# Predictive Failure Analysis
**Entity class:** Storage firmware / failure prediction
## Definition and history
IBM's Predictive Failure Analysis used drive measurements to issue an impending-failure notification in products from 1992. [[articles/Intelligence Inside|Intelligence Inside]] identifies this firmware judgment as an early classifier resident in ordinary hardware. A threshold could turn multiple measured mechanical conditions into a binary report sent toward the controller. That outward result is distinct from a device learning its own threshold during use.
## Mechanism and limits
Failure prediction shifts maintenance from waiting for damage to acting on an estimated future condition. The warning depends on which parameters are monitored and how thresholds were set, so silence is not a proof of health. Later SMART reporting expands the drive-health interface, while large operators can aggregate readings across fleets for separate modeling.
## Relationships
**Precursor to:** [[wiki/SMART Drive Monitoring|SMART Drive Monitoring]]. **Function:** [[wiki/Classifier|Classifier]] and [[wiki/Telemetry|Telemetry]]. **Comparison:** [[wiki/Intel Thread Director|Intel Thread Director]]. **Article route:** [[articles/Intelligence Inside|Intelligence Inside]].
## Sources
[SMART project history and PFA documentation](https://smartlinux.sourceforge.net/smart); [[articles/Intelligence Inside|Intelligence Inside]].