Downsampling keeps long histories affordable and charts readable. Readings taken every second become one-minute averages after a week, hourly summaries after a month, and daily figures after a year; a time-series database usually does this through continuous queries tied to the data retention policy. Dashboards also downsample on the fly, since a screen cannot display more points than it has pixels.
The trade-off is information loss. An average hides the short spike that explains a fault, so good policies keep minimum and maximum alongside the mean, or keep raw data around events of interest. Downsampling can also happen on the device as edge analytics, which saves bandwidth but makes the decision irreversible.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.