Segmenting a small list

Segmentation is sold as a sophisticated practice requiring software, and at a small size it is three or four groups kept by hand that make every message slightly more relevant. The sophisticated version is for lists large enough that the differences are measurable, which is not most lists.

The splits that are worth making

Each of these changes what you would actually write, which is the only test of whether a segment is real.

• Customers and non-customers. The single most valuable split. One group needs to be persuaded and the other needs to be looked after, and sending a first-purchase offer to somebody who bought last week is the most visible waste in email.
• What they came for. Somebody who subscribed for a checklist about pricing is not the same person as one who came for a guide about shipping.
• How long ago they arrived. New subscribers need context that long-standing ones find repetitive.
• Whether they still open anything. Not a segment you message differently so much as one you stop messaging, which is covered in list hygiene.

Notice what is absent: demographics. For a small business, what somebody did with you predicts what they want far better than who they are, and you already have the first and not the second.

How to do it without a system

1. 01Tag at the point of entry. Which form, which offer, which page. A tag added when somebody arrives costs nothing; reconstructing it later is impossible.
2. 02Tag at purchase. Customer, and what they bought.
3. 03Keep the number of tags small enough to remember. Ten tags nobody uses is worse than three that get used.
4. 04Before each send, ask only whether anybody on the list should obviously not receive this. That question does most of the work of segmentation with none of the machinery.

When to stop

Segmentation has a cost: every split halves the audience for a given message and doubles the work of writing. There is a point where you are writing four versions of the same email for groups too small for the difference to matter.

A reasonable stopping rule: if a segment would receive a message that differs only in a sentence, it is not a segment. If you would write something genuinely different, it is. Most small businesses need two or three and build nine.

What you actually do with them

Segments are usually built and then not used, because nobody decided what would be different. Four concrete uses, which is most of the available value.

• Exclude rather than target. The first and best use. A first-purchase offer goes to everybody except customers; a product announcement skips the people who already own it. Exclusion is easier than targeting and prevents the errors customers actually notice.
• Send the same message with a different opening line. Two sentences changed for people who came through a particular offer is often the whole of useful personalisation, and it costs minutes rather than writing two emails.
• Change the frequency rather than the content. New subscribers can take more; long-standing quiet ones should get less.
• Hold a segment back. Send to your most engaged group first, see what happens, then send to the rest a day later with anything you learned fixed. This is the cheapest testing available to a small list and almost nobody does it.

That last one is worth more than split testing at a small size, because it tests the whole message against a real audience rather than one variable against noise.