A customer health score is a structured summary of evidence about an account, not a prediction that the customer will renew. Start with signals your team can explain, record when they were observed and keep missing data visible. The worksheet below combines an illustrative score with evidence coverage and a separate next-action review.
Decide what the score should help someone do
Use the worksheet to prioritize a conversation, investigate a change or review an outcome. Do not begin by choosing colors. Define the account segment and lifecycle stage first: a new implementation and a mature customer should not be judged by the same expected behavior. Name an accountable person who will review exceptions.
Keep this page distinct from a customer success plan. A plan records shared goals and work. A health worksheet evaluates the evidence available about the account at a point in time. Link the two, but do not treat a completed internal task as proof that the customer achieved its goal.
Choose a small set of explainable signals
Useful inputs include progress toward the customer’s stated outcome, adoption of the relevant workflow, relationship continuity and unresolved support impact. Define each input precisely: which people, which period, which source and which threshold. “Lots of activity” is not a definition.
Gainsight’s documentation treats customer health as a scorecard with measures and history. That is useful context for an explainable model, but the weights and thresholds below are an original illustrative worksheet, not a vendor recommendation or a validated churn model.
Illustrative signal definitions
| Signal and weight | Evidence to record | Common mistake |
|---|---|---|
| Outcome progress, 35% | Customer-confirmed milestone and dated review | Using your own delivery activity as the customer’s outcome |
| Relevant adoption, 30% | Defined users and workflow over a matched period | Counting every login as valuable usage |
| Relationship continuity, 20% | Confirmed champion, sponsor and open decisions | Treating a friendly reply as budget authority |
| Support impact, 15% | Current unresolved impact and operational source | Treating missing support data as no issues |
Calculate the score and its evidence coverage
Assign an illustrative 0–100 value only when the input meets its evidence definition. Multiply each known value by its weight, sum those contributions and divide by the total known weight. Report the known-weight percentage beside the score. Never fill an unknown input with zero or with the account’s previous favorable value.
Worked example: outcome progress is 80, adoption is 60 and relationship continuity is 75. Support impact is unknown. The weighted contributions are 28, 18 and 15, totaling 61. Known weight is 85%, so the illustrative score is 61 ÷ 0.85 = 71.8 out of 100, with 85% input-weight coverage. It is not a 71.8% renewal probability. The unknown support input still needs review.
Keep critical facts outside the average
A weighted score can hide a serious unresolved issue. Maintain explicit review flags for a departing sponsor, a blocked core workflow, an unverified outcome or a material commercial concern. The reviewer decides the next action based on the evidence, even when the composite score looks favorable.
Choose a freshness rule for each signal and segment. When evidence expires, mark it stale or unknown; do not preserve a green status indefinitely. Compare the account with its own prior periods using the same definitions before interpreting a change.
Turn the worksheet into a useful weekly review
Review changed signals, stale evidence and the next action, not every field on every account. Each action needs an owner, an agreed date and evidence of completion. A task created is not a task completed, and a message sent is not a customer outcome.
Use Finta to retain the relationship context and follow-up. Ask Aurora to summarize approved account material and identify missing evidence. Product-usage, support and renewal inputs may live in other systems; this worksheet does not imply Finta natively collects or calculates every measure.
Weekly review record
| Question | Answer to capture | Next step |
|---|---|---|
| What changed? | Signal, old and new period, source | Check whether definitions remained constant |
| What is unknown? | Missing input or stale evidence | Assign an evidence owner |
| What needs a conversation? | Specific outcome or risk question | Prepare a reviewed account follow-up |
| Did the action help? | Customer-confirmed result, if available | Update the evidence, not just the color |
Try it with Aurora, Claude or Codex
Copy this prompt and add only material you are authorized to use. In Aurora, use the context available in your Finta workspace. Other assistants need the relevant inputs supplied separately.
Help me review this customer-health worksheet from the evidence I provide.
Use the stated signal definitions and weights.
Calculate known-input score and input-weight coverage separately.
Keep unknown and stale inputs explicit.
Flag material risks that an average could hide.
Return evidence gaps, questions and next actions with owners to confirm.
Do not predict renewal or invent product-usage data.Review the result against the original sources before acting. A generated draft is not an approval or a completed external action.
Sources and method
Sources reviewed October 28, 2026. Examples and worksheets are original editorial teaching tools, not customer results.
Gainsight scorecard definitions and history.