If you wear an Apple Watch and an Oura ring on the same night, the two devices will report different HRV numbers — sometimes by a factor of two. Neither device is wrong. They’re computing different metrics from the same underlying heartbeat data, and the two metrics are correctly producing different values.
This page is the disambiguation. If you’re here because your numbers don’t match across devices, this is why.
What each metric is
Both metrics start with the same input: the time interval between consecutive heartbeats, measured in milliseconds, across a window of several minutes. They then summarize that interval series in different ways.
SDNN — Standard Deviation of Normal-to-Normal intervals. The classical statistical spread. Take all the intervals in the window, compute their standard deviation, report the result. It captures the total variation across the whole window, including both fast vagal-mediated variation and slower respiratory and circulatory components.
RMSSD — Root Mean Square of Successive Differences. Take each pair of adjacent intervals, compute their difference, square it, average all the squares, take the square root. This deliberately emphasizes short-term beat-to-beat variation — the variation driven directly by the vagus nerve.
Why the numbers differ
For the same person on the same night, SDNN values are typically 1.5 to 2 times higher than RMSSD. This is mathematically expected: SDNN is summarizing more sources of variation than RMSSD. If your overnight RMSSD is 35 ms, your overnight SDNN is probably somewhere between 55 and 75 ms.
The relationship is consistent enough that the ratio between the two metrics is itself a useful number. A high RMSSD/SDNN ratio suggests strong vagal modulation; a low ratio suggests the longer-period autonomic variation is doing more of the work. Most consumer dashboards don’t expose this ratio because it would confuse more than it would help.
Which device reports which
| Device | Metric | Measurement window |
|---|---|---|
| Apple Watch | SDNN | 30–60s during stillness |
| Apple Health (clinical) | SDