The Churn Diagnosis Playbook. 9 ways to find the real reason customers leave, and where the operators disagree.

Before you touch a single feature, look at the shape of your retention curve. On Y Combinator's Startup School breakdown of how to keep users, the test is simple and brutal. Do the cohort curves flatten out, or do they decay toward zero? A flattening curve means a stable base of people who keep coming back, which is the only real evidence the product retains anyone. The absolute numbers matter far less than the shape.

A curve that never flattens is a leaky bucket no amount of acquisition will ever fill, and it is the first thing to read before you decide churn is a feature problem.

Jason Cohen built two unicorns, and on Lenny's Podcast he names logo churn as the single most critical problem to solve, ahead of almost everything else. The reason is mathematical, not emotional. Every customer who cancels has to be replaced before you grow at all, so as you scale, churn quietly sets a ceiling on how big you can ever get, no matter how good your acquisition is. You cannot out-market a leaky bucket.

The first diagnosis is to stop staring at new-logo growth and look at the rate customers reject the product after going through your entire onboarding gauntlet, because that number is the one that caps the business.

Jason Lemkin's rule on 20VC is that a churn number means nothing until you set it against the segment you actually sell to. Enterprise SaaS needs triple-digit net revenue retention to even count as SaaS. Small businesses, on the other hand, churn 3 to 4 percent a month as a rule, because they cancel everything the moment it stops paying back. The same raw number is healthy in one segment and fatal in another.

So before you panic or relax, benchmark your churn against your buyer, not a number from a blog post. If it is abnormal for your segment, the cause is usually that you sold to the wrong customer, not that the product is weak.

Alex Hormozi draws a line most founders miss. Some of your churn is structural, and you should not lose sleep over it. He tells a founder running a fulfillment tool with 33 percent monthly churn that the number is the nature of the segment, since 99 percent of the traffic is brand-new businesses that mostly fail on their own. Shopify, he points out, churns more than they do.

The diagnosis is to split your churn in two. One part is structural, baked into who the segment is, and no feature fixes it. The other part is the churn you actually caused with weak activation and delivery, and that part is yours to fix. Confusing the two wastes quarters.

Bob Moesta, who co-created Jobs to Be Done, says the richest signal you have is sitting in the people who already left. When a customer churns, something in their context changed, and the job they originally hired you for either went away or got handed to something else. The switch that brought them in has reversed. Interviewing churned users surfaces that struggling moment with a clarity current users can never give you, because the people who stayed are not feeling the friction.

Do not survey them. Talk to them, and trace backward to the exact moment the product stopped fitting the job they had.

The counterweight, also from Moesta, is a warning against your loudest users. Complaints are not churn. People who gripe about a missing feature are often the ones who never leave, and people who quietly cancel rarely complain first. Bitching ain't switching, as he puts it, because switching takes real friction and most complainers never pay that cost.

So weight your diagnosis by behavior, not volume. The feature ten vocal users demand may have nothing to do with why the silent ones are leaving. Look at what churned users did, not what current users say, or you will rebuild the roadmap around the people who were always going to stay.

When Hormozi's team fixed a churn problem, they did not touch the product. They fixed onboarding, reset the expectations set before the customer started, and aligned the incentives. The net change was a 61 percent reduction in cancellations and a 2.4x lift in customers ascending to higher-priced tiers. The lesson for diagnosis is to look hard at the first thirty days before you look at the feature set.

Most churn is decided in the gap between what the customer expected and what they actually got in their first sessions. If you have a high-touch window at the start and you are not using it to get people to a first real win, that, not the roadmap, is very likely where your churn is being manufactured.

Albert Cheng worked on retention at Duolingo, Grammarly, and Chess.com, and on Lenny's Podcast he gives a consumer benchmark worth holding. A day-one retention rate of 30 to 40 percent is healthy. The deeper point is that the right number depends entirely on your model, and for a mature product, retention mechanics are a more powerful growth lever than acquisition. So part of the diagnosis is choosing the right yardstick.

A consumer app lives or dies on day-one return and habit loops. A B2B tool lives on net revenue retention and expansion. Diagnosing churn against the wrong benchmark for your model sends you chasing the wrong fix entirely.

Most churn advice assumes a subscription, which quietly breaks for one-time-purchase and physical-product businesses. On My First Million, the CEO of Ridge Wallet, a 200-million-dollar brand, names the trap directly. Over-relying on theoretical lifetime value while you are unprofitable on the first purchase. For non-subscription businesses, retention is a repeat-purchase-rate game, not a cancellation-rate game, and the discipline is to make money on order one rather than betting on a lifetime that may never arrive.

So if you sell a one-time product, a service, or anything non-recurring, stop importing SaaS churn math. Diagnose retention as reorder rate and first-order profitability, the metrics that actually govern your model.