Your goal: retention

Retention Diagnostic & Sprint

Keep the revenue you already won.

In six weeks we find which customers leave, when and why, and design the first experiments to keep them. Then, if you want it, a one-quarter sprint with a new experiment live every week.

Book a free intro call →How the 3W Loop works
For product teams that own a retention, churn or renewal number, with enough customers leaving each month to see a pattern.
The 3W Loop: where it hurts in product analytics, why it happens from story-based interviews, whether the fix works in a scoped experiment

The 3W Loop, run on conversion with product teams at

  • Taxfix
  • elyps
  • UX Pilot
  • Deviniti

Churn Costs More Than It Looks

Most growth plans start with winning more customers. Whether that growth adds up depends on how many of them stay.

You already paid for them.Every customer cost something to win: ads, content, sales time, onboarding. Each one who leaves before paying that back turns it into a loss, and takes next year's invoices with them.
Small monthly churn is a big yearly number.Lose 3% of your customers a month and you lose about 31% of them in a year. Cut that to 2% a month and you lose about 22%: one point a month is about nine points a year.
The fix keeps paying.Acquisition spend stops working the day you stop spending it. A reason for leaving, found and fixed, keeps working on every customer you win after it.

Where Retention Actually Breaks

Retention is an outcome. It moves when you find the specific cause underneath it, and it does not move when you attack it in general. Three cuts usually find the cause:

Cut by segment.An average retention number blends a segment that loves the product with one that never should have signed up. If one keeps 80% of its users and the other 30%, the fix is often who you sell to, not what you build.
Cut by the first two weeks.The account that leaves at month five usually never reached the moment where the product proved itself. That is activation wearing a retention costume, and it is fixed at the start of the funnel: see the Funnel Conversion Sprint.
Talk to the ones who left.Customers who canceled made a real decision, remember what triggered it, and gain nothing from softening the answer. We ask what made them cancel and what they switched to, not what would have kept them.

Why There Is No Retention Result on This Page

Retention results take longer to read than conversion results: a change that keeps customers shows up in renewals months later, and we have not published a retention result. We would rather show you none than one we cannot back. Here is what we can show:

Checked firstDeviniti: the question that decided where the work wentBefore any experiment, Deviniti's own Tableau data answered one question: was this a retention problem or an activation problem? Retention came back healthy, with every customer segment staying beyond twelve months, so the work went to activation: trial-to-paid went from 26% to 59% in three months, and monthly recurring revenue grew 33%. The diagnostic on this page asks that question first, too. Read the case →
One loopThe same loop, on conversionWhere it hurts in the data, why it happens, whether the fix works in an experiment. It is the 3W Loop, the same one the Funnel Conversion Sprint runs, and on conversion its numbers are public: at UX Pilot, one experiment lifted free-to-paid conversion 44.7%, tested on 120,000 users. See the conversion results →
In the openThe method, written downHow to read your retention against the benchmark for your kind of business, and how to get honest answers from the customers who left: both published, so you can judge the thinking before you book a call. The benchmarks → The interviews →

Common Problems We Diagnose

If retention is your goal, some of these will sound familiar:

Churn is up, and nobody agrees why.Sales blames the product, product blames onboarding, and the meeting that could settle it keeps moving.
The exit survey says “other”.The dropdown was written by the team, so it can only confirm what the team already suspected. The reason you lost the customer is the option nobody wrote.
Renewals look fine. Usage does not.Revenue holds while fewer people open the product each month. That is churn in slow motion, and it arrives at the next renewal.
A retention initiative every year, and the same number.Re-engagement emails and an onboarding checklist, again, because nobody established which customers were leaving or why.
The customers who leave at month five never really started.They signed, set up half the product, and faded. The cause sits in their first two weeks, long before the renewal.
You only hear from the customers who stayed.Every interview passes through the filter of people the product already works for. The ones it stopped working for are the data set nobody collects.

How It Works

Diagnostic · Weeks 1 to 3

Where: find who leaves, and when

We start in your product analytics and billing data, not in the backlog. First, the question that decides everything after it: is this a retention problem or an activation problem? Then your cohorts, cut by segment and by start month and read against the benchmark for your kind of business, and the point where the customers who leave stop using the product. You also get a simple model of what each point of retention is worth in revenue.

Diagnostic · Weeks 3 to 6

Why: hear it from the ones who left

Data shows who leaves and when. It cannot show why. Story-based interviews with five to eight customers who canceled, and a few who stayed, walk through the day they decided, what they were trying to get done, and what they switched to. The diagnostic ends with the reasons ranked, the early behavior that best predicts staying in your data, and the first test cards.

Sprint · One quarter, if you continue

Whether: test the fix

Each reason becomes a test card: what we believe, what we will change, which early behavior has to move and by how much, and what must not get worse. Scoped experiments go live through your normal release path with guardrail metrics: the first in week two, then a new one every week.

Sprint · End of the quarter

Learn: keep the loop

A learning card with the numbers: how far the early behavior moved, the retention it predicts, and what that retention is worth in revenue, with the renewal date that will confirm it. What worked, what did not, and why. The next loop, already scoped. And a team that runs the loop without us.

