Definition
Percentage of SON-initiated parameter changes that are reversed or counter-adjusted by the same or a neighboring cell within a defined window, indicating optimization churn rather than convergence.
The Measurement Gap
Decision Supported: Is the SON producing stable, beneficial network configurations, or is it oscillating between competing local optima without net improvement? Legacy RAN metrics (call drop rate, throughput, spectral efficiency) report the outputs of optimization but cannot reveal whether the optimizer is churning through parameter changes that cancel each other out. A cell cluster can show acceptable throughput while the SON reverses 30% of its own changes within hours: consuming signaling resources, creating transient quality dips during each reconfiguration, and never reaching a stable configuration. NOOR makes this invisible churn visible.
Measurement Gap: SON platforms from SON platform, Nokia NSP, and Samsung log every parameter change, but no standard metric tracks reversals. Network performance dashboards show post-optimization outcomes but not the path, a cluster that reaches 95% of optimal through three stable changes is operationally different from one that reaches 95% through thirty changes and twenty-seven reversals. The reversal pattern reveals whether the SON's optimization model is converging or trapped in a local-optimum cycle, a distinction no existing telecom KPI captures (Hamalainen et al., 2012; 3GPP TS 32.500 on SON management).
Formula
Formula (Plain Language): Count every SON parameter change event in the measurement period. Then count how many of those were followed by a reversal: defined as the same parameter on the same cell reverting within N hours, or the same parameter on an adjacent cell moving in the opposing direction within N hours. Divide reversed-or-counter-adjusted changes by total changes.
Notation:
NOOR = Reversed_or_counter_adjusted_changes / Total_SON_parameter_changes x 100
Variable Definitions:
- Reversed_or_counter_adjusted_changes: SON parameter change events where the same parameter on the same cell reverts within N hours, OR the same parameter on an adjacent cell moves in the opposing direction within N hours. Unit: count, dimensionless.
- Total_SON_parameter_changes: All SON-initiated parameter modifications in the measurement period. Unit: count, dimensionless.
- N: Oscillation detection window. Proposed starting point: 4 hours for intra-cell reversals, 8 hours for inter-cell counter-adjustments. Adjust based on the SON platform's execution interval and the specific parameter type. Faster-cycling parameters (CIO, handover thresholds) may warrant shorter windows; slower parameters (antenna tilt) may warrant longer windows.
Sign Convention: Lower is better. 0% means every change was durable. Rising NOOR means the SON is churning.
Edge Cases: Legitimate rapid adjustments during declared mass events (concerts, emergencies, sports events) should be excluded or segmented separately. The SON may correctly cycle parameters during these periods. Parameter changes with multiple valid states that naturally cycle with traffic patterns (e.g., tilt adjustments for day/night traffic distribution) should be distinguished from reactive counter-adjustments using the SON's change-reason metadata. For cells with very low change volume, aggregate to cluster or market level for statistical meaning; proposed starting point for minimum sample is 50 parameter changes per measurement period per cluster.
Proposed formula, pending validation pilots.