The churn you already paid for

Most retention plans start with the same artefact: a list of accounts approaching renewal, sorted by ARR, colour-coded by health score. It is the wrong list.

By the time an account reaches renewal, the decision has usually already been made — quietly, months earlier, by someone who stopped opening the product and never told you. The renewal call is where you find out. It is not where it happened. It happened in week three.

The arithmetic almost nobody does

Take a company I will call Meridian: forty people in Wellington, selling compliance workflow software to mid-sized accounting firms. Average contract value is NZ$14,400 — $1,200 a month. They closed sixty new logos last year. Their CAC payback period is thirteen months, which is respectable; the 2026 Aleph and Benchmarkit joint report puts the median B2B SaaS payback at sixteen months across 342 companies, with the top quartile at six months or fewer and the bottom quartile at twenty-four or more.

Thirteen months of payback means each new logo cost roughly thirteen months of its own revenue to acquire. Call it $15,600 in blended sales and marketing spend per customer.

Now the part that hurts. Of those sixty logos, twenty-two never reached what Meridian considers activation — importing a client list and completing one compliance cycle — within ninety days. Fifteen of the twenty-two were gone by month seven.

Fifteen customers at $15,600 of acquisition cost is $234,000 spent. Against that, those fifteen paid an average of six months of subscription before leaving: roughly $108,000 collected. The net destroyed is about $126,000, on sixty logos, in one year. That is a senior engineer, or most of a trade-show budget, or the runway difference between raising in March and raising in June.

Then comes the second bill. To replace fifteen logos, Meridian spends another $234,000 acquiring them. The pipeline does not care that the last fifteen were preventable.

Here is the comparison that should decide the quarter's priorities. Recovering a third of those fifteen accounts — five customers, $78,000 of already-spent CAC redeemed plus their ongoing revenue — required knowing why they stalled. Twenty-two phone calls, twenty minutes each. Under eight hours of talking.

Why an already-signed customer is the cheapest thing you can save

New-logo acquisition is expensive because it front-loads every hard problem at once. You have to find the account, earn attention, prove the category exists, differentiate, survive procurement, pass a security review, and get budget approved in a fiscal year that was planned before you called.

A customer who signed and then stalled has cleared every one of those. The budget is approved. The security review passed. Someone inside the building has publicly attached their name to the decision — which means they have a personal stake in it working, and they would rather it worked than have to explain a cancelled contract to the person who signed off.

That last point is underrated, and it cuts both ways. The champion who cannot get your product working is often the least likely person to raise their hand, because raising their hand means admitting the thing they advocated for is not delivering. Silence from a stalled account is not neutral. It is frequently the sound of someone hoping the problem resolves itself before anyone notices.

The trap: knowing that they stalled tells you nothing

This is where most teams stop, and it is why the lever stays unpulled.

Product analytics will tell you, accurately and in real time, that fourteen accounts have not completed step three. That is a symptom with roughly forty plausible causes, and the dashboard cannot distinguish between them — your funnel tells you where they stopped, never why.

The actual reasons, in my experience of reading these conversations, look like this:

The data export from the customer's previous system came out in a slightly non-standard CSV that your importer rejected with a generic error, and the champion spent forty minutes on it, felt stupid, and quietly put it back on the pile.

Their IT team would not whitelist your sending domain. The ticket sat in a queue for five weeks. Nobody thought to mention it to you because from their side it was an internal problem.

The person who signed the contract left in month two. Their replacement inherited eleven tools and no context on which ones mattered. Yours was not the one with an angry stakeholder attached.

The product does exactly what was promised, but the promise was made to a VP and the daily work is done by three people who were never in the room and have their own way of doing it, which works well enough.

None of those four are visible in an event stream, and the survey you would normally send cannot recover the reason either. All four are recoverable, cheaply, if you learn about them in week three rather than month eleven. Two of them are fixed with a single email. One is fixed with a fifteen-minute call to the new owner. One means the account was mis-sold and you should know that, because it is a sales problem wearing an onboarding costume, and it will keep happening.

Two caveats on the numbers, because you will be asked

The most-quoted activation benchmark in this space is an average of 37.5 per cent — the share of new signups who reach a product's core value event. It comes from Userpilot's 2024 benchmark work, drawn from 62 B2B companies using their activation dashboard. Treat it as directional and nothing more. The sample is self-selected (companies that bought an onboarding tool), it is one vendor's definition of activation, and 62 companies is a small base for a number that gets quoted like a law of physics. Your own activation rate is the only one that matters, and you can calculate it this afternoon.

The other number you will meet is "a 5 per cent increase in retention increases profits by 25 to 95 per cent." It traces back to Frederick Reichheld and W. Earl Sasser's 1990 Harvard Business Review article "Zero Defections: Quality Comes to Services." The 25 per cent figure was specific to financial services, and the upper end of that range is a widely repeated distortion of a finding about one bank's branch system. The underlying claim — that retained customers are disproportionately profitable — is sound. The range is not a benchmark for your business, and a CFO who knows the provenance will take the rest of your slide less seriously.

Say what you can source. Where you cannot source it, say that too. It costs you nothing and it is the difference between an argument and a vibe.

What to actually do this quarter

Define one activation event. Not a funnel with five stages — one moment that means this customer is now getting the thing they bought. If your team cannot agree on what it is in a single meeting, that disagreement is itself the finding.

Compute ninety-day activation for the last four quarterly cohorts. You are looking for a rate and a trend, not precision.

Pull the list of accounts that did not make it. This is the list your retention plan should have started with.

Then call them. Not a survey — a call. Twenty minutes, no agenda beyond understanding what happened, no attempt to save the account on the first pass. You are collecting reasons, not defending the product, and there is a working script for the 30-day call if you would rather not invent one.

The reassuring part is that this is a finite exercise. Abbie Griffin and John Hauser's 1993 Marketing Science paper "The Voice of the Customer" — still the best empirical work on this question — found that twenty to thirty customer interviews surface ninety to ninety-five per cent of customer needs, and that twenty interviews captured over ninety per cent of what thirty revealed. You are not signing up for a permanent research function. You are signing up for roughly twenty-five conversations, after which the reasons start repeating and you can stop. Set against $15,600 of acquisition cost per logo, what an interview costs per customer is a rounding error.

Count the reasons. Fix the top two. Rerun the cohort.

The uncomfortable part

Most companies do not do this, and the reason is not cost. Twenty-five phone calls is not a budget line.

The reason is that the calls are unpleasant. You are ringing people who bought something from you and did not get value from it, and you are asking them to explain why, in their own words, to someone who works for the company that sold it. Sales does not want to make those calls. Customer success is measured on the accounts still alive. Product has a roadmap that was agreed in July.

So the calls do not get made, the reasons stay unknown, the next cohort stalls in exactly the same place, and the retention plan starts again with a list of accounts approaching renewal.

The customers who stalled in onboarding are the ones you have already paid full price for. Everything after this point is margin.

---

Where Nosie fits

The exercise above is finite and it is cheap, and it still does not happen — because the calls are unpleasant and no quarter has eight spare hours of someone senior in it.

Nosie is the part that makes them happen anyway. You give it the cohort that did not activate; it runs the interviews as an outbound voice study, rings each account, asks what actually happened, and follows the answer wherever it goes. The reasons come back transcribed, tagged by cause and counted — which is the ranked list the twenty-five phone calls were for, without anyone having to make twenty-five phone calls.

Try it on yourself. Your first self-test interview is free: Nosie rings you, so you hear exactly what a customer would hear before a single customer is contacted.