Choose what a healthy customer does
Start with the customers who stayed longest and ask what they did in the first six months. They probably used the product or service on a rhythm, reached out when they had questions, brought colleagues in, paid on time and attended reviews.
Write down five or six behaviours that you can see in data. That list is your first set of signals. Do not start with fifty.
Pick signals in four groups
Cover each group with one or two signals.
- Usage: how often they log in or use the service, and whether that is rising or falling.
- Depth: how many features, users or projects are active.
- Relationship: whether they reply, attend meetings and have a named champion.
- Commercial and support: payment problems, open tickets, and how they feel about the replies.
Check that each signal is something you can get today without a project. A signal that needs a month of engineering will not happen.
Pull the data into one place
The simplest start is a spreadsheet in Google Sheets with one row per customer and one column per signal. Fill it weekly, by export or copy.
If you have a CRM such as HubSpot, you can build properties for each signal. Dedicated customer success tools such as Gainsight, Vitally or Custify can pull usage in directly. Start with the spreadsheet if you have fewer than a hundred accounts.
Give each signal a simple score
Turn each signal into green, amber or red with a clear rule, such as:
- Logins in the last 14 days: 3 or more is green, 1 or 2 is amber, none is red.
- Open support tickets older than a week: none is green, one is amber, two or more is red.
- Last meeting with the champion: within 60 days is green, 60 to 120 days is amber, longer is red.
Write the rules where everyone can read them. A rule you cannot explain to a customer is a rule you cannot defend.
Combine them into one number
Give each signal a weight. Usage and champion contact usually deserve the most. Add the points and map the total to an overall status. Keep the arithmetic so simple that a colleague can recalculate it on paper.
The first weights are a guess, and that is fine. You will correct them against real churn in Run quarterly health score reviews.
Set thresholds that trigger action
Decide in advance what each colour means for the team. Green: normal contact. Amber: the account owner checks in within two weeks. Red: the owner contacts the customer within five working days and tells a manager.
What to say and do for red accounts is the topic of Proactive outreach to at-risk accounts. Here the point is that the colour must lead to a task.
Show it where people already look
Add the status to the customer record, a weekly list in Slack, or the top of the Monday meeting. Show the trend as well as the colour. An account that went from green to amber in a fortnight is more urgent than one that has sat in amber for a year.
Common mistakes
- Tracking too many signals and trusting none.
- Using only logins, which hides unhappy customers who log in every day.
- Building a score nobody owns.
- Ignoring customers with no data at all. Silence is a signal.
- Treating the score as fact, not as a prompt to check.
How you know it works
Look back at the last ten customers who cancelled. Would your score have flagged most of them in the 90 days before? If yes, the signals work. If not, adjust the weights and try again. After a few weeks your team should be using the colours in conversations without being asked.
