The Churn Problem

San Francisco · April 24, 2026

The CEO is standing at the whiteboard, and she's drawn a line that goes down and to the right. Nobody in the room is saying anything. The line represents our monthly retention rate for the last six quarters, and it looks like a ski slope.

"Four point two percent," she says. "Monthly logo churn. Four point two."

She lets the number sit there. In the silence, I can hear the espresso machine in the kitchen two rooms away. Someone's phone buzzes. The VP of Engineering is looking at his laptop screen, not at the whiteboard, which is how you know he already knew.

Four point two percent monthly churn means we lose roughly 40% of our customers every year. At our current MRR of $3.1 million, that's $1.24 million walking out the door annually — not in one dramatic exit, but in a steady drip of cancellation emails and expired credit cards and accounts that just go dark.

I've been in growth long enough to know that churn is the problem everyone talks about and nobody fixes. Not really. You can patch it. You can slow it down. But fixing it requires admitting something uncomfortable about your product, your market, or both.

The first sign was in the cohort data. I pull cohort analyses every Monday morning — it's the closest thing I have to a ritual — and in January I noticed that our November 11, 2025 cohort was retaining at 61% after sixty days. Our benchmark was 72%. That's an eleven-point gap, and I flagged it in our weekly growth meeting.

"Could be seasonal," the head of product says. She's not wrong. November cohorts are always a little weird because of the holiday signup patterns — people sign up before Thanksgiving, forget about the product, come back in January or don't.

"Could be," I say. "But December's at 63%."

She writes something down. I notice the VP of Sales is checking his email. This will become important later.

The thing about churn is that it's a lagging indicator of a leading problem. By the time you see it in the numbers, the damage happened weeks or months ago. The customer who churns in April made the decision to leave in February. They just hadn't gotten around to canceling yet.

The customer who churns in April made the decision to leave in February. They just hadn't gotten around to canceling yet.

So I started doing something I hadn't done in a while: I called churned customers. Not emailed. Called. Old-fashioned, pick-up-the-phone, hope-they-answer calling.

Out of thirty-seven attempts, I got fourteen conversations. Here's what they told me, in order of frequency:

Seven said they'd switched to a competitor. When I asked which one, five named the same company — a startup that had launched eight months earlier with a free tier that covered 80% of what our basic plan offered.

Four said they'd "outgrown" our product, which is a polite way of saying we didn't have the features they needed and they'd moved to an enterprise solution.

Two said budget cuts. Fair enough.

One said — and I'm quoting directly — "I honestly forgot we were paying for it."

We set up what the CEO called a "churn war room," which in practice meant we booked Conference Room C for two hours every Wednesday and put a lock on the calendar invite. The attendees: me, the CEO, head of product, VP of Engineering, and a data analyst named Priya who had been at the company for four months and already understood our metrics better than anyone.

Priya built what she called the "churn autopsy dashboard." It tracked thirty-two signals across the customer lifecycle — login frequency, feature adoption depth, support ticket volume, billing page visits, API call patterns. She ran a logistic regression and found that three variables predicted churn with 78% accuracy: decline in weekly active users within an account, fewer than three integrations set up, and no usage of our reporting module.

"So they're not using the product," the CEO says.

"They're using parts of it," Priya says. "The parts that are easy to replicate."

That was the moment. Right there. The parts that are easy to replicate. Our churn problem wasn't a churn problem. It was a differentiation problem. Customers were using us for basic functionality that three other products could provide, and when a cheaper option showed up, they left.

The solution — if you can call it that — came in three pieces.

First, we identified what Priya called "sticky features." These were the parts of the product that, once adopted, correlated with 89% twelve-month retention. The reporting module was the big one. Custom workflows was another. API integrations with more than two endpoints was a third.

Second, we rebuilt the onboarding flow entirely. The old onboarding was five steps and took about four minutes. The new onboarding was twelve steps and took about twenty minutes, but it got users to their first custom report within the first session. Our activation rate — defined as completing three key actions in the first seven days — went from 34% to 51%.

Third, and this was the controversial one, we killed the free tier.

"You can't kill the free tier," the VP of Sales says. "Forty percent of our pipeline comes from free-to-paid conversion."

"Forty percent of our pipeline," I say, "converts at 2.1% and churns at twice the rate of direct signups."

He doesn't have a response to that, because the numbers are the numbers. We'd been optimizing for top-of-funnel volume for so long that we'd forgotten to ask whether the people coming in the top were the right people.

We'd been optimizing for top-of-funnel volume for so long that we'd forgotten to ask whether the people coming in the top were the right people.

We replaced the free tier with a fourteen-day trial. Full-featured, no credit card required, but with a hard cutoff. The theory was that this would filter for intent — people who wanted to evaluate the product seriously versus people who wanted a free tool.

It's been six weeks since we made the changes. Here's what I can tell you:

New signups dropped 58%. That number made the VP of Sales physically uncomfortable. I watched the color drain from his face in the meeting where Priya presented it.

But trial-to-paid conversion went from 2.1% to 11.3%. And the early retention data — it's early, I want to be careful here — shows the March cohort retaining at 78% after thirty days, compared to 68% for the same period last year.

MRR is flat. We're adding roughly the same revenue we were before, but with fewer, higher-quality customers. The LTV projections, if the retention improvements hold, suggest we'll be net positive within two quarters.

If.

That's the word that keeps me up. If the retention improvements hold. If the cohort data at sixty and ninety days confirms what we're seeing at thirty. If the competitor with the free tier doesn't adjust their strategy. If the board, which meets next month, has the patience to wait for the math to work.

The CEO stopped me in the hallway yesterday. "How are you feeling about the numbers?" she asked.

"Cautiously optimistic," I said.

"That's the most Colombian thing you've ever said," she said, smiling. "Just tell me we're not going to die."

"We're not going to die," I said. Then, because honesty is the only thing I have: "Probably."

The churn rate for April, as of this morning, is 3.1%. Down from 4.2%. It's a number. It's moving in the right direction. But I've been doing this long enough to know that one month doesn't mean anything. The real test is whether the line on the whiteboard, the one the CEO drew in that silent conference room, starts bending back up.

I'll let you know.