Resources retention

The customer who cannot leave, and the one who was always going to

Some customers churn because the product failed them. Others churn because they finished. Those are not the same problem, and averaging them together hides the only segmentation decision that matters.

·7 min read

Ask a founder why customers leave and you will get a list of product complaints. Ask which customers leave and you usually get a shrug. The second question is the useful one, because in most businesses the answer is not scattered — it is a segment, and it was predictable on the day they signed up.

There is a clean way to draw the line. A forever customer is one for whom your product has been built into something they sell:

Unless they rip that feature out, they're gonna be a customer forever. Since it's a built-in feature, they're just gonna stay forever unless they rip the feature out — and typically features are not ripped out.

Jason Cohen

Everyone else is a temporary customer. Not a bad person, not a failed sale — someone with a finite job who will finish it. The mistake is treating the second group as a retention problem to be solved rather than a population to be recognised.

The plateau this produces

Walling names only three causes of an early plateau, and the third one is this exact shape:

You've built a one-time-use product rather than building something people want to subscribe to.

Rob Walling

That is worth sitting with if you sell anything that helps someone work through a list. The customer signs up, works through the obvious items, gets the value you promised — and the subscription's job is finished. Nothing went wrong. The product did precisely what it said. It simply had an end.

The test that separates the two, from a teardown of a keyword tool, is unusually crisp:

The agency needs the tool every month, forever. The single brand needs it twice a year.

Rob Walling

Same product. Same price. Same quality. One of those customers has a structurally infinite need because their list refills every time they win a client; the other one has a list that runs out. The forever customer is not the one who likes you most. It is the one whose work does not end.

The curve that has to flatten

There is one picture that tells you which kind of business you have. Plot each signup cohort's retention over time.

Plot the retention curve. It must flatten and hold steady, not decay to zero. A flattening curve lets cohorts layer and compound; a curve that reaches zero means you restart from scratch every period.

Ruben Gamez

A curve that flattens at 60% means every month adds a durable floor to the business. A curve that keeps sliding toward zero means January's customers are gone by autumn and September is starting over — you are not building a base, you are running a series of unrelated quarters.

The flattening point is more informative than the headline churn number, because it tells you whether there is any cohort that stays. If the curve flattens for one segment and decays for another, you have found your ICP, and it took no research.

What to do about the temporary ones

The instinct is to try to fix them. Usually you can't — their need really was finite — and the effort is expensive. The more valuable move is to stop letting them distort your decisions:

That's drowning out the other metrics, which are also important. You would tune things like, well, how much can I spend on marketing or even tech support on a person that is definitely going to churn in the next ten days. Definitely. And you go, a little bit, but very little.

Jason Cohen

Concretely, three things:

  • Segment the reporting at around 100 days. Anyone who cancels inside that window is a different population; report them separately rather than blending them into a churn number that then describes nobody.
  • Ask on signup what the work is for. One question — a project, or an ongoing motion — sorts most of it, and it is the cheapest data you will ever collect.
  • Price so the two groups self-select. If a short, intense use is genuinely worth serving, sell it as a short intense thing at a price that reflects it, rather than as a subscription that will end.

The version that applies to your target list

Everything above is about your own customers. The same axis works one level down, on the list you are about to email, and almost nobody sorts on it.

For each prospect, ask: when this works, does it become part of something they sell to their own customers, or does it stay inside their building? A marketing agency that puts your output in front of its clients has made you part of its service. A single brand that uses the same output internally has bought a project.

Built in Consumed internally
Their list Refills with every client they win Runs out
Removing you Visible to their customers Invisible
What you sell them next More of the same, at more accounts An upgrade they may not need
Renewal conversation Does not really happen Happens every year

You are allowed to sell to both. The point is to know which one you are talking to before you decide how much to spend winning them, how much to discount, and how seriously to take their feature requests.

The four things to change

  1. Plot the retention curve by cohort and find out whether it flattens at all. That single chart decides whether growth compounds or resets.
  2. Add one field to your target list: does this company resell what we do? Sort descending. Work the top.
  3. Report short-life cancellations separately so the blended number stops describing an average customer who does not exist.
  4. Stop spending equally. Support, onboarding effort and acquisition budget should follow the segment that stays — which is a decision you can only make once you can see the two groups apart.

The reframe worth keeping: churn is often not a verdict on your product. It is a verdict on who you sold it to, delivered several months late.

Keep reading