The 90-Day Renewal Health Check: Catching the Churn Before It's a Save Call

A save call is a call you lost the right to avoid.

I have run them. Six weeks out, the renewal date is on a calendar somewhere in procurement, the champion has gone quiet, and you are calling to ask a customer to re-decide something they have already quietly decided. Every lever you have at that point is a discount, a favor, or a promise about the next twelve months. None of those are strong. You are negotiating from behind and both sides know it.

The fix is not a better save call. The fix is a check you run ninety days out, on every account, whether or not it feels healthy. It takes about twenty minutes per account if your data is in one place and an hour if it isn't. It is boring. It is also the single highest-leverage hour I know of in an account management week, because the thing you find at ninety days is almost always fixable, and the same thing found at forty-five days is a negotiation.

This is the renewal risk checklist I run, what each red flag actually means, and the specific call to make the week you find it.

Why ninety days and not sixty

Ninety days is not a magic number. It is the shortest window that still lets you do something real.

If a customer has stopped using a workflow, you need time to get them back on it and time for that usage to show up in the data. That is not a two-week project. If your champion left, you need to find and build a new one, and building a champion means getting somebody a win they can take to their own boss. Wins take a quarter. If the metric you sold against has gone the wrong way, you need time to either fix it or reframe it honestly, and reframing it honestly requires a conversation you would rather not have and then a second conversation where you show progress.

At sixty days you can do one of those things. At forty-five you can do none of them, which is why the six-week call is a save call and not a renewal call. By then your only remaining product is price.

The other reason for ninety is budget. In a lot of companies the person who can protect your line item is doing next year's planning three months before your renewal date, not six weeks before it. If you show up after that planning is done, you are asking someone to reopen a spreadsheet they already sent upward. Almost nobody wants to do that for you.

The five checks

Run them in this order. The first two catch most of what is going to kill you.

1. Usage trend against the last three months, not against onboarding

This is the one that gets measured wrong more than anything else, and it is the reason healthy-looking dashboards produce surprise churn.

Most usage reporting compares current activity to some baseline set during implementation, or to a target that was written into the business case. That comparison flatters you. Onboarding usage is inflated — you had a project team, a deadline, a training session, and somebody's boss watching. Of course logins were high. Comparing today to that is like judging a gym membership by January.

What you want is the shape of the last three months against the three before it. Not a number. A direction. Is this account doing more of the thing it bought the product to do, the same amount, or less?

And be specific about which thing. Total logins is a vanity metric. If you sold a reporting tool, the question is whether reports are being built and shared, not whether people opened the tab. If you sold a workflow tool, the question is whether records are moving through the workflow to completion. Pick the one action that means the customer got what they paid for and trend that.

Red flag: the core action is down against the prior quarter, even slightly, and especially if it is down while headcount at the account is flat or up.

The play: you are not calling to ask why usage is down. That question puts the customer in the position of defending themselves, and the honest answer — we stopped bothering — is embarrassing for them to say out loud. You call to ask what changed in the process. "When we set this up in March, Dana's team was running the intake through the queue every morning. I looked at the last quarter and it looks like that shifted. Did the process move somewhere else, or did the team change?" That is a neutral question about their operation, and people will answer it. Usually the answer is that someone left, a different tool absorbed the step, or a reorg moved the work to a team that never got trained.

All three are fixable in ninety days. None are fixable in thirty.

2. Does your champion still hold the same job

Not "is my champion still at the company." Is my champion still in the same seat, with the same scope, reporting to the same person.

I check this on LinkedIn before I check it in the CRM, because the CRM tells you what someone typed in six months ago and LinkedIn tells you what happened last week. Look for a title change, a team change, a new manager, or a company reorg announcement. Then look at whoever signed the contract, if that was a different person. Then look at whoever owns the budget line, if that was a third person.

A champion who got promoted is not automatically good news. Promotions move people away from the day-to-day problem you solved. The new scope is bigger, your tool is now one of forty things in their portfolio, and the person who inherited their old job has no emotional stake in a decision they didn't make.

Red flag: any change at all in the champion's role, or a new manager above them, or a new person occupying the seat where the daily user used to sit.

The play: congratulate the promotion and ask for the handoff in the same call. "Congratulations — genuinely. Two things. Who picked up the queue work, and would it help if I ran them through what we built with you so they aren't starting cold?" You are offering to do free onboarding for the new person, which is a favor to your old champion, and it gets you a first meeting with the person who will actually decide the renewal.

If the champion is gone entirely, do not wait to be introduced. Find the two or three people who use the product most and call them directly. Not the manager first. The users, because they can tell you in five minutes whether the product still matters, and a manager cannot.

3. How many people from the buying committee have logged in

Go back to the original deal. Write down every name that was in the room, on the thread, or in the security review. Usually four to eight people. Now check which of them have touched the product in the last quarter.

The answer is often one. Sometimes zero.

This matters because a renewal is not decided by the daily user. It is decided by people who formed an impression of your product at some point and have not updated it since. If the VP who approved the purchase has not logged in since the pilot, their opinion of you is frozen at whatever they thought during the pilot, plus whatever they have heard in hallways since. You have no idea what that is, and you have no way to influence it from a save call.

Red flag: fewer than two or three of the original committee have any recent contact with the product or with you. Single-threaded accounts renew fine until the single thread breaks, and then they don't renew at all.

