In an effort to help users avoid data loss, drive manufacturers are now incorporating logic into their drives that acts as an "early warning system" for pending drive problems. This system is called Self-Monitoring Analysis and Reporting Technology or SMART. The hard disk's integrated controller works with various sensors to monitor various aspects of the drive's performance, determines from this information if the drive is behaving normally or not, and makes available status information to software that probes the drive and looks at it.

The fundamental principle behind SMART is that many problems with hard disks don't occur suddenly. They result from a slow degradation of various mechanical or electronic components.
SMART evolved from a technology developed by IBM called Predictive Failure Analysis or PFA. PFA divides failures into two categories: those that can be predicted and those that cannot.

Predictable failures occur slowly over time, and often provide clues to their gradual failing that can be detected. An example of such a predictable failure is spindle motor bearing burnout. This will often occur over a long time, and can be detected by paying attention to how long the drive takes to spin up or down, by monitoring the temperature of the bearings, or by keeping track of how much current the spindle motor uses.
An example of an unpredictable failure would be the burnout of a chip on the hard disk's logic board: often, this will "just happen" one day. Clearly, these sorts of unpredictable failures cannot be planned for.

The main principle behind failure prediction is that some failures cause gradual changes in various indicators that can be tracked to detect trends that may indicate overall drive failure.

• Head Flying Height: A downward trend in flying height will often presage a head crash.
• Number of Remapped Sectors: If the drive is remapping many sectors due to internally-detected errors, this can mean the drive is starting to go.
• ECC Use and Error Counts: The number of errors encountered by the drive, even if corrected internally, often signal problems developing with the drive. The trend is in some cases more important than the actual count.
• Spin-Up Time: Changes in spin-up time can reflect problems with the spindle motor.
• Temperature: Increases in drive temperature often signal spindle motor problems.
• Data Throughput: Reduction in the transfer rate of the drive can signal various internal problems.

Using statistical analysis, the "acceptable" values of the various characteristics are programmed into the drive. If the measurements for the various attributes being monitored fall out of the acceptable range, or if the trend in a characteristic is showing an unacceptable decline, an alert condition is written into the drive's SMART status register to warn that a problem with the drive may be occurring.

Clearly, SMART is a useful tool but not one that is foolproof: it can detect some sorts of problems, but others it has no clue about. A good analogy for this feature would be to consider it like the warning lights on the dashboard of your car: something to pay attention to, but not to rely upon. You should not assume that because SMART generated an alert, there is definitely a drive problem, or conversely, that the lack of an alarm means the drive cannot possibly be having a problem. It certainly is no replacement for proper hard disk care and maintenance, or routine and current backups.

If you experience a SMART alert using your drive, you should immediately stop using it. Some support approaches consider a SMART alert sufficient evidence that the drive is bad, and will immediately replace it. Others require other steps to be performed, such as running diagnostic software on the drive. In no event should you ignore the alert.