Definition
PARDR measures the trend in response time to predictive alerts over time.
The Measurement Gap
The Decision This KPI Enables:
Are building engineers responding to predictive and AI-generated alerts with the same urgency over time, or has alert fatigue from low-accuracy alerts degraded response behavior so that genuine critical alerts receive delayed or diminished attention?
Why Legacy Metrics Miss This:
Traditional alert response metrics capture average response time at a point in time. They do not track whether response times for critical alerts are degrading over time: whether the same type of alert that received a 15-minute response six months ago now receives a 4-hour response. Average response time across all alert types conceals the degradation because high-volume, low-priority BAS alerts (which engineers respond to quickly by acknowledging and dismissing) hold the average down while critical medical gas and life safety alerts stretch longer.
The degradation pattern is behavioral, not technical. The alert system works correctly: it generates the right alert at the right time. The monitoring infrastructure is functioning. The notification reaches the engineer. The engineer chooses to deprioritize the response because experience has taught them that most AI-generated alerts in this category are false positives or demand-driven misattributions. Each false positive reinforces the behavior. The degradation is invisible until a genuine alert with patient safety implications receives the same delayed response as the false positives that preceded it.
PARDR measures this behavioral degradation by tracking response time trends for critical alert categories over rolling periods. A rising trend in response time for medical gas alerts, even if the current average is within target, indicates that alert fatigue is eroding response discipline. The trend is the signal: by the time the average exceeds the target, the behavioral degradation has been underway for months and a patient safety event may have already occurred.
Formula
PARDR = ((RT_current − RT_baseline) / RT_baseline) × 100
Where:
- RT_current = Average response time for a critical alert category in the current measurement period (e.g., trailing 30 days)
- RT_baseline = Average response time for the same critical alert category during the baseline period (e.g., first 90 days after system deployment or last known "good" period)
Unit of Measure:
Percentage change from baseline (positive values indicate degradation, negative values indicate improvement)
Measurement Frequency:
Monthly, with rolling 30-day window; weekly monitoring for categories with recent degradation trends
Data Requirements:
- Alert log with timestamp of alert generation and timestamp of first response action (not just acknowledgment: actual investigation or dispatch)
- Alert categorization by system type (medical gas, fire/life safety, BAS comfort, BAS critical, predictive maintenance)
- Baseline period response times by alert category
- Alert investigation outcome (confirmed fault, demand-driven, false positive, nuisance) to correlate source accuracy with response degradation
- Engineer identification for individual response pattern analysis
Calculation Notes: Response time should be measured to first meaningful action, not to alert acknowledgment. Many organizations allow engineers to acknowledge alerts from their phone without investigating: acknowledgment time may remain constant while investigation time degrades. Separate PARDR calculation by alert category is essential; BAS comfort alert degradation is a nuisance, while medical gas alert degradation is a patient safety concern. The same PARDR value carries fundamentally different risk depending on the alert category.
Proposed formula, pending validation pilots.