Repeatability is how closely an instrument agrees with itself when it measures the same quantity repeatedly under the same conditions over a short time. A hygrometer that reads 51, 52, 51, 52, 51 percent on the same air has a repeatability of about ±1 percent. It says nothing about whether 51 is the true value; that is accuracy. Repeatability sets the smallest change an instrument can detect with confidence, which for monitoring is often the more useful property.
Repeatability, reproducibility and accuracy
| Property | What varies | What it tells you |
|---|---|---|
| Repeatability | Nothing; same everything, short time | Random noise and short-term stability of the instrument |
| Reproducibility | Instrument, operator, day or location | How much the method itself scatters |
| Accuracy | Compared with the true value | Systematic error, calibration, drift |
Precision is the umbrella term covering repeatability and reproducibility. See accuracy vs precision.
What limits repeatability
- Noise. Electronic and sensor noise scatter readings from sample to sample. Averaging reduces it.
- Resolution. An instrument cannot repeat better than its display step. A 1 °C display has ±0.5 °C repeatability at best. See range vs resolution.
- Hysteresis. Some sensors read differently approaching a value from above versus below. Humidity sensors show this after wet or dry excursions.
- Settling. Readings taken before the sensor has stabilized scatter with time. See why sensors need time to stabilize.
- The quantity itself. Air in a room is never perfectly still or uniform; part of the apparent scatter is real.
Typical figures
| Instrument | Typical repeatability |
|---|---|
| NDIR CO₂ monitor | ±10 to 20 ppm |
| Capacitive humidity sensor | ±0.1 to 1% RH |
| Digital thermometer (good sensor) | ±0.05 to 0.2 °C |
| Laser particle sensor | ±2 to 5 µg/m³ or ±5 to 10% |
| 6000-count multimeter, DC volts | ±1 to 2 digits |
| Thermal camera spot temperature | ±0.1 to 0.5 °C |
Why it matters
The smallest change you can trust is roughly twice the repeatability. A CO₂ monitor with ±15 ppm repeatability can reliably show a 30 ppm change; a humidity sensor with ±1 percent repeatability can show a 2-point change. For monitoring trends in one place with one instrument, repeatability is the property that decides whether the chart is meaningful, and accuracy can be mediocre without harm. For comparing rooms or instruments, accuracy takes over.
Testing it yourself
- Put the instrument in a stable environment: a sealed box for humidity, a large glass of water for temperature, outdoor shade for CO₂.
- Wait for settling.
- Record ten readings a minute or two apart.
- The spread between the highest and lowest is roughly twice the repeatability; the standard deviation is a better summary.
If the spread is much larger than the manufacturer’s figure, the environment is not stable, the sensor has not settled, or the instrument is faulty. Related: what measurement uncertainty is.
Recommended Tools
Fluke 117 Digital Multimeter, Non-Contact AC Voltage Detection, Measures Resistance/Continuity/Frequency/Capacitance/Min
View on Amazon
Products shown because their documented specifications fit this guide. Price and availability are on Amazon and can change.
Frequently Asked Questions
What is the difference between repeatability and reproducibility?
Repeatability: same instrument, same operator, same conditions, short time. Reproducibility: different instruments, operators, locations or days. Reproducibility is always worse.
Can an instrument be repeatable but inaccurate?
Yes, and it is common. A drifted sensor gives the same wrong answer every time. Repeatability is about consistency, not truth.
How do I test repeatability?
Measure a stable reference (a glass of water at room temperature, a sealed jar for humidity, outdoor air for CO₂) five to ten times a few minutes apart and look at the spread.
Affiliate disclosure: Akermin earns a commission from qualifying Amazon purchases made through links on this page. Our editorial picks are based on documented specifications and owner feedback, not commissions.