Health Scoring
Part of Customer health scoring
Combining usage and relationship signals carefully
Combine product use and customer relationship evidence while keeping missing data, stale assessments and conflicting signals visible.
Combine usage and relationship signals only after defining what each one tells you. Product records can show whether the intended work happened in the product. Conversations can reveal whether the work still matters, who owns it and what prevents a useful result. Preserve both views beside any combined score so one favourable measure cannot conceal a serious concern in the other.
Put the signals on comparable footing
Choose the account unit, eligible users and observation period first. Usage from one division should not be silently combined with a relationship assessment for the entire organisation. Identify the task the customer expects to perform and give it enough time to occur. A quiet week has a different meaning for a daily task and a quarterly one.
For each usage measure, specify the event, completion condition, account mapping and exclusions for test or automated activity. For each relationship measure, record the person consulted, their role, what they confirmed, when they said it and when the assessment must be revisited. A sponsor's attendance and a process owner's confirmation of value answer different questions.
How to Combine Usage and Relationship Signals Safely
- Define the account unit and observation periodEnsure usage and relationship data align at the same level (e.g., division vs organisation) and timeframe (daily vs quarterly).
- Specify event conditions and exclusionsSet clear criteria for what counts as completed usage (e.g., successful workflow execution, excluding test accounts).
- Record relationship assessments with metadataDocument who was consulted, their role, confirmation statement, date, and review date.
- Keep observations separate from judgementsNote factual customer statements separately from interpretations about health impact.
Keep observation and judgement distinct
An account owner may hear that a critical handoff fails despite steady product use. Record that customer statement as a dated observation. Label the owner's judgement about its effect on account health separately, with a reason and review date.
Likewise, a drop in recorded actions could reflect a tracking change or a task that did not arise. Verify the cause before calling it disengagement.
A two-view account note can help:
| Usage view | Relationship view | First question |
|---|---|---|
| Expected work is completed | Customer confirms the result is useful | Are both observations current and about the same task? |
| Expected work is completed | Customer reports that the result fails its purpose | Which part of the outcome is missing? |
| Completion has fallen | Customer says the work is paused by agreement | Does the expected-use period need updating? |
| Completion has fallen | The relevant owner cannot be reached | Are the task, contact and data records still current? |
These are investigation prompts, not automatic health ratings. The customer's explanation may change the interpretation of the product record.
Usage vs Relationship Signals in Customer Health Scoring
- Usage ViewMeasures whether the expected work occurred in the product (e.g., workflow completion, event tracking).
- Relationship ViewCaptures customer confirmation of usefulness, ownership, and barriers to outcome (e.g., process owner feedback, audit team concerns).
- Key DifferenceUsage is observable and data-driven; relationship is subjective and context-dependent.
Test the combination rule
If the team needs one number, document the scale and weight of each component. Check whether two inputs count the same event twice. For example, workflow completion and a usage-based engagement measure may both reward the same activity. Consider whether a material issue should remain visible as a separate flag even when the weighted average is favourable.
Decide how unavailable inputs behave. Gainsight NXT documentation describes excluding measures from the overall score by assigning zero percentage weight. Check the rule in the system actually used, and show which components were present, which were missing and when each was last refreshed. A relationship assessment from months ago should not appear current merely because product events updated yesterday.
Consider a hypothetical account where the core workflow is completed regularly, but the customer's process owner says the resulting records cannot be used by its audit team. The usage observation is favourable; the customer-confirmed outcome is not. Present both and assign investigation of the record problem. No numerical weight can resolve that disagreement by itself.
Pros and Cons of Combining Usage and Relationship Signals
- ProsProvides a holistic view of customer health by combining objective product behaviour with subjective business context.
- ConsRisk of masking critical issues if one signal overrides or obscures the other; stale or duplicate data can distort results.
Key Data Points for Health Score Validation
- Last refreshed
- Product events: yesterday
- Last relationship assessment
- Months ago
- Weight assigned to relationship signal
- Zero percentage (excluded)
Review disagreements as useful evidence
When the two views diverge, check the particular workflow, stakeholder and period involved. Record what the customer confirmed and what remains an inference. If the data was wrong, correct the component and note why. If the data was right but incomplete, retain both views and the resulting action.
Revisit the combination rule when the product, customer workflow or available data changes. The account view should make it possible to see when the total is misleading.
