Definition
SCCI measures the percentage of sensors that directly feed AI decision systems that are currently within their calibration validity window, weighted by the number of AI systems consuming each sensor's data and the consequence severity of decisions those AI systems make, creating a single index that reflects the calibration health of the AI system's data inputs rather than the facility sensor fleet as a whole.
The Measurement Gap
Sensor maintenance programs track calibration dates for the full sensor fleet (temperature, humidity, power, pressure, flow) with no distinction between sensors that feed human dashboards and sensors that feed AI decision systems. A temperature sensor used only for historical trending that drifts out of calibration has very different implications than a sensor whose readings directly drive AI cooling setpoint decisions. Standard sensor calibration KPIs (percentage of sensors in calibration, overdue calibration count) measure the overall fleet without the consequence-weighted view that AI dependency requires. An AI-optimized facility can have 95% of sensors in calibration while the 5% out of calibration are precisely the sensors the cooling optimizer relies on most, creating silent accuracy degradation in the AI system's most critical inputs.
Decision Supported:
SCCI determines whether the sensors that AI decision systems depend on are being maintained to a calibration standard consistent with the consequence of errors in AI decisions based on those sensors, rather than a uniform standard that treats all sensors as equivalent.
Formula
Plain language: Build a list of all sensors that feed AI decision systems. For each sensor, check whether it is within its calibration validity window. Assign a weight based on how many AI systems consume it and how consequential errors in those systems are. SCCI is the weighted percentage of those sensors currently in calibration.
Notation:
SCCI = Σ_s (w_s × In_calibration_s) / Σ_s w_s
Variables:
| Variable | Definition | Unit |
|---|---|---|
| s | Sensor index: only sensors in the AI-critical sensor inventory (those whose data feeds one or more AI decision systems) | n/a |
| In_calibration_s | Binary: 1 if sensor s is within its defined calibration interval (last calibration date + calibration interval ≥ today), 0 if overdue | Binary |
| w_s | Weight for sensor s: (Number of AI systems consuming this sensor's data) × (Decision consequence score 1–5) × (Drift sensitivity score 1–3) | Dimensionless weight |
The AI-critical sensor inventory must be explicitly maintained: it is not the full sensor fleet but the subset whose readings drive AI decisions rather than just human monitoring. Default calibration interval to 12 months when manufacturer specification is not available. Auto-calibrating sensors still require periodic verification of calibration function. Sensor replacement resets the calibration clock to the replacement date.
Proposed formula, pending validation pilots.