time series data

Time series data is a sequence of data points collected and stored in time order, where each value is associated with a specific timestamp. In industrial and manufacturing environments, it commonly refers to time-stamped measurements from equipment, sensors, control systems, and software applications.

Unlike transactional data, which describes discrete business events (such as a purchase order or a work order release), time series data captures how a variable changes over time. Typical examples include machine temperatures sampled every second, line speed every minute, OEE components per shift, or the count of good and bad parts by time interval.

Key characteristics

• Time-stamped: Each record has an explicit time reference (date/time), often in a standardized time zone.
• Ordered: Data is logically and usually physically ordered by time, enabling sequence and trend analysis.
• Often high-volume: Industrial sensors or control systems may generate values every second or faster.
• Typically numeric: Most time series values are numeric (e.g., pressure, counts, KPIs), though status codes or categorical states can also be recorded over time.
• Granularity-dependent meaning: The interpretation depends on the sampling interval (per second, per cycle, per batch, per shift, etc.).

How time series data is used in manufacturing

In regulated and industrial operations, time series data commonly supports:

• Operational performance metrics: Calculating KPIs such as OEE, availability, performance, and quality over defined time windows.
• Compliance and traceability: Providing evidence of process conditions (temperatures, pressures, cycle times) during production of specific lots or serial numbers.
• Condition and asset monitoring: Tracking vibration, current, temperatures, or error codes to assess equipment health.
• Alarm and event analysis: Correlating alarms, mode changes, or recipe changes with process variables and product outcomes.
• Capacity and utilization analysis: Using time-stamped machine states (run, idle, down, setup) to analyze downtime and throughput patterns.

Systems that generate and store time series data

Multiple layers of the industrial stack generate and manage time series data, including:

• PLC/SCADA and distributed control systems, capturing real-time process signals.
• Historians and time series databases, optimized for high-frequency time-stamped data.
• MES and production tracking systems, recording states, counts, and KPI values by time.
• Quality and test systems, logging measurement results and test outcomes over time.
• IoT platforms or data lakes, aggregating time series data from multiple plants or assets.

Time series data and KPI auditability

For auditable KPIs, such as those aligned with ISO 22400, time series data provides the raw evidence for calculations. Each KPI value (for example, OEE for a shift) can be traced back to the underlying time-stamped events and measurements from which it was computed. This often requires:

• Consistent timestamping and time zones across data sources.
• Versioned and documented aggregation or transformation logic from raw time series to KPIs.
• Retention and controlled access to historical time series used in past KPI calculations.

What time series data is not

• It is not limited to financial data, although finance and forecasting are common uses.
• It is not the same as master data (such as part numbers or BOMs), which generally does not change at high frequency over time.
• It is not just event logs; events may be part of a time series, but continuous or regularly sampled measurements are typical.

Common confusion

• Time series data vs. event logs: Event logs capture discrete occurrences (e.g., a batch start), each with a timestamp. Time series data often involves continuous or periodic measurements. In practice, both can be combined for analysis.
• Time series database vs. historian: In manufacturing, a plant historian is a specialized form of time series database. The terms are sometimes used interchangeably, but historians are typically tuned for OT and process data.