The Accuracy Trap
Your predictive maintenance model has an impressive accuracy score. It generates alerts. Work orders get cut. But here’s a question most maintenance organizations never ask: after you subtract the labor, materials, false alarm investigations, and production disruptions those alerts triggered, did the model actually make money?
Most PdM programs are evaluated on statistical metrics — precision, recall, F1 scores. Those metrics tell you whether the model is correct. They don’t tell you whether the model is profitable.
Alert Net Economic Yield
Alert Net Economic Yield (ANEY) is a proposed KPI that produces a single number: the net dollar value generated per AI-issued maintenance alert.
The formula takes the cost you avoided by catching the failure early, subtracts the full cost of the intervention the alert triggered (labor, parts, production loss during planned downtime), subtracts the cost of any false alarms, and divides by total alerts. The result is a dollar figure per alert that tells you whether your AI investment is producing economic value — not just statistical performance.
Why This Matters
Research from Spiegel et al. (2018) demonstrated that a model optimized for the best F1 score — the standard statistical measure — can actually be economically suboptimal. When failures carry asymmetric costs (and in maintenance, they always do), the statistically “best” threshold and the economically best threshold are different. Most PdM programs are tuned for the wrong one.
The Data Already Exists in Maximo
If you’re running Maximo 9 and MAS Predict, every data point in this formula is already sitting in your WORKORDER actuals, FAILUREREPORT hierarchy, and Predict alert logs. The data exists. The join just hasn’t been built yet.
Full Deep Dive
The full article covers the complete formula, walks through a worked example with real numbers, maps every component to specific Maximo fields, and covers the supporting research.
Read the complete deep dive on Alert Net Economic Yield — formula, worked example, Maximo field mapping, and supporting research.
Read Full Article on LinkedInAbout the Advanced KPI Series
This is the first deep dive in the Advanced KPI Series — 45 proposed metrics for AI-era maintenance that address blind spots legacy KPIs like MTBF, MTTR, OEE, and PM Compliance structurally cannot see. None of these KPIs are proven standards. They’re grounded in research and designed for the community to test.