Churned MRR

Monthly subscription revenue always looks respectable as a single figure. Last month was $30,000, this month is $31,500, so the month added $1,500 and reads as a good one. That one line says nothing about the state of the business.

The same $1,500 comes out of two completely different months. In one, new customers brought in $2,300 and departures took away $800. In the other, new customers brought in $7,000 and departures took away $5,500. The second company's revenue drops the moment it stops advertising, but on a table that records only the total, the two months are identical.

The difference surfaces constantly in investor meetings. A team that arrives with a growth rate and can't answer "how much went out?" doesn't get to the next part of the conversation, because a growth rate alone can't distinguish a sound bucket from one with a hole in it being filled fast.

Churned MRR is the field that records the outbound side on its own. Monthly recurring revenue (MRR) — the sum of what is contracted to arrive again each month — gets split into what came in and what went out, and the outbound figure is this number.

Take only the customers who were in last month's closing recurring revenue and add up everything that disappeared from them this month. Four steps.

1. Freeze the population to last month's closing list. A customer who first paid this month isn't in it. If that customer leaves within the month, the movement gets handled on the new MRR side instead.
2. Add up customers who left entirely, at the amount they were paying last month. One account at $460 a month plus three at $230 comes to $1,150.
3. Add up the amounts that shrank. This covers a downgrade, where the contract survives on a cheaper plan, and contraction, where the plan holds but the seat count falls. Dropping from 80 seats to 50 and losing $350 counts as $350, exactly as a full cancellation at that amount would.
4. Add the two and record one positive number. The example comes to $1,500. The rule that matters is not netting this against the inbound side first. Recording zero churn because $1,500 of expansion arrived erases two facts at once.

Two policies have to be settled in advance so every month gets counted the same way.

The first is one-time money. Implementation fees and training fees don't arrive again next month, so they were never recurring revenue, and the end of that work doesn't belong in churned MRR.

The second is failed payments. A card that expired and cut the subscription off isn't a customer deciding to leave, and most of those accounts recover within days. Counting them immediately makes the number swing every month, so the usual practice is a retry window of about two weeks, with only the accounts that don't recover in it counted.

The amount by itself separates nothing, because churned MRR grows naturally as the company grows. Divide it by last month's closing recurring revenue instead. Losing $1,500 out of $30,000 is 5 percent a month.

The benchmark depends heavily on who the customer is. Products bought directly by individuals and small teams commonly run 3 to 5 percent a month. Products sold on annual enterprise contracts aim for under 1 percent a month: contract terms leave fewer exit points, and a single departure carries so much revenue that the same percentage hurts far more.

One thing is worth reading in absolute terms — the distribution behind it. A month where $1,500 leaked evenly out of forty small accounts and a month where it left with one enterprise account carry the same figure but demand different work. The first points at the product or the pricing and gets worked on in batches; the second sends one person to find out what happened at that single account.

Set it beside expansion MRR and one more thing appears. When the money added by customers who stayed exceeds the money lost to customers who left, revenue grows even in a month with no new customers at all. A company in that state can cut marketing spend without falling over, which is why investors ask for the two figures separately rather than the net.

"Net new is positive this month, but churned MRR has grown two months running."

Total revenue is up, but the leak is widening and deserves attention now.

"Does churned MRR include the failed payments? I think we need to look at it without them."

Check whether recoverable amounts got mixed in, so the figure reflects customers who actually left.

"Count the downgrade deltas in churned MRR too. Cancellations alone aren't the whole number."

Revenue lost while the contract survives is still revenue lost.

At Pagedeck, Grace opens this table in the first week of every month. New, expansion and churned sit side by side, and the first thing she looks at is not the total but the churned column over the last two or three months. When it's rising, the follow-up questions are always the same: which accounts it came from, and why those accounts went. The total gets read last, because it can stay flat while both sides double.

Last quarter half of the amount came from one enterprise account that cut from 80 seats to 50. Sofia checked, and a whole department there had been folded into another team. The conclusion was that the product wasn't at fault — but having no way to see that change coming stayed on the list as an open problem.

The most common mistake is assuming new revenue covers it. Losing $1,500 while gaining $2,300 nets out positive, so nothing looks wrong. But acquiring a customer costs advertising and sales salary, while keeping one costs far less. The same $800 is much cheaper to hold onto than to replace. Paper over churn with new sales and each month needs more spend to produce the same growth rate, and revenue falls the moment the spend stops.

The other is counting cancellations while ignoring shrinkage, because a steady account count looks like nobody left. But if 100 customers are still 100 customers while total seats fall from 4,000 to 3,200, customer churn is 0 percent and revenue is down 20 percent. Leave that running for a few months and the table shows nothing happening while revenue keeps sliding.