fan segmentation

Segmentation is a list operation. You are not describing a person; you are deciding who receives which send, before anyone opens a thread. That is what separates it from fan types, which has its own page: a fan type tells whoever is writing one conversation what to do next, while a segment decides who is on a list at all. The two use overlapping information and answer different questions, and confusing them produces the familiar result: a blast worded for one kind of fan landing on everyone.

Which criteria actually pay?

The ones built on money and timing. Four cuts do most of the work on most accounts.

Cut	What it separates	Why it changes the message
Has bought / never bought	Two completely different audiences	One needs an offer, the other needs a conversation
Days since last purchase	Warm from cooling	Recency predicts the next purchase better than volume does
Spend band	Normal from very large	A whale never receives a discount list
Subscription state	Active from expired	An expired fan gets a win-back, not a sales message

Notice what they have in common: each one changes the text that gets sent. That is the only test a criterion has to pass. Everything else is a note.

Which criteria do nothing?

Most of the ones a platform hands you for free, which is why they get used. How many messages he has sent, how long ago he subscribed, any “engagement score” without a purchase behind it. The demographic labels are covered under fan types and they do no more work on a list than they do in a thread. They produce segments that end up receiving the same message anyway.

There is one honest exception, and it is not a property of the fan: the hour you send. Our data shows conversion collapsing between 2am and 6am and peaking in the evening. That is a scheduling decision rather than a segment, and unlike any property of the fan, it is measured.

How many segments should you run?

Four or five. The limit is not the software, it is the writing: every segment needs its own message, and a segment nobody writes for is a segment that receives the generic version.

Three rules that keep the list honest:

1. Merge two segments that get the same text. They were one segment.
2. Rebuild the cuts on a schedule. They go stale within weeks. A buyer from March is sitting in the wrong group by June.
3. Never segment on something you would not act on this week. A cut you admire and never send to is overhead.

Does segmentation replace writing to one person?

No, and this is where a well-segmented account still loses money. A segment gets the right message to the right group; it does not decide when to name a price inside a thread. Our data puts that decision elsewhere: pitching a sale before the sixth message drops conversion by roughly a third, and the optimum sits after about ten real exchanges, regardless of which group the fan was in when the mass message reached him.

Segmentation is upstream of the conversation, not a substitute for it. What it earns is relevance at scale, plus the ability to leave people out (the underrated half of the job), since every irrelevant send costs you a reader you will want later. It also feeds conversation automation, which needs a defined group before it can do anything at all.