Measurement-definition audit · Oura and Garmin · Checked November 21, 2026
Two resting-heart-rate numbers can disagree before either sensor makes a mistake. “Lowest during the night,” “nightly average” and “lowest 30-minute average during a day” select and summarize data differently. Printing bpm beside all three does not make them interchangeable.
Before explaining a sudden change as improved fitness or poor recovery, check whether the measurement changed. This is especially important when switching devices, joining exports or reading a dashboard that shortens every field name to RHR. The abbreviation is too small to carry the method by itself.
Four documented fields
| Field | Documented summary | Window |
|---|---|---|
| Oura Average RHR | Average resting heart rate captured during the night | Night |
| Oura Lowest RHR | Lowest resting heart rate captured during the night | Night |
| Garmin daily RHR | Lowest 30-minute average | 24-hour period |
| Garmin seven-day average RHR | Average of daily resting values across the last seven days, on devices offering it | Rolling seven days, not a calendar-week reset |
Oura's help page identifies both nighttime fields and says RHR values are captured every ten minutes. Garmin's calculation guide defines the daily and rolling fields and notes that missing overnight wear can reduce estimate accuracy. Those are documented definitions. The pages do not disclose every sampling, filtering, weighting or missing-data decision needed to reproduce the commercial output.
We therefore leave those implementation details unknown in the download. We do not assume that “average” always means an unweighted mean of every visible point, or that a displayed graph contains all data used internally.
One invented trace, three correct summaries
To isolate the arithmetic, imagine six consecutive ten-minute bins. Their average heart rates are 50, 55, 60, 65, 60 and 58 bpm. All bins are complete and equally long. This is a synthetic one-hour teaching trace, not a person's night and not an emulation of Oura or Garmin.
| Rule in this example | Calculation | Result |
|---|---|---|
| Lowest ten-minute bin | Minimum of the six bin means | 50 bpm |
| Lowest aligned 30-minute mean | Minimum of four consecutive three-bin means | 55 bpm |
| Whole-hour mean | (50 + 55 + 60 + 65 + 60 + 58) ÷ 6 | 58 bpm |
Nothing about the underlying trace improved between those rows. The difference is entirely the calculation. The 30-minute windows here start only at our ten-minute bin boundaries; a real device may consider other alignments, data or exclusions. Calling 55 a reproduced Garmin result would go beyond this example.
Now consider a separate invented week of daily resting values: 57, 55, 56, 54, 56, 55 and 59 bpm. Their mean is 56 bpm, while the latest day is 59 bpm. A daily tile and a seven-day tile can therefore differ by 3 bpm without contradicting one another. Again, this is transparent arithmetic, not a diagnosis or a claim about a particular app's missing-data behavior.
A device change can create a break in the series
Imagine exporting a month of nightly averages, then appending a new device's daily low-window values under the same column name. The resulting chart may look continuous, but its measurement definition is not. A before-and-after comparison now mixes a possible physiological change with a change in method.
Keep the source and field name with every value. Mark the switch date. If you have an overlapping period, compare like-for-like fields where they exist and preserve the wear conditions. An overlap can describe agreement in that setup; it does not automatically establish which device is clinically accurate or justify a universal correction offset.
Subtracting a fixed number to make the lines meet is especially tempting. It also assumes that the difference stays constant across nights and activity patterns. These documentation pages do not establish that assumption. A visible discontinuity with an explanation is more honest than a smooth line built on an untested adjustment.
What to check when the number surprises you
- Read the complete label. Is it a current heart rate, nightly average, nightly low, daily RHR or multiday average?
- Check the period and date. Confirm whether the two views refer to the same completed window.
- Check coverage and source. Overnight nonwear, a changed source or an incomplete record can complicate comparison.
- Check the documentation for that feature and version. Keep unspecified processing rules unspecified.
- Only then consider what the trend might mean in your own context. A calculation mismatch should not become a confident health explanation.
This guide does not supply a “good” resting heart rate, a training clearance or a threshold for illness. If the reading accompanies symptoms or raises a health concern, seek appropriate medical advice rather than using a source-definition audit to dismiss it.
Data and limits
We manually checked the two manufacturers' public definitions on November 21, 2026. The CSV and JSON contain the four fields, their sources, the complete synthetic sequences and calculation assumptions. No device was worn or tested for this article. It addresses definition mismatch, which is separate from sensor accuracy.
For a related example involving beat-to-beat intervals rather than bpm, see why HRV numbers are not interchangeable. Publisher disclosure: we make Vora, which combines proactive AI with wearable, nutrition and workout context. Keeping a measurement's meaning attached is part of making that context useful. Explore the product; the reference requires no account. Corrections: support@askvora.com.