IoT Signal Degradation Index (IOTDI)

GMI-INS-014InsuranceInfrastructure-Dependent

Definition

Estimated percentage of loss-predictive signal in IoT sensor data that is destroyed by privacy-mandated aggregation, anonymization, or retention policies before the data reaches the actuarial or loss prediction model.

The Measurement Gap

Insurance carriers are investing heavily in IoT data for risk assessment and pricing. Water flow sensors, smart home devices, telematics, commercial property monitors, and wearables all generate granular data streams that can predict loss events. But between the sensor and the actuarial model sits a privacy compliance pipeline that aggregates, anonymizes, and truncates the data to meet retention policies.

The sensor works fine. The data pipeline destroys the signal. A water flow sensor that reports per-minute readings might be aggregated to daily averages before the data reaches the pricing model. A telematics device recording per-second driving behavior might be summarized to trip-level scores. Each aggregation step removes information that could predict losses. The model never sees what was lost.

Legacy metrics evaluate IoT model performance against the data the model receives. They measure how well the model uses the aggregated data it has. They do not measure how much predictive power was lost before the data arrived. A model can perform well on degraded data while being structurally incapable of reaching the performance it would achieve on granular data.

ISDI closes this gap by comparing model performance on raw granular data against the same model architecture trained on privacy-processed data. The performance difference quantifies the signal destroyed by privacy processing, making the cost of compliance data processing visible and actionable.

Decision Supported:

How much actuarial value is being lost to privacy compliance data processing, and are there privacy-preserving alternatives (differential privacy, federated learning, edge computing) that could retain predictive signal without violating data retention policies?

Formula

Plain Language: Build two models on the same historical IoT data: one using the raw granular data before privacy processing, one using the data as it arrives after privacy-mandated aggregation. Compare their predictive performance. The gap is the signal destroyed by privacy processing.

Formula:

ISDI = 1 - (AUC_aggregated / AUC_granular)

Variables:

  • AUC_granular: area under the ROC curve for the loss prediction model trained on raw, granular IoT data (dimensionless, 0.5 to 1.0)
  • AUC_aggregated: area under the ROC curve for the same model architecture trained on privacy-processed (aggregated/anonymized) data (dimensionless, 0.5 to 1.0)
  • ISDI of 0 means no signal lost (privacy processing did not degrade predictions). ISDI of 0.20 means 20% of the model's discriminative power was lost to privacy processing.

Units:

Ratio (0 to 1). Higher values indicate greater signal loss.

Data Requirements:

Data Element Source System Classification
Raw IoT sensor data (pre-aggregation) IoT platform in research sandbox NOT CAPTURED (requires controlled access or prospective collection)
Aggregated IoT data (post-privacy processing) Data warehouse / actuarial data feed NATIVE
Loss outcomes for IoT-monitored risks Claims management system NATIVE
Privacy/retention policy documentation Compliance / data governance documentation MANUAL

Proposed formula, pending validation pilots.

At a Glance

Unit
Index (0 to 1)
Trend Direction
Lower; approaching 0
Reporting Frequency
Annually
Category
AI Assessment Accuracy -- Vision and IoT
← Damage Estimation Segment Bias (DESB)Loss Control Intervention Effectiveness (LCIE) →

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