The sampling rate is the first design decision in any sensing application. A vibration sensor for bearing diagnostics may need thousands of samples per second to capture the relevant frequencies; a soil moisture probe is fine with one reading an hour. The Nyquist criterion gives the lower bound: to capture a signal component, sample at more than twice its frequency.
Every extra sample costs something. It takes energy to wake the ADC, radio time to transmit, and storage to keep, so the rate is a lever on battery life and cloud cost. A common pattern separates sampling from reporting: measure often, compute a summary locally, send rarely. Where raw history is kept, downsampling later reduces it further.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.