Confidence Scoring
Every Execution in frugally.app receives a confidence score — a value from 0 to 100 that indicates how reliably the Execution is expected to achieve its goal. Higher scores mean greater confidence that the right resources will be acted on successfully.
What the confidence score represents
The confidence score is a composite metric that considers multiple factors about the Target, its resources, and historical performance. It answers the question: "How confident is frugally.app that this Execution will do what you expect?"
| Score range | Label | Meaning |
|---|---|---|
| 80–100 | High | Strong resource match, good historical success, well-configured Target |
| 50–79 | Medium | Likely to succeed but some factors could be improved |
| 0–49 | Low | Significant risk of unexpected results — review before scheduling |
Scoring factors
The confidence score is calculated from the following factors:
Resource match quality
How well the Target's filters match the intended resources.
| Signal | Positive impact | Negative impact |
|---|---|---|
| Tag coverage | All matched resources have the expected tags | Resources matched by broad filters (e.g. no tags, region-only) |
| Resource count stability | Consistent count across recent scans | Count fluctuates significantly between scans |
| Service type match | Resources are the expected type and size | Mixed resource types that may behave differently |
Historical success rate
How often previous Executions against this Target have succeeded.
| Signal | Positive impact | Negative impact |
|---|---|---|
| Past success rate | 90%+ Executions succeeded fully | Frequent partial failures or full failures |
| Consistency | Results are predictable across runs | Inconsistent outcomes (some runs succeed, others fail) |
| Recency | Recent Executions succeeded | Last few Executions had issues |
Configuration completeness
How well the Target and its associated resources are configured.
| Signal | Positive impact | Negative impact |
|---|---|---|
| IAM permissions | All required permissions are in place | Missing or outdated IAM policies |
| Connection health | All linked Connections are healthy | One or more Connections have issues |
| Tag specificity | Filters use specific key:value pairs | Filters are overly broad (key-only or no tags) |
Where confidence scores are displayed
| Location | Detail shown |
|---|---|
| Execution detail page | Full score with per-factor breakdown |
| Executions list | Score badge (High / Medium / Low) |
| Onboarding Wizard | Per-recommendation score at the review step |
| Slack notifications | Score included in Execution result messages |
| Dry run results | Score for the simulated Execution |
[SCREENSHOT: confidence-score-breakdown.png -- Execution detail showing confidence score with factor breakdown]
How to improve a low score
| Factor | Action |
|---|---|
| Low tag coverage | Add specific tags to your AWS resources and update the Target's tag filters |
| Fluctuating resource count | Investigate why resources are being created/destroyed — consider more specific filters |
| Low historical success rate | Review failed Executions for error patterns — often an IAM or connectivity issue |
| Broad filters | Narrow your Target's filters with additional tag key:value pairs |
| Unhealthy Connection | Check the Connection status and update the IAM role if needed |
A confidence score below 50 does not prevent an Execution from running. It is an advisory signal to help you prioritise which Targets to review and improve. Running a dry run is a safe way to investigate before taking action.
Confidence scores and the Onboarding Wizard
The Onboarding Wizard uses confidence scores to rank its Target recommendations. The Select All High Confidence button selects only recommendations with a score of 80 or above, ensuring your initial setup starts with the most reliable Targets.