Definition
Percentage of historical fiber failures that occurred through degradation modes the predictive model can structurally observe (OTDR-detectable degradation signatures), versus failure modes the model cannot predict because they produce no advance signal in the model's data sources. PMORS defines the model's theoretical detection ceiling: its maximum possible prediction rate.
The Measurement Gap
Fiber predictive models are evaluated against total fiber failures, but a significant fraction of fiber failures occur through mechanisms that produce no advance signal in the model's data sources (third-party dig-ins, vehicle strikes, vandalism, sudden catastrophic events). Evaluating the model against total failures sets an unfair and misleading performance target. A model with 50% overall prediction rate might be performing near its theoretical ceiling if 45% of failures are inherently unpredictable by its data sources. Without quantifying the observational boundary, the organization cannot determine whether the model is underperforming or whether it is performing near its structural limit and the unpredicted failures require a different risk mitigation strategy entirely.
The Decision
What fraction of fiber failures is the predictive model capable of predicting, and for the unpredictable fraction, what alternative risk mitigation is needed? PMORS reframes how the organization evaluates the predictive model's value: a model with 50% overall prediction rate and PMORS of 0.55 is performing at 91% of its theoretical ceiling (50/55). A model with 50% rate and PMORS of 0.9 is underperforming badly (50/90 = 56% of detectable failures caught). The distinction drives completely different management actions.
Research Foundation
Analysis of fiber failure root causes across major North American networks shows that 30-45% of fiber cuts result from external damage events (dig-ins, vehicle strikes, construction accidents) that produce no detectable precursor signal in OTDR monitoring, while 40-55% result from degradation modes with measurable advance signals (splice aging, connector contamination, macro-bending from sustained stress, fiber aging) (Fiber Optic Association Technical Bulletin: Root Cause Analysis of Fiber Plant Failures, 2024). Predictive fiber analytics vendors report detection rates of 40-60% against total failures, but when evaluated against only detectable-mode failures, effective detection rates rise to 70-85% (Light Reading: Fiber Predictive Analytics Market Assessment, 2023).
Formula
PMORS = Failures_from_detectable_modes / Total_failures
Variable Definitions (every variable resolved)
- Failures_from_detectable_modes = historical fiber failures with root causes that produce observable advance signals in the model's data sources (e.g., splice degradation, fiber aging, connector contamination, macro-bending from sustained mechanical stress) (count)
- Total_failures = all fiber failures in the historical analysis period (count)
- Detectable mode classification = determined by engineering analysis: does this failure mode produce an OTDR signature, splice loss trend, or other measurable precursor before failure?
Units:
Ratio, 0 to 1. 0.6 means the model can theoretically predict at most 60% of failures: the other 40% are inherently unpredictable by its data sources.
Edge Cases
Failure modes with ambiguous detectability (rodent damage that sometimes shows as slow attenuation before the cut and sometimes is sudden): classify as "partially detectable" with an estimated detection fraction. Failure causes that are unknown or disputed: classify as "indeterminate" and exclude from both numerator and denominator. PMORS is a ceiling, not an accuracy measure: the model's actual prediction rate will be below PMORS because it will not catch every detectable-mode failure. Minimum: analyze a minimum of 200 historical fiber failures for a meaningful PMORS estimate.
These are proposed starting points: calibrate to your operational context during pilot: PMORS above 0.6 means the model can theoretically predict the majority of failures. Below 0.4 means most failures are outside the model's reach, and alternative risk mitigation (physical patrols, one-call coordination, route hardening) is the primary defense. Evaluate the model's actual prediction rate as a percentage of PMORS to determine true performance.
Proposed formula, pending validation pilots.