What I Do in the First 48 Hours After a Health Score Alert

When a health score alert fires, I do three things within 48 hours. I check that the drop is real, I find the reason before I contact anyone, and I agree one next step with the customer, with a date. A health score tells you which account needs attention, but the outcome depends on what happens in those first two days.

In the last edition I covered signals that point to growth. This one covers the opposite: signals that point to risk.

Why do alerts get ignored?

Usually because nobody's name is on them.

I saw this on a partner team that had a well-built score and a dashboard nobody opened. Red accounts sat for weeks because everyone assumed someone else had them. The score was fine. The fix was putting one name next to every alert and a deadline next to every name. After that, the same alerts started getting answered.

What do I check in the first hour?

Before I write to anyone, I confirm that the drop is real:

• Data: did a tracking change, an integration outage or a renamed field cause it?
• Calendar: is it a holiday, a fiscal close or a quiet season for this customer?
• People: did the main contact change role or leave?
• Support: are there open critical tickets or a recent escalation?

When the first two checks explain the drop, I close the alert with a short note on the cause. That note matters later, because it shows which signals keep creating noise.

How do I find the reason before calling?

A walkthrough with illustrative numbers: an account's score drops from 78 to 60 in 14 days, weekly active users fall by 35%, and renewal is 80 days away.

I don't call to say "your score dropped". First I look at which users stopped logging in. In a case like this it is often one team, not the whole account. Then I check that team's tickets. If they hit a bug two weeks ago and gave up, I have my reason. My first message names it: "I saw your operations team stopped using the booking export after the issue on the 3rd. It's fixed now. Can I take 20 minutes with them to get them back on it?"

A specific message gets a reply. A generic check-in usually doesn't.

How do I close the loop?

Every alert ends with one of three outcomes, logged on the account:

Outcome	What I record	Next step
False alarm	Cause (data, season, other)	Adjust the trigger if it repeats
Recovered	Reason and action taken	Recheck the score in 30 days
Still at risk	Reason and what is blocking	Escalate to the team lead with the renewal date

That log feeds the monthly review. The full set of playbooks, with owners and response windows, is in Customer Health Score Playbooks: Turn Alerts Into Action, and how to set the alert thresholds has its own article. You can test your own weights with the health score calculator, and you can track whether recovered accounts actually renew with the renewal rate calculator.

Try this this week

• Open your red accounts and check that each one has a named owner and a date.
• For the next five alerts, run the four checks before contacting anyone.
• Write each first message around one specific reason, not the score.
• Log every alert as a false alarm, recovered or still at risk, and count them at the end of the month.

In the next edition: how to tell whether a retention playbook actually saved an account, or whether it would have renewed anyway.