A short one from me. I build GTM systems and write up what actually worked, receipts only. This one is about the half of the customer relationship almost nobody measures: the goodbye.
- The moat moved. Software is commoditized. Features get copied in a quarter and pricing converges, so the only durable edge left is how you make people feel. That includes the exit.
- Everyone onboards, nobody offboards. The welcome is a five-star hotel. The goodbye is a parking ticket.
- Churn is a list, not a wall. A third of your churned accounts were still getting value the month they left. That is a boomerang list, and it is invisible until you look.
- Win-back is re-onboarding. The customers who come back do not come back for a discount. They come back because the door was left open.
The reflex, and the scar
When churn ticks up, everyone reaches for the same move: a better cancel-flow save offer, a bigger win-back discount, one more retention play bolted onto the exit. Reduce the leak. An SDR leader hears "pipeline is soft" and adds activity. A marketer hears "MQLs are down" and buys more traffic. Same reflex, different seat: when a number moves the wrong way, add more of the thing that was already not working.
Here is my scar. At a company I was at, mid-market SaaS, we had boomerang customers constantly, people who churned and came back months later. For a long time I filed it under luck and moved on. It cost me, because I was busy optimizing the save offer at the exit while the thing that actually brought people back was something I never measured and never designed.
The reframe: score the goodbye
The move is not "reduce the leak." It is "score the goodbye." Churn is a loud symptom. Drill past it and the real thing is quiet and specific.
A customer who leaves feeling respected is a boomerang. A customer who leaves feeling trapped is gone for good.
Same move, other seats. A CSM drills "churn is up" down to which accounts left with real usage and which health signal missed them. A marketer drills "the brand feels transactional" down to the exact moment a customer stops feeling chosen and starts feeling processed, which is almost always the cancel flow nobody designed.
What the operators say
I am not the first to argue the experience is the lever. But watch where the smartest people on retention stop.
Retention is the single most important thing to get right, and improving onboarding, the new-user experience, is often a bigger lever on retention than improving the product itself.
Retention comes from connecting users to real product value fast, and the durable version of it is built as loops, not one-time pushes.
Onboarding is not done to prevent churn, it is done to get the customer to their desired outcome, and retention follows. And the exit matters: customers should leave on a good note, because leaving now is not leaving forever.
The method: solve, stack, split
Here is the workflow, worded off you and the work that matters, not off "reps" and "seats." Every step is a question you are already asking.
Not "the cancel flow leaks." The real problem is you never score the exit and you have no way to tell a winnable churn from a gone one. Score the welcome on feeling, not activation, and score the exit on one question: bridge, or bridge burned.
The right signals, not a shopping trip: your product analytics for who churned while still active, your CRM for the exit history and the champion, and the two audit rubrics as the brief. The CSM version reads adoption; the AE version reads the deal history; the marketer version reads the funnel the customer walked to the door.
The scoring is production work, and it comes off your plate. The audits grade every moment, the boomerang-list read finds who left with value on the table, and none of it needs a human until there is a decision to make.
The win-back conversation stays human, and it should. Timing the reach-back to a real change, bringing back the person who knew them, and giving an honest goodbye instead of a desperation discount is judgment, not automation. One owner, one weekly review of who came back and who did not.
The receipt
I ran the offboarding audit on a standard cancel flow: email support to cancel, a 40-percent-off save offer, a five-day processing delay, then delete the account. It scored about 1.2 out of 5, illustrative but not far off. The burned bridge was the ending: deleting the config so a returning customer starts from zero, and a discount that tells them the price was never real. The fix was not complicated. One honest cancel button, a real human goodbye for accounts over a threshold, keep the config warm for 90 days, schedule the reach-back.
The bigger receipt is what the other end looks like when you take it seriously. At that same company, making retention proactive, an early churn-and-expansion read plus an exit that did not burn bridges, is what moved us from bleeding out every month to a real annual profile, gross retention at 100 percent and net revenue retention at 125 percent. Not from a better discount. From treating the whole arc as designed.
Drop your own cancel flow in and score it. You will find the cold spot faster than you expect.
- The win-back does not start when they come back. It starts with how you let them leave.
- Price is almost never the real reason. "Too expensive" usually means value not felt, which is usually an onboarding that never finished.
- You cannot fix what you never scored. The exit had no number, so nobody owned it, so it stayed a cancel button instead of a channel.
The move this week
Score your offboarding. One honest walk through your own cancel flow, graded bridge or bridge burned, whether your number is a funnel or a lifecycle. If you greet customers like a five-star hotel and dismiss them like a parking ticket, the gap between those two scores is the most honest retention metric you have.
Two places to go next. The skill that grades the exit: offboarding-audit, which "scores the moment most companies treat as a cancel button." And the proof it is real, not a vibe: every skill and play we ship is read against the full text of the skills it chains and run on a real scenario before it earns the badge.