What is a data historian?
A data historian is a time-series database built to record readings from industrial equipment and control systems at high speed and keep them for years. It collects named signals, called tags, such as a motor’s current draw, a tank’s level, a line’s speed, or a valve’s position, and stores each value with a timestamp. A mid-sized plant may have tens of thousands of tags, and a utility network may have far more.
How a data historian works
The historian connects to programmable logic controllers, SCADA systems, and distributed control systems through industrial protocols such as OPC UA or Modbus, and through gateways for newer IoT sensors. Because raw readings arrive many times per second, historians compress them, often by storing a new value only when it moves by more than a set deadband. Collectors buffer data locally and forward it later if the network drops. Most products also maintain an asset hierarchy that links each tag to the equipment it belongs to, so a query can ask for every signal on Compressor 3 rather than for a cryptic tag name.
Why a data historian matters
The historian holds the record of how equipment actually behaved, which is the raw material for condition monitoring, predictive maintenance, energy management, and root cause analysis. On its own it answers operational questions, such as what the discharge pressure was at a given minute. Its value multiplies when it is joined with business data that lives elsewhere: work orders from the maintenance system, production orders from the ERP, and quality results from the lab.
A water utility, for example, keeps pump station readings in its historian and repair history in its asset management system. A model that combines both can flag a pump whose power draw has drifted upward at the same flow rate, weeks before the bearing fails, and the planner can schedule the repair alongside other work in the same district.
What to watch for
- Compression settings tuned for storage can erase the short spikes a model needs to see, so review deadbands before using historian data for analytics.
- Tag names differ between plants and between eras of engineering, so budget time to map tags to assets before any cross-site analysis.
- Getting data out can be harder than getting it in, so confirm that the product offers an open interface for replicating data into a lakehouse and check how licensing treats that export.
- Clocks on controllers drift, and a few seconds of offset between sources is enough to break a cause-and-effect analysis.