Definition
Ratio of avoided event cost to preemptive action cost across prediction-triggered interventions, measuring whether acting on the model predictions creates net economic value.
The Measurement Gap
The measurement gap is straightforward: no existing water utility metric captures the net economics of prediction-driven action. Utilities report prediction accuracy (did the model get it right?) and action counts (how many work orders did the model generate?) but never connect the two to answer the fundamental question: is acting on predictions saving more money than it costs? Without PVR, a utility cannot distinguish between a predictive program that returns $3 for every $1 spent and one that burns money on false positives faster than it saves on true positives. Both programs look identical in a prediction accuracy report.
Formula
Layer 1: Plain Language
For every preemptive action triggered by a prediction, estimate two things: what the action cost (crew time, materials, chemicals, disruption) and what the event would have cost if the prediction was correct and no action was taken. Classify each action as true positive (event was real) or false positive (event would not have occurred). Sum the avoided costs for true positives, divide by the total cost of all actions.
Layer 2: Mathematical Notation
PVR = Σ (C_avoided_i × TP_i) / Σ C_action_i
Where:
- C_avoided_i = estimated consequence cost avoided for preemptive action i, if the prediction was correct (dollars; source: historical event cost data plus engineering estimate)
- C_action_i = actual cost of preemptive action i, including crew hours multiplied by rate, materials, chemicals, and any service disruption (dollars; source: CMMS/EAM work order cost fields)
- TP_i = 1 if action i was a true positive (event confirmed or conditions validated the prediction), 0 if false positive (source: post-event engineering review or statistical classification)
Proposed formula, pending validation pilots.