Definition
Percentage of fiber predictive model corrective action recommendations where the recommended action addressed the actual degradation source, measured by whether the corrective action resolved the degradation trend or whether a second intervention was needed to address the true source. CADA isolates the model's diagnostic accuracy from the technician's execution quality. If the technician followed the model's recommendation and the trend did not resolve, the model's diagnosis was wrong.
The Measurement Gap
Fiber predictive models detect degradation trends and recommend corrective actions ("re-splice at location X," "clean connector at patch panel Y"). When the recommendation is executed and the degradation appears to stop, the model gets credit. But nobody systematically checks whether the trend actually resolved or whether it resumed weeks later and required a different corrective action at the actual degradation source. Without post-action verification of trend resolution, the model's diagnostic accuracy is assumed rather than measured. A model that correctly detects degradation but misattributes it to the wrong component produces unnecessary work (the first intervention), extended degradation (the real source remains unaddressed), and a false sense of predictive maintenance effectiveness.
The Decision
When the fiber predictive model says "re-splice at location X," is location X actually the degradation source, or is the model systematically misattributing symptoms to the wrong component? If CADA is low for a degradation type, technicians should independently verify the model's diagnosis before executing the recommended action, converting blind execution into informed execution. The model may be excellent at detecting degradation but poor at localizing it.
Research Foundation
Fiber predictive models demonstrate significantly different diagnostic accuracy by degradation type, with splice failure localization achieving 80-90% accuracy (clear OTDR signature) while macro-bending localization achieves only 40-60% accuracy (OTDR signature overlaps with splice degradation) (IEEE/OSA Journal of Optical Communications and Networking: Diagnostic Accuracy in Predictive Fiber Analytics, 2024). Field validation studies found that 25% of predictive model corrective actions in fiber networks required a follow-up intervention at a different location, with the model's initial recommendation addressing a correlated symptom rather than the root cause (Corning Fiber Reliability Conference Proceedings, 2023).
Formula
CADA = Recommendations_that_resolved_degradation / Total_executed_recommendations × 100
Variable Definitions (every variable resolved)
- Recommendations_that_resolved_degradation = predictive model corrective actions where the degradation trend stopped or reversed within N days of execution and no follow-up action was needed (count)
- Total_executed_recommendations = all predictive model corrective actions that were executed (count)
- N = resolution verification window (proposed starting point: 30 days for fiber, 14 days for connector/patch panel issues)
Units:
Percentage. 100% = every recommendation addressed the right component. Lower = the model is misattributing degradation sources.
Edge Cases
Degradation with multiple contributing sources (splice AND macro-bending AND connector contamination): the recommended action may resolve one source while others persist; if the trend improves but does not fully resolve, classify as "partial resolution" and track separately. Corrective actions that resolve the trend but the trend was going to resolve on its own (temperature-dependent loss that seasonally reverses): attempt to control for seasonal/environmental patterns. Corrective actions where the technician deviated from the model's recommendation: exclude. These test the technician's judgment, not the model's diagnosis. Minimum sample: 30 executed recommendations per failure mode category per measurement period.
These are proposed starting points: calibrate to your operational context during pilot: CADA above 80% means the model is usually pointing at the right component. Below 60% means technicians should independently verify the model's diagnosis before executing the recommended action. For any degradation type with CADA below 70%, require independent diagnostic testing before executing model-recommended corrective actions.
Proposed formula, pending validation pilots.