Role Mining Stage 3
Discover repeatable access patterns before you turn them into candidate roles.
Stage 3 turns the clean Stage 2 footprint into pattern seeds. Compare cohort stability, shared entitlements, and confidence so you can separate real job-based access from narrow exceptions or privileged noise before role generation starts.
Discovery in progress
Pattern discovery metrics will update as the Stage 3 queue is rendered.
Pattern Signal Matrix
Measure which access bundles repeat with enough cohesion, cohort size, and separation to survive downstream candidate review.
Signal pulse
In queue
0
Need calibration
0
0%
of pattern seeds in the current queue are already strong enough for candidate generation.
Pattern signal commentary will appear here once the discovery queue is rendered.
A strong discovery queue mixes repeatable access bundles with clear business anchors. The goal is not to force every cluster into a role, but to identify the few seeds that stay stable when reviewed against risk, ownership, and separation of duties.
Priority Pattern Seeds
Rank the most promising Stage 3 clusters by cohort strength, shared-access depth, and candidate readiness.
Priority queue
Pattern seeds most likely to survive candidate generation
The seeds below are ordered by confidence first, then by how much shared access they preserve once baseline noise has been removed.
Top seeds will be highlighted here when the table is rendered.
| Pattern | Cohort | Shared Access | Confidence | Risk / State | Next Step |
|---|
Analyst Checklist
Quick pattern-quality checks before a seed graduates into a role candidate.
- Confirm the business anchor Every strong seed should map back to a real job family, operating process, or approval lane that an analyst can explain.
- Trim privileged overlap early Clusters carrying administrative or sensitive access may still be valid, but they need tighter scoping before candidate generation.
- Watch for thin cohorts High-confidence clusters with only a few people can be real, but they often need more sampling before they become reusable roles.
- Preserve separability If two seeds share most of the same access, split them by business signal now or they will collapse into noisy candidates later.
Pattern Quality Mix
See how much of the current queue is ready to promote, still under review, or needs more calibration.
A healthy queue has a few obvious promotions, several review-worthy seeds, and only a small number of clusters that still need major recalibration.
Cohort Spread
Compare cluster size against shared-access depth so thin but privileged patterns do not outrank broader, more reusable bundles.
- Large and shallow Wide cohorts with little shared access often indicate remaining baseline noise rather than a role-worthy pattern.
- Focused and deep Smaller cohorts that keep a dense shared core are usually the best seeds to inspect before candidate generation.