Risk Score Concentration Bias Index (RCBI)

GMI-TEL-045TelecommunicationsIntegration-Required

Definition

Degree to which the AI risk-scoring model's maintenance and capital replacement resource allocation is proportional to the economic risk across all physical asset classes, specifically identifying asset classes with disproportionately low attention relative to their failure consequence. RCBI reports both a per-class ratio (1.0 = proportional) and a portfolio score (0 to 1, where 1.0 = perfectly proportional allocation).

The Measurement Gap

AI risk-scoring models prioritize maintenance and capital replacement based on the data available to them. Asset classes with rich sensor data and abundant failure history (batteries, rectifiers) dominate the model's attention because they produce the strongest risk signals. Asset classes with sparse data (standby generators, structural steel, outside plant cabinets) receive lower risk scores not because they are lower risk but because the model has less information to assess their risk. The result is systematic concentration bias: resources flow to data-rich classes while data-sparse classes accumulate unmanaged risk that only becomes visible when an unplanned failure occurs in a class that was never in the model's priority queue.

The Decision

Is the risk model allocating capital and maintenance resources based on actual risk, or is it concentrating on data-rich asset classes while data-sparse classes accumulate unmanaged risk? The most dangerous finding is asset classes with RCBI near zero AND high failure consequence: for example, standby generators that power critical sites but are rarely included in the predictive model's portfolio because they run infrequently and produce minimal telemetry.

Research Foundation

AI-driven asset prioritization models in infrastructure management demonstrate systematic bias toward asset classes with higher data density, allocating 2-3x more resources per unit of economic risk to data-rich classes compared to data-sparse classes with equivalent failure consequences (McKinsey Global Infrastructure Initiative: AI in Asset Management, 2024). Capital allocation studies in telecom physical infrastructure found that 68% of predictive maintenance budgets were consumed by battery and power system maintenance, while structural and environmental assets representing 35% of portfolio economic risk received 12% of predictive maintenance attention (CTIA Network Infrastructure Report, 2023).

Formula

RCBI_c = Maintenance_share_c / Risk_share_c

RCBI_portfolio = 1 − Σ |RCBI_c − 1.0| / (2 × N_classes)

Variable Definitions (every variable resolved)

  • Maintenance_share_c = percentage of total maintenance/capital replacement budget allocated to asset class c (percentage)
  • Risk_share_c = percentage of total portfolio economic risk represented by asset class c, computed as failure_probability_c × failure_consequence_c × installed_base_c / total_portfolio_risk (percentage)
  • N_classes = number of asset classes in the portfolio (count)
  • c = asset class (towers, batteries, rectifiers, generators, fiber plant, outside plant cabinets, HVAC, small cells, microwave dishes, etc.)

Units:

Per-class ratio (1.0 = proportional allocation) and portfolio score (0 to 1, where 1.0 = perfectly proportional). Report per-class to identify the specific neglected classes.

Edge Cases

Asset classes with zero maintenance/capital allocation and non-zero risk: RCBI_c = 0. These are the most dangerous findings. Risk estimation for data-sparse classes relies on engineering judgment or manufacturer data rather than historical failure rates; flag the confidence level of the risk estimate. New asset classes (recently deployed small cells, new technology generations) with no failure history: assign conservative risk estimates and flag as "data-sparse. Risk estimate is conservative." Minimum: compute across all asset classes with 10+ installed units.

These are proposed starting points: calibrate to your operational context during pilot: RCBI_portfolio above 0.7 indicates reasonably proportional allocation. Below 0.5 means systematic concentration bias. Individual RCBI_c below 0.3 flags a severely neglected class. Any asset class with RCBI_c below 0.3 AND failure consequence rated high or critical should receive a mandatory minimum capital allocation regardless of the risk model's scoring.

Proposed formula, pending validation pilots.

At a Glance

Unit
Composite; Ratio per class; portfolio score (0 to 1)
Trend Direction
Toward 1.0; proportional
Reporting Frequency
Quarterly
Category
Asset Risk Scoring and Capital Allocation Quality
← Cost Assumption Currency Score (CACS)Experience-Proxy Divergence Index (EPDI) →

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