Remediation Learning Rate (RLR)

GMI-SWD-014Software DeliveryIntegration-RequiredPattern O-10: Deferral Compounding

Definition

Rate at which auto-remediation events convert into permanent fixes rather than silently recurring interventions, measuring whether automation is resolving problems or hiding them.

The Measurement Gap

Measurement Gap and Decision

The Gap: MTTR is excellent by design since automation resolves faster than humans. Incident count may be low because auto-remediated events aren't logged as "incidents." Availability is high because automation keeps services up. Every reliability metric says you're improving while organizational understanding of failure modes stagnates. Change failure rate (DORA) doesn't measure whether failures produce learning. SLA compliance measures uptime, not capability to prevent future failures.

The Decision:

Is automated remediation creating a "fix and forget" culture, or is there a functioning feedback loop from automated fixes to systemic improvements? The metric tells you whether your automation pipeline operates in pure symptom-suppression mode or produces organizational learning.

What Changes When You Have This Number:

Above 0.40, the team is converting a significant fraction of automated actions into systemic improvements. Between 0.15–0.40, learning happens for severe events but routine remediation produces no improvement. Below 0.15, mandate a lightweight root-cause review for every Nth automated remediation (e.g., every 5th OOM restart requires a 30-minute investigation). The intervention is structural: add learning triggers to automation workflows rather than hoping teams investigate voluntarily.

Formula

Formula:

RLR = Auto_remediations_with_learning / Total_auto_remediations

Where Auto_remediations_with_learning is the count of automated remediation events producing a substantive learning artifact within N days (14 recommended), and Total_auto_remediations is the total count in the period.

Sign Convention:

Higher is better. 1.0 = every automated remediation produces learning. 0.0 = pure symptom suppression.

Worked Example:

Over Q3, automated remediation triggered 67 times:

  • Pod restarts (OOM detection): 23
  • Circuit breaker activations: 18
  • Auto-rollbacks (canary failure): 12
  • Auto-scaling events: 9
  • Certificate rotations: 5

Within 14 days, learning artifacts produced for 27 events:

  • 8 OOM restarts → memory profiling → code fixes merged
  • 4 circuit breakers → dependency team conversations → timeouts tuned
  • 7 auto-rollbacks → post-mortems → specific fixes and test additions
  • 5 latency events → capacity planning → scaling thresholds adjusted
  • 3 certificate rotations → monitoring additions

40 events had no follow-up investigation.

RLR = 27 / 67 = 0.403

Learning Artifact Definition:

Must be substantive. "Restarted pod, service recovered" is NOT a learning artifact. "Identified allocation in OrderService.processRefund() creating unbounded list on high-volume days; fix PR #2847 merged" IS a learning artifact.

Proposed formula, pending validation pilots.

At a Glance

Unit
Rate
Trend Direction
Higher
Reporting Frequency
Quarterly
Category
Operational Resilience
← Orchestration Depth Impact Index (ODII)Rollback Cascade Coherence Index (RCCI) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment