Churn shows up in the data first

Most subscription cancellations look sudden from the outside. The email arrives, the account closes, and the first question is always the same: did we see this coming? Almost every time, the answer is yes, but the data was not in front of the right person.

A cancellation is the end of a story that usually begins six to eight weeks earlier with a change in behavior. Not a dramatic one. The account does not slam the door. It goes quiet.

Five signals that show up before the exit

Each signal alone can mean many things. Together, they almost always mean the same thing.

• Sign-in recency stretching. The account signs in once a week, then once every ten days, then not at all for three weeks. The gap grows. Nobody at the company is sitting down to evaluate the product on those missing days. They stopped coming back.
• Usage trend bending down. Feature use drops month over month. The new workflow they were building is abandoned partway through. Usage breadth shrinks. The platform is shrinking to one person doing one thing.
• Seats going quiet. The account has twelve seats and eight were active last quarter. Now three are. The decision-maker who approved the contract may not notice that the people doing the work have moved on.
• A spike of support tickets. Not one thoughtful question: multiple tickets over a few weeks, some of them repeats. Confused users are burning out. When a team is struggling to make a product work and the support does not land well, they stop trying.
• A failed payment. Often the first mechanical signal and often the last one before cancellation. By the time a payment fails, the account has usually already decided. The billing card was not updated because updating it for a product you are leaving is nobody's priority.

Why teams miss them

The reason these signals go unnoticed is not negligence. It is that they live in different places. Sign-ins are in the analytics tool. Tickets are in the support system. Seats are in the billing account. Payment failures are in Stripe. Nobody's Monday morning job is to pull all five, line them up against a book of two thousand accounts, and decide which six deserve a call today.

The Monday morning job is responding to the accounts that are making noise: the renewal that is overdue, the expansion call that got scheduled, the support escalation that arrived in the inbox. The quiet accounts, the ones that are slipping without saying so, do not make it onto the list until they are gone.

An account that churned with six weeks of visible warning

A mid-market software company that uses Churnora today lost an account worth around forty thousand in ARR last year, before they came on, that they still talk about. Not because it was the biggest loss, but because of what they found when they went back through the data afterward.

The account had dropped from four active users to one over a six-week period. Sign-ins stretched from daily to weekly to once in three weeks. Two support tickets in month four, neither resolved cleanly. Month five, one payment failed and recovered. Month six, the cancellation request arrived.

Six weeks of signal, in data the company owned. The account executive had not called. Nobody had flagged it. There was no system that looked at sign-in recency alongside seat activity alongside support volume and put the account on a list.

The data had the story. Nobody had read it.

Maya Kessler

The argument for ranking attention

Dashboards do not save accounts. Looking at a dashboard tells you what is happening. A ranked list tells you where to start. The difference is that one requires a person to make a judgment every time they open it, and the other has already made the judgment for them.

When a customer-success team scores every account by churn risk and works the list from the top, the accounts with the clearest warning signals get the call before the cancellation. Not because anyone was faster or smarter. Because the system was pointed in the right direction.

Churnora ranks accounts by the behavioral model's estimate of churn probability, not by which account made noise most recently. The quiet ones at the top of the list are there for a reason. That is the whole point.

6-8 wk

how far ahead of most cancellations the behavioral signals typically appear in usage data

5

behavioral patterns, from sign-in recency to payment failures, that appear most consistently ahead of a cancellation

1 in 4

at-risk customers who can typically be saved when someone reaches out before the cancellation request arrives