The play: manufacture a reason for the non-users to see value without asking them to log in. A one-page summary of what their team accomplished with the product this year, sent to the exec who signed, with an offer of fifteen minutes to walk through it. Not a QBR deck. One page, their numbers, their team's names. Most executives will not take the meeting. That is fine — the point is that they read the page, and the frozen impression thaws slightly. Some of them take the meeting, and those are the ones who were about to cut you.

4. What happened to the metric you sold against

Every deal was sold against something. Time to close a ticket. Cost per load. Days in accounts receivable. Hours of documentation per clinician per shift. Whatever it was, it is written in your notes from the discovery call, or it should be.

Go find out what that number is doing now. Not what your product's dashboard says about it — what the customer's own version of it says. These are frequently different, and the customer believes theirs.

This is the hardest of the five checks because it usually requires asking, and asking risks getting an answer you don't like. Ask anyway. At ninety days a bad answer is a project. At six weeks it is an obituary.

Red flag: the metric got worse, the metric is no longer tracked, or nobody at the account can tell you what it is. That third one is the most common and the most dangerous. A metric nobody measures is a benefit nobody can defend in a budget meeting.

The play: if the number got worse for reasons unrelated to you — volume tripled, a plant closed, the department lost half its staff — get that context on the record with your champion, in writing, in their words. That sentence is the thing that saves the renewal later. If the number got worse partly because of you, say so first, before they do, and bring a plan with a date on it. Owning a problem ninety days out reads as competence. Owning it six weeks out reads as a defense.

If nobody is tracking it anymore, your job for the next quarter is to make somebody track it again. Rebuild the report. Send it monthly. Put a human name on it.

5. Has anyone asked you a support question in eight weeks

The last check is the cheapest and I like it the most. Open the support queue and your own inbox. When did anyone at this account last ask you or your team a question?

Silence is not health. Customers who are getting value ask questions, because using something seriously generates friction, and friction generates tickets. Accounts that go completely quiet are usually not quietly succeeding. They are quietly not using it.

There is a version of quiet that is genuinely fine — a mature deployment that has been stable for two years and needs nothing. You will know the difference by cross-referencing with check one. Quiet plus flat-to-rising usage is fine. Quiet plus declining usage is a countdown.

Red flag: no inbound of any kind in eight weeks, from anyone, while the core action is trending down.

The play: do not send a check-in email. "Just checking in" produces nothing. Send something with a specific, small, useful thing in it — a setting they don't have turned on, a report their peer company built, a bug you fixed that was annoying them last spring. The goal is to get a reply, because a reply reopens a channel, and an open channel at ninety days is how you find out what is actually going on.

What to do with the results

Don't build a score. I have seen teams turn this into a weighted health index with a color, and the color becomes the artifact instead of the action. What matters is the count and the pairing.

One red flag on its own is usually noise. A champion changed jobs but usage is up and the team asks you questions every week — you have work to do, not a crisis. Two red flags together is a real risk, and the combinations that scare me most are champion change plus usage decline, and metric-nobody-tracks plus single-threaded committee. Three or more and you should assume the renewal is already lost internally and you simply haven't been told. Escalate it that week. Tell your manager. Get a second person on the account.

Then write the play down with a date and a name. Not "re-engage the account." "Call Marcus Thursday, ask who inherited the intake queue, offer to train them the week of the 14th." Renewal risk work dies when it stays abstract.

When the check comes back bad anyway

Sometimes you run this at ninety days and the account is already gone. The champion left in March, you found out in September, usage has been flat on the floor for two quarters. That happens, and when it does you are running a save call whether you like it or not.

Run it properly. The structure I use for a software account that has drifted is in the SaaS renewal call script for a save call you're six weeks from losing, and it is built around getting the customer to say the honest thing out loud before you start proposing anything. If you are in logistics, where the account usually stopped using the visibility layer because a broker relationship changed rather than because your product failed, the freight and 3PL version handles that specific conversation. And when the renewal is against a firm that had a genuinely bad year and is cutting everything with a line item, the professional services script is the one that deals with the budget-first objection without collapsing to a discount in the first four minutes.

Use them. But notice what all three have in common: they are recovery documents. Every one of them starts from a position you would rather not be in. The health check exists so you need them less often.

The honest argument for doing this

Nobody gets promoted for a renewal that closed uneventfully. That is the whole problem. Save calls are visible, dramatic, and easy to get credit for. The quiet quarter where you noticed a champion had changed jobs in July and had a new one by September produces no story at all.

Do it anyway, and do it on a calendar rather than on instinct. Instinct checks the accounts you already worry about. The renewals that hurt come from the accounts nobody worried about, where the usage was fine last time anyone looked and the last time anyone looked was February.

Block two hours every Monday. Pull the accounts renewing in ninety to a hundred days. Run the five checks. Make the calls that week.

If you want to get sharper at the calls themselves — the champion handoff ask, the "what changed in your process" question that doesn't make someone defensive, the exec one-pager follow-up — that is exactly the kind of thing we built DrillCall for, running the same awkward opening thirty times against a customer who pushes back until it stops sounding like a script. What I would do this week, though, is simpler than that. Open your renewal list, find the four accounts closing in the next quarter, and check whether your champion still holds the same job. That one takes ten minutes and it will tell you more than your CRM has told you all year.