How Risk Threshold Hysteresis Works
Up: Risk Thresholds & Hysteresis See also:
Definition
Risk Threshold Hysteresis works by establishing two distinct limits: an Entry Threshold (to enter a penalty state) and a stricter Exit Threshold (to leave it). Once a rule is broken, the account must "over-correct" to return to normal standing.
Why it matters
It creates a "trap." If a monitoring program limit is 0.9% disputes, you enter at 0.91%. But to exit, the rule might require <0.6% for 3 months. Simply returning to 0.8% is insufficient.
Signals to monitor
- Current Metric (Real-time rate)
- Target Threshold (Exit requirement)
- Probation Counter (Clean months accrued)
- Cohort Performance (New vs Legacy traffic risk)
- Burn-Down Rate
Breakdown modes
- "The Almost Reset" (Spiking in month 3 of probation, resetting counter to zero)
- "Permanent Flagging" (Stuck in entry/exit loop)
- Misaligned expectations (Aiming for Entry threshold instead of Exit threshold)
Implementation notes
Observability should visualize the path to the Exit threshold and track the "Gap" between current reality and required performance. Key focus is sustained stability.