The month cohort analysis exposed my real churn at $44K MRR: a founder diary (2026)

I asked myself that in 2026, sitting at $44,000 MRR, staring at a churn number that had barely moved in twelve months. On paper I was fine. In my gut I was not. It took me a full weekend building a cohort chart to understand that my dashboard had been telling me a flattering half-truth the whole time.

Quick answer (2026): This is a composite founder diary about running a proper cohort retention analysis at $44K MRR and discovering that my stable-looking blended churn was hiding newer customers who were leaving much faster than my old ones. Blended monthly churn averages your best and worst cohorts into one comforting number. When you split customers by the month they signed up and track each group separately, you can see whether your most recent cohorts are retaining worse, which is the difference between a business that is quietly healing and one that is quietly bleeding. The fix was not a churn-save campaign. It was learning to read the leading edge instead of the average.

The number that never moved

For most of a year my monthly logo churn sat at roughly 3.5%. It did not spike, it did not crater, it just sat there. I treated it like a thermostat reading: boring, stable, nothing to see. Revenue was climbing, so I let the flat churn line reassure me and looked at other things.

The problem with a blended churn number is that it is an average, and averages are where uncomfortable truths go to hide. My 3.5% was the blur of two very different realities stacked on top of each other. My oldest customers, the people who had been with me for two or three years, almost never left. They were sticky, they were happy, and they were quietly propping up the whole number. Underneath them, my newest customers were churning at nearly double that rate, and I could not see it because the loyal old guard kept dragging the average back down to comfortable.

I only noticed because the mood did not match the metric. New signups were up, revenue was up, and yet every month felt like running up a down escalator. That gap between what the dashboard said and what the business felt like is what finally made me build the chart I had been avoiding.

What a cohort actually is

A cohort is just a group of customers who signed up in the same window, usually the same month, tracked over time as its own line instead of being folded into the blended total.

Blended churn asks "what percentage of everyone left this month." Cohort retention asks a harder question: "of the people who joined in March, how many are still here in month one, month three, month six." You end up with a grid, one row per signup month, each row showing that group thinning out over time. Read down the columns and you can see something a single number can never show you: whether the March cohort is retaining better or worse than the January cohort, and whether that trend is improving or rotting as you grow.

The first time I built mine, the pattern was ugly and obvious. Every recent row started dropping faster than the rows above it. My cohorts were not stable. They were degrading, month over month, and the only reason the headline churn looked flat was that my old cohorts were so loyal they masked the decay at the front.

The distinction I had been sloppy about

Before I could act on the chart I had to get honest about two words I had been using interchangeably: retention and revenue.

Logo retention counts customers. Revenue retention counts dollars. They can point in opposite directions, and the gap between them is the whole game. You can lose a pile of small logos and still grow revenue if the accounts that stay expand. That is why the industry watches net revenue retention so closely. SaaS Capital's 2025 survey of private SaaS companies put median net revenue retention in the $25K to $50K ACV band at 102%, with the top quartile at 111% and the bottom at 97%, and it makes the point that net retention is always higher than gross retention because it folds in upgrades and expansion (SaaS Capital, 2025).

That framing stung, because I had been quietly leaning on it without earning it. My revenue retention looked passable only because a handful of growing accounts were expanding fast enough to paper over a rising tide of small-account churn. On a logo basis my newest cohorts were leaking badly. On a revenue basis the expansion masked it, exactly the way my loyal old customers masked the blended churn. Two different flattering averages, hiding the same problem.

The research that reframed it

Once I knew what to look for I read the people who study this at scale, instead of trusting one founder's weekend spreadsheet.

The most sobering read was ChartMogul's retention study of more than 2,500 SaaS businesses (ChartMogul SaaS Retention Report, 2021 to 2024 data). Two findings rearranged how I thought. First, retention is not a vanity metric, it is the growth engine: the median company holding net revenue retention at or above 100% grew about 48% year over year, more than twice as fast as companies below 100%. Second, and this is the part that scared me, retention gets structurally harder as you add customers. In their data only about 6% of companies with more than 12,000 subscribers held net revenue retention at or above 100%, while smaller companies cleared it far more often. Growth itself dilutes your retention if you are not watching the incoming cohorts.

The same report showed the mirror image of my problem. Companies at or above 100% net revenue retention get more than half their revenue from expanding existing customers, while the weakest performers lean on new business for around 70% of theirs, and companies with net revenue retention below 60% churn at roughly double the rate of the healthy group. I was drifting toward the wrong side of that split, running harder on new signups to outpace a leak I had refused to measure. I had lived the acute version of this once already, back in the month my churn doubled, but that was a visible spike I could react to. This was the slow kind, the kind that never trips an alarm.

What the cohorts actually told me

When I stopped looking at the average and started looking at the rows, the story got specific fast.

The degrading cohorts were not random. They clustered around a stretch where I had leaned hard into a cheap-to-acquire top-of-funnel push, chasing signup volume because the signup chart felt good. Those customers arrived with weaker intent, activated less, and left inside ninety days. My older cohorts, acquired slowly through word of mouth and content, retained beautifully. I had been averaging a healthy business and an unhealthy one together and reading the mean as truth.

A big slice of the front-edge leak was also involuntary, which is its own quiet tax. Failed cards, expired cards, and silent dunning failures were pushing customers out who had never actually decided to leave, and it hit my newest cohorts hardest because they had the least attachment to fight through a billing hiccup. I had written about that trap before in the month failed payments were eating my growth, but seeing it isolated inside specific cohorts made it undeniable rather than theoretical.

What I changed

I did not launch a win-back campaign. I changed what I measured and what I chased.

I killed the cheap top-of-funnel push, because a cohort that churns in ninety days is not growth, it is a loan against next quarter. I moved my retention dashboard from a single blended churn line to a cohort grid I look at weekly, and I set one rule: judge the business on the newest three cohorts, not the blended average, because the leading edge is where the future lives and the average is where it hides. I tightened onboarding for the segment that was leaking, and I fixed the boring, unglamorous dunning flow so involuntary churn stopped silently thinning my youngest customers.

None of it was dramatic. The drama was in finally seeing the thing I had been structurally blind to for a year.

What actually happened

Over the next quarter my blended churn barely changed, which is the last twist of the knife: the metric I used to trust would not have shown me any of this even after I fixed it. But the cohort grid moved. My newest cohorts started flattening into a plateau instead of sliding toward zero, which is the shape you actually want, a curve that levels off because a real core of customers has decided to stay.

MRR moved from $44K to about $46K over the quarter, but the number that mattered was quieter: the gap between my best and worst cohorts narrowed, and for the first time in a year the business felt like it was filling instead of leaking. The headline churn had been lying by omission the whole time.

The one thing I would tell you

Stop reading your churn as one number.

A blended churn rate is an average of your happiest and unhappiest customers, and it will stay comfortable long after your newest cohorts have started to rot. Build the cohort grid, split customers by the month they joined, and judge yourself on the leading edge, not the mean. Watch logo retention and revenue retention as two separate stories, because expansion can hide a logo leak exactly the way loyal veterans hide a churn problem. The day your dashboard and your gut disagree, believe your gut and go build the chart. Mine was right, and it took me a year and a weekend to admit it.