What's Included

Retention Diagnostic (6 weeks) - who leaves, when and why
  • Your retention, read against the right benchmark: six-month user retention and net revenue retention for your kind of business, not someone else's
  • Cohorts cut by segment and start month, with the point where the customers who leave stop using the product
  • A revenue model of your retention: what each point of it is worth in monthly revenue
  • Five to eight interviews with customers who canceled, plus a few who stayed, and a synthesis your team can use
  • The reasons customers leave, ranked by how often they came up and how much revenue sits behind the customers who gave them
  • The early behavior that best predicts staying in your data, agreed as the number the experiments have to move
  • The first test cards, each naming the number it has to move, with success criteria and guardrail metrics
  • A readout at weeks three and six for the person who owns the number, in revenue terms
Retention Sprint (1 quarter, optional) - a new experiment every week
  • A new experiment live every week from week two, run with your team in your release path
  • Weekly working sessions in the flow of work, not in a separate workshop
  • A readout every two weeks for the person who owns the number, in revenue terms
  • End of the quarter: a learning card with the numbers, and the next loop already scoped
Format
Diagnostic: six weeks · Sprint: one quarter, a new experiment every week · Remote or on-site
Recommended involvement
The product manager who owns retention, a designer, an engineer or tech lead, whoever owns product analytics, someone from customer success, and the leader who owns the number

What We Need From You

A diagnostic only finds something if the setup lets it. Before we start, we agree on five things:

  • Access to your product analytics and your billing or subscription data, even if only as an export
  • A list of the customers who canceled in the last three months, and a way to reach them. We write the invitations; they go out in your name
  • A team that owns the number: product, design and engineering, with time to act on what we find
  • For the sprint, room in the release path for a new experiment every week, behind a flag or to a share of customers
  • Enough customers leaving each month to learn from. If too few leave for that, we will tell you on the first call, and it is good news

FAQs

Why start with a diagnostic and not with experiments?
Because retention does not move when you attack it in general. The usual retention initiative fills the backlog with re-engagement emails before anyone knows which customers are leaving or why. Six weeks of cohorts and churned-customer interviews show where the experiments should go, and whether they should be retention experiments at all.
Can you measure retention within a quarter?
Not all of it, and we will not pretend otherwise. A change that keeps customers shows its full effect at the next renewal. So the diagnostic looks for the early behavior that best predicts staying in your data, and every experiment is read on that behavior within weeks, together with the retention and revenue it predicts. If your data does not show one strong enough to steer by, we say so at the end of week six, before you decide on the sprint. The full effect keeps arriving after the quarter, and the learning card says which number to watch.
What if it turns out to be an activation problem?
Then we will say so, usually by week three, and the fix belongs at the start of the funnel, not at the renewal. At that point the Funnel Conversion Sprint is the better next step.
Do we have to continue to the sprint?
No. The diagnostic stands on its own: you keep the cohorts, the interviews, the ranked reasons and the first test cards, and your team can run the experiments. You decide at the end of week six, with the reasons on the table.
How many churned customers do you need to talk to?
Five to eight is usually enough to see a pattern. One vivid cancellation story is an anecdote; five people describing the same broken moment is a finding. We write the invitations, and where we can, we ask within days of a cancellation, while the moment is fresh.
What if churned customers will not talk to us?
Some will not, and that is expected. Three things raise the response: ask within days of the cancellation, while the moment is fresh; send a short personal note from the product manager or the account's customer success manager asking for fifteen minutes, with no survey link; and if you need to, pay for the time, because a gift card is a fair trade for the most honest interview in your research. We write the notes; they go out in your name.
Why is there no retention case study on this page?
Because we have not published a retention result. We would rather show you the method and the public conversion numbers of the same loop than a claim we cannot back.
Does this work for consumer products?
Yes, if enough customers leave each month to learn from. The benchmarks differ, and consumers need a different invitation to an interview, but the questions are the same: who leaves, when, and what made them go.

Reading for Retention Teams

Six field notes on the thinking behind this diagnostic:

Metrics

What good retention actually looks like for B2B products

Around 60% of users active after six months is good for SMB and mid-market SaaS. The benchmarks, and how to use them without chasing the number.

User Research

Interview your churned users, not just the ones who stayed

They already decided, so they have no reason to be polite. Three questions to ask about the day they left.

Metrics

How to improve user activation: find the behavior that predicts a paying customer

Rule out retention, find the early behavior your paying customers share, then test the smallest fix with guardrails.

Product-Market Fit

Iterate on WHO, not what

An average can hide a segment that should never have signed up. How to tell an audience problem from a product problem.

User Research

Talk to five customers this week

The cheapest way to de-risk your roadmap costs one hour and zero engineers: talk to five customers.

Product Discovery

What is the 3W Loop? Where, Why, Whether

Three questions, asked in order, each with its own instrument. One loop run end to end on a real onboarding problem.

Service Area

Designed for product teams in B2B and consumer products with enough monthly cancellations to learn from. Available remotely (global) and on-site across Europe. Aleksander is based in Wrocław, Poland.

Losing people before they ever pay? That is a conversion problem, and it has its own page: the Funnel Conversion Sprint.

Keep more of the customers you already won.

Book a free intro call →