Skip to main content
Every time series in CDF has a type: numeric, string, or state. You set the type when you create the time series, and you can’t change it later. The type determines which values a data point can hold, how CDF interprets the time between data points, and which aggregates you can request.

Compare the time series types

Numeric, string, and state time series differ in the values they store, how CDF validates and interpolates them, and which aggregates they support. For the aggregates in each group, see Aggregates by time series type.

Choose a time series type

Pick the type based on what the values represent and how you want to analyze them:
  • Use numeric when each value is a measured or calculated quantity, such as temperature, pressure, flow rate, or a production KPI.
  • Use state when the values come from a known, finite set of conditions, such as ON/OFF or RUNNING/IDLE/ERROR. CDF rejects values outside the set and can calculate how often and how long equipment is in each state.
  • Use string for free-form text that doesn’t fit a fixed set, such as diagnostic messages, batch identifiers, or firmware versions.

Numeric time series

Set the isStep property to match how the measured process behaves. Use isStep: true for values that hold until the next data point, such as setpoints and counters, and isStep: false for values that change gradually, such as most physical measurements. CDF uses isStep for interpolation and for time-weighted aggregates such as average. You can change isStep after you create the time series. For how each interpretation affects the results, see Stepwise vs continuous.

String time series

Each string value can be up to 1023 bytes when UTF-8 encoded. CDF stores values exactly as you ingest them: an empty string is a valid value and is different from a missing data point, and "OK" and "Ok" are two different values.

State time series

Each state time series references a state set that defines its valid states. To learn how to define state sets, ingest state data points, and analyze state changes, see State time series.

Set the type when you create a time series

Set the type property on the CogniteTimeSeries view to numeric, string, or state. A state time series also needs a stateSet property. See State time series.
Create a numeric time series
For more information, see Time series in data modeling.

Next steps

Aggregation

Learn how CDF calculates each aggregate and how granularity works.

State time series

Define state sets, ingest state data points, and analyze state changes.

Time series in data modeling

Create time series as CogniteTimeSeries instances in your data model.

Time series and data points

Insert, retrieve, and query data points with the API.
Last modified on November 16, 2026