The month my churn quietly doubled at $20K MRR: a retention diary (2026)

Every founder learns to read one chart before coffee. Mine was monthly recurring revenue, and for four months in early 2026 it kept going up. From $16,250 to $20,560. A 27 percent climb in a quarter. On paper, the best stretch the company had ever had.

Underneath it, the money leaving every month had more than doubled. I almost did not notice, because the number on top kept getting bigger.

This is a first-person diary of that quarter: how gross churn quietly doubled while my headline MRR grew, how net revenue retention slid under 100 percent without anyone raising a hand, and the boring cohort table that finally showed me what was happening.

Editor's note: this is a composite operator diary. The arc is real and drawn from my own dashboards, but the figures are rounded and identifying details are changed. Treat it as one founder's honest account, not a benchmark.

Quick answer (2026): MRR growth can hide rising churn because new sales are added on top of the leak, not netted against it. The metric that exposes the leak is net revenue retention (NRR): starting MRR plus expansion, minus contraction and churn, divided by starting MRR, with new customers excluded. In my case top-line MRR grew 27 percent in a quarter while gross monthly revenue churn doubled from 2.7 percent to 5.7 percent and NRR fell from about 100 percent to about 95 percent. For SMB SaaS in 2026, median NRR sits near 97 to 100 percent (SaaS Capital, April 2026), so a slide toward 95 percent is a real warning, not noise.

The month the number lied to me

The first month I felt it was February. Nothing was wrong. New signups were healthy, a small annual deal had just landed, and MRR was up again.

But three cancellations came in the same week, and two of them were customers I would have bet money on. I told myself it was a coincidence. Churn is lumpy at $17K MRR. One bad week does not make a trend.

It was a trend. I just could not see it yet, because I was looking at the wrong line.

What net revenue retention actually measures

Most early founders watch MRR and gross new sales. Both are addition metrics. They tell you what came in. Neither tells you what stayed.

Net revenue retention isolates the existing book of business. You take the customers you already had at the start of the month, add their upgrades, subtract their downgrades and cancellations, and ignore every new logo you signed. If that number is above 100 percent, your existing customers are worth more this month than last, even before you sell anyone new. If it is below 100 percent, you are refilling a leaking bucket.

David Skok made this point years ago in his essay SaaS Metrics 2.0: churn is the silent killer, because a small monthly leak compounds into a ceiling on growth that no amount of new sales can push through. I had read it. I had even quoted it. I still missed it in my own numbers, because I never put NRR on the dashboard next to MRR.

The cohort table that finally showed it

In April I stopped looking at the top line and wrote out the monthly movement instead. New, expansion, contraction, churned. One row per month. It took twenty minutes and it ended a quarter of denial.

Read the Ending MRR column and everything is fine. Read the Churned column and the floor is on fire. The money leaving each month went from $440 to $1,150, roughly 2.6 times, in one quarter. New sales kept the ending number climbing, so the leak never showed up on the chart I actually looked at.

Expansion was quietly dying too. Upgrades fell from $610 to $430 while downgrades climbed. NRR crossed under 100 percent in February and kept sliding. My "best quarter ever" was a quarter in which my existing customers became worth less every month.

How SMB churn hides inside growth

The reason this is so easy to miss at the SMB end of SaaS is that the benchmarks give you cover. Median net revenue retention for bootstrapped SaaS companies with $3M to $20M ARR was 103 percent in 2026, and for the smallest SMB accounts it runs closer to 97 percent (SaaS Capital, April 2026). So when your NRR reads 98 percent, you can tell yourself you are roughly average and move on.

Average is not the same as safe. A company sitting at 97 percent NRR with flat new sales shrinks. Mine was masking a fast deterioration with a temporary run of new logos from a launch spike, exactly the trap that pushed one team to shut down a growing SaaS at $12K MRR because the churn was hiding inside the growth. I read their post-mortem the same month I built my table. It read like a letter to me.

The five things I changed

None of these are clever. All of them were things I should have been doing already.

1. NRR moved onto the main dashboard, next to MRR. Same size, same screen. If I only get to look at one chart before coffee, it is now the retention one.
2. I started reading the churn reason field. Half my cancellations shared one cause: a workflow gap two customers had asked me to fix in December that I had deprioritized for a shinier feature.
3. I split the cohort by how they arrived. Launch-spike signups churned nearly twice as fast as customers who found us through search. I had been optimizing acquisition for the worse cohort.
4. I built a save path before cancel. A one-question exit survey plus a real offer to downgrade instead of leave. It recovered a slice of the contraction I had been treating as inevitable.
5. I named the fork out loud. New features versus retention work is a real tradeoff, and I had been making it by default every sprint. Writing it on the wall forced an honest choice.

Are you default alive?

The frame that made all of this click was Paul Graham's Default Alive or Default Dead: given your current growth and burn, do you reach profitability before you run out of money, without raising again? A bootstrapper's version of the same question is simpler. If you never sold another new customer, would your existing book grow, hold, or shrink?

For four months my honest answer had drifted from "hold" to "shrink," and the top-line chart never told me. NRR did, the first time I bothered to look.

The one thing to take from this

Growth metrics and retention metrics answer different questions, and only one of them tells you whether the business underneath the chart is healthy. Put net revenue retention on the same screen as MRR, read it every month, and write out the movement table the first time your stomach disagrees with your dashboard. The leak is always cheaper to fix in February than in April.

I learned that the slow way, right around the same stretch I finally admitted I could not keep doing everything alone, which is its own diary about making a first hire at $9,120 MRR.

Keep reading

• Why we killed our SaaS at $12K MRR (a post-mortem) - the churn-hiding-inside-growth story that read like a letter to me.
• First hire at $9,120 MRR: a SaaS founder's diary (2026) - what I did once the numbers forced my hand.