What is predictive maintenance software, and when is it worth it?

Predictive maintenance software is the system that decides when to service an asset based on its measured condition rather than on a schedule. Where preventive maintenance runs on a calendar or run-hours interval, predictive maintenance, also called condition-based maintenance, watches signals from the asset (vibration, temperature, run current, building system data, inspection readings) and triggers the work when those signals show failure is heading your way. The software collects the condition data, models what "healthy" looks like, flags the deviation, and opens the work order before the breakdown happens.

The reason to care is cost, and the number most people quote comes from the US Department of Energy. According to the DOE's Operations and Maintenance Best Practices Guide, a functioning predictive program runs roughly 8 to 12 percent below a preventive one, and as much as 40 percent below reactive run-to-failure. That figure is real, but it is an average over the assets predictive is suited to. It is not a promise that monitoring everything saves money.

Predictive vs preventive maintenance, in one table#

For the full portfolio-level version of this comparison, see preventive vs predictive maintenance.

When predictive maintenance software is worth it#

The test is two questions, and both have to be yes. Can the asset's condition be measured before it fails? And is a failure expensive enough that catching it early pays for the monitoring? On a rooftop HVAC unit over a busy store, both are yes: the condition degrades detectably, and a failure means a hot or closed site and an emergency repair at premium rates. On a cheap light fixture, monitoring costs more than the fixture, so predictive never earns its keep.

For a multi-site occupier, the assets that clear both bars are consistent across the portfolio: HVAC and rooftop units, commercial refrigeration, elevators, and electrical switchgear. Concentrating condition monitoring there, and leaving the long tail of cheap assets on a sensible preventive or reactive footing, is where the real saving comes from. The mistake is buying predictive maintenance software and pointing it at everything.

What the software actually needs#

Three things. Condition data, from sensors, meters, building automation systems, or disciplined manual inspection. A baseline per asset, so the software can tell a real signal from normal noise, which takes a little time to establish. And a work-order path, so a flagged signal becomes a scheduled repair by a named vendor rather than an alert nobody actions. That last piece is why predictive maintenance works best inside a broader maintenance management system rather than as a standalone sensor dashboard.

REAL approaches predictive maintenance as a prioritization problem across sites: it reads asset and warranty detail out of documents, watches condition where it is available, and ranks the work by the cost of failure and the value of the location, next to lease and portfolio data. See how that plays out in predictive maintenance for multi-site portfolios and how it connects to asset performance management.