What it is.
It pulls deals from HubSpot, collects linked support tickets and feature usage from a Google Sheet, runs sentiment analysis on the tickets to produce a customer health score, then has an AI step weigh deal age, sentiment, and usage trend against thresholds you set.
When risk crosses the line it emails the person who owns the account, with the data and a suggested next step attached. Each of those three signals is weak on its own and useful together.
What you get.
- A customer health score built from CRM, support, and usage data in one pass.
- Sentiment analysis over the ticket text, which carries more than a ticket count.
- Risk evaluation against thresholds you define for deal age, sentiment, and usage.
- An email to the responsible owner carrying the data and a suggested next step.
How to set it up.
Connect HubSpot and confirm which deal pipeline and stages you want watched.
Build the Google Sheet with support tickets and feature usage in the columns the template expects.
Check the sentiment step against tickets you have already read, because sentiment scoring is the piece most likely to be wrong.
Set your own thresholds for deal age, sentiment, and usage decline; the defaults are the author's.
Point the alert at the named account owner, so somebody owns each one.
Run it against last quarter's churned accounts first, and see whether it would have caught them.
Use cases
Health score from what you already have
Build one score per account from the data you already collect.
Read the ticket sentiment
Let the analysis pass over the ticket text nobody has time to read.
Give CS a call list
Sort the scores and hand over the accounts worth a conversation this week.
Best for
Accounts you only hear from at renewal
Silence reads as health until it does not, and this watches the signals that move first.
Teams without a customer success platform
A health score from HubSpot plus a sheet covers most of what a dedicated tool would show.
Testing a churn hypothesis
Backtesting against accounts that already left tells you whether your thresholds mean anything.