The renewal was lost in month two. Nobody noticed until month eleven.

Here is a thing I have watched happen more than once. The renewal forecast is green. The account is marked healthy. Everyone is relaxed. Then, in month eleven, the customer does not renew, and the room is genuinely surprised.

They should not have been. The renewal was decided months earlier. Nobody was reading the signals that said so.

Churn is a lagging number. By the time it lands in a report, the decision is old. The customer stopped using it around month three. The champion who fought to buy you left in month five, and the person who inherited the account never wanted it. The one thing that made you worth the money was never switched on. None of that showed up in the renewal forecast, because the renewal forecast was a feeling wearing the costume of a number.

Health scores measure vibes, then act surprised

Most customer health scores are built backwards. Someone picks a few inputs that feel right - a satisfaction survey, a support ticket count, whether the account manager likes the client - weights them by instinct, and calls it a health score.

Then the green accounts churn and the red accounts renew, and everyone quietly stops trusting the score.

A health score is only worth building if it predicts. Which means building it from the behaviours that actually preceded a renewal, and the ones that actually preceded a churn, in your own data - not a vendor template. Adoption depth. Breadth of real users, not seats sold. Time to the first real outcome. Whether the person who championed you is still there. Those are boring. They are also the ones that move before the renewal does.

Build the score from what happened, not from what feels important. Then test it against last year: would it have caught the accounts you actually lost? If not, it is decoration.

The renewal is won in onboarding

The highest-leverage moment in the whole customer lifecycle is the first thirty days. Not the quarterly review. Not the renewal call. Onboarding.

A customer who reaches a real outcome early builds a habit and a story - a reason they can repeat to their own boss for why this was a good decision. A customer who does not gets a thing they paid for and never quite used, and eleven months later that unused thing is the easiest line item to cut.

Most onboarding is a tour. It shows people the buttons. It does not get them to the one outcome that makes the rest matter. The gap between those two is most of your retention rate.

Expansion is the same job, done again

Net revenue retention - the number investors actually care about - does not come from a clever upsell sequence. It comes from solving more of the customer's problem than you did last quarter.

Which is the same job as keeping them: understand what they are actually trying to do, and remove the next obstacle. The teams that expand best are usually solving their customer's customer's problem, not mirroring the buyer's data back at them. Expansion is not a motion you bolt on. It is retention that kept going.

None of it works on dirty data

All of it - the health score, the onboarding trigger, the expansion signal - runs on the record. If the record is wrong, the signals are wrong, and you are back to guessing with extra steps.

Data hygiene is the first domino here, not a back-office chore. A customer success team working off a CRM nobody trusts is a team that finds out about churn from the churn.

Retention feels like a lagging problem because the number arrives late. It is not. Every renewal is decided in the quiet months - in onboarding, in adoption, in whether the right person is still in the room. Read those, and the renewal call is a formality.

Miss them, and you will keep being surprised by a decision the customer made in the spring.