The situation
A company serving three or four recognisable customer types with one product. One of those types retains far better than the others, expands, and refers. The rest churn at a rate that would be alarming if it were not buried inside a blended average.
Marketing addresses all of them equally, because all of them buy, and because narrowing feels like turning down revenue. Ask who the ideal customer is and you get a description of the market rather than a definition of a buyer.
What the company usually believes
That the churn was a product gap, and that the segments leaving would stay once a few missing features shipped. Roadmap was being allocated against the complaints of accounts that were never going to be profitable.
Which of the three it is
A positioning problem, of the kind that looks like a retention problem in every dashboard the company owns.
How the other two get ruled out
Not a downstream leak in the ordinary sense. The leak was real, but it was predictable from the account’s attributes on the day it signed up, which makes it an acquisition decision rather than a retention failure. Nothing built after signup would have changed the outcome.
Not a lack of urgency. These buyers acted, quickly, and regretted it later.
The evidence
Split retention, expansion and support load by segment rather than looking at the blended figure. The blend is what hides this, and almost every company looks at the blend.
Then look at when the difference becomes visible. If the segments diverge in the first fortnight, the company is acquiring its own churn.
What we do
We write an ICP definition specific enough to exclude people, because a definition that excludes nobody is a description. Then we rebuild the positioning around the segment that stays, say what changes for that buyer rather than what the product does, and carry it across every surface so the promise made in an ad matches the promise the product keeps.
The other segments do not come out of the product. They come out of the writing.
What changes
Signups fall, which belongs in the plan before it happens so that nobody panics in month two. Average contract value and retention improve, mostly by not acquiring the accounts that used to leave.
What the company stops doing
Allocating roadmap against complaints from accounts that were never a fit. Writing landing pages that try to be true for four buyers at once and are compelling to none of them.
The lesson
An ICP is useful when it disqualifies. Deciding who you are not for feels like turning down money, and for a quarter or so it is turning down money. It is also the fastest way to make everything downstream cheaper.