Triage Scoring.
How AI scores feedback and how to customize the weights.
How Scoring Works
When feedback is submitted, the AI triage pipeline analyzes the content and assigns sub-scores across multiple dimensions. Each sub-score is a value from 0 to 1. The final priority score is a weighted average of these sub-scores, mapped to a 1–10 scale. Higher scores indicate feedback that requires more urgent attention.
Developer Dimensions
Developer projects score feedback on these 8 dimensions:
| Dimension | What It Measures |
|---|---|
actionability | How clearly the feedback describes something you can act on. |
specificity | Level of detail: reproduction steps, expected vs actual behavior. |
severity | Impact of the issue: crash, data loss, cosmetic, etc. |
frequency_signal | Whether the feedback suggests a widely experienced problem. |
sentiment_intensity | Strength of emotional tone, positive or negative. |
user_effort | How much effort the submitter invested in the report. |
technical_clarity | Quality of technical information: logs, error messages, stack traces. |
scope_estimate | Estimated scope of the fix: quick patch vs architectural change. |
Business Dimensions
Business projects use the same base dimensions minus technical_clarity and scope_estimate, and add three business-specific dimensions:
| Dimension | What It Measures |
|---|---|
revenue_proximity | How directly the feedback relates to revenue-impacting activities. |
public_visibility_risk | Risk of the issue becoming a public complaint or negative review. |
customer_retention_signal | Indicators of whether the customer is likely to churn without intervention. |
Score Anchors
Use these anchors to interpret the final 1–10 priority score:
| Score | Label | Meaning |
|---|---|---|
| 9 – 10 | Drop everything | Critical issue requiring immediate action. |
| 7 – 8 | Prioritize this week | Important feedback that should be addressed soon. |
| 5 – 6 | Worth reading | Useful feedback to consider during planning. |
| 3 – 4 | Background noise | Low-priority items. Review when you have time. |
| 1 – 2 | Noise / Spam | Not actionable. Likely spam, test data, or empty submissions. |
Custom Weights
You can customize the weight of each scoring dimension via the API to match your priorities. Weights are decimal values that determine how much each dimension contributes to the final score. Higher weights give more influence to that dimension.
PATCH /projects/:id/scoring-weights
Content-Type: application/json
x-grova-key: gv_your_project_key
{
"weights": {
"severity": 2.0,
"actionability": 1.5,
"frequency_signal": 1.2,
"specificity": 1.0,
"sentiment_intensity": 0.8,
"user_effort": 0.5
}
}Omitted dimensions keep their default weight of 1.0. Changes apply to all future feedback. Existing scores are not recalculated automatically.