at all in the last 90 days? That number decides whether segmenting is worth it — below 200 active contacts, usually not.
  2. Write rules, do not maintain lists. A segment is a condition that applies automatically, not a distribution list filled by hand. Hand-maintained lists are out of date within three months.
  3. Write one message per segment. Before the segment goes live. If nothing comes to mind that differs from the others, delete the segment.
  4. Settle the overlaps. A contact can meet several conditions. There has to be a precedence — customer beats active, dormant beats everything, for instance.
Careful Step four gets skipped readily and leads to the most embarrassing mistake there is: an existing customer receiving the welcome sequence for prospects. A clear precedence, written down in one place, prevents that once and for all.
From practice

The most common state after two years: seventeen segments, five of them empty, six overlapping, and three that have received no message in a year. Nobody knows any more by what rule they get filled.

Clearing up goes fastest through one question per segment: when did this group last receive a message no other segment also received? If the answer is "never" or "over a year ago", the segment can go — with no replacement.

One legal point

Segmenting by behaviour — who opened and clicked what — is processing of personal data and, depending on how it is done, profiling. Three points belong settled:

  • Transparency: the privacy notice states that and how analysis takes place.
  • Basis: with extensive behavioural analysis, a legitimate interest does not always suffice.
  • Objection: it has to be possible and then has to end the segmentation too, not only the sending.
Prompt
Help me cut our contact segments down to what makes sense.

Our situation:
- Contacts in total: [number]
- Of those, reacted in the last 90 days: [number]
- Offerings we have: [list]
- Languages: [list]
- Segments that exist today: [list with condition, where known]
- Messages we have sent per segment in the last six months:
  [details]

Tasks:
1. For each existing segment, state whether it has ever received
   a message no other segment also received. If not, recommend
   deleting it.
2. Name segments that overlap and propose a precedence that
   decides unambiguously.
3. Propose four to six segments matching our situation – each
   with a condition, content and frequency.
4. For each proposed segment, state in one sentence what this
   group receives that no other does. If you cannot manage that
   for one of them, delete it.
5. Name the condition from which a contact should receive no
   further messages, and what gets tried once before that.

Do not invent figures about our database.

In closing

Segmentation is not a question of fineness but of sustainability. Four groups that genuinely receive different messages do more than twenty that all receive the same one.

And the most effective segment is the one nothing goes out to: consistently retiring inactive contacts improves deliverability for everyone else and saves licence cost as well.

Common questions

How many segments do you need?

Four to six is the upper limit of what a small company can sustain: new, active and fitting, customer, dormant — plus perhaps one per very different offering or per language. The test for each: does it receive a message no other segment receives? If not, it is a filter, not a segment.

Which characteristics should you segment on?

Three: buying stage (problem unclear, looking for an approach, comparing providers, customer), fit with the audience, and how recent the last reaction was. Not useful as segments in their own right are gender, location with no connection to the offering, or sign-up date.

Why should inactive contacts be removed?

Because they lower the delivery rate for everyone else: mailbox providers factor in how many recipients actually engage with the messages. They also cost money every month on pricing tiered by contact count. A maintained list of 800 active contacts is worth more than an unmaintained one of 3,000.

From what point is segmenting worth it at all?

From roughly 200 contacts that have reacted in the last 90 days. Below that, the effort of separate messages exceeds the gain — one good message to everyone is then the better decision than four mediocre ones to four groups.

What has to be considered legally when segmenting by behaviour?

Analysing opens and clicks is processing of personal data and, depending on scale, profiling. What is needed: transparency in the privacy notice, a sound lawful basis — with extensive analysis, a legitimate interest does not always suffice — and an objection route that also ends the segmentation, not only the sending.

Marketing that sets itself up

The Studio Engine beta is live. Claim your spot and help shape it from the start.

Join the beta →
← Back to overview