Five signals, one consensus
Five independent signals cross-check every outcome's metadata. The engineer and the manager see the same signals at the same time. The goal isn't catching liars — it's making sure a record is uncontested before it counts.
Single-number evaluation systems collapse to a single number
When self-reported time is the only signal, it gets inflated. When token counts are the only signal, they get hidden. RUQA uses five independent signals so no single one carries the entire weight — and so disagreement between signals becomes its own information.
Self-report
What the engineer says happened
AI estimate
Reasonable time given the work
Mechanical signals
Git timestamps, API sessions
Volume regression
Time predicted from output volume
Peer median
Team baseline for similar work
How signals combine
Median of T2-T5 anchors the value. T1 is compared to that anchor. If T1 deviates by more than 30%, it's flagged for manager review — never silently.
Trust, but verify
RUQA is not a surveillance tool. The goal is to make evaluation fair — protecting honest engineers from being out-gamed by less honest ones, and giving managers ground truth they can defend.
- Right to know about every flag, instantly
- Right to dispute (council review)
- Right to view your full data
- Calibration period (90 days, no consequences)
- Algorithm transparency (this page exists)
- ×Keystroke logging
- ×Screen recording
- ×DM monitoring
- ×Off-hours surveillance
- ×Algorithm-only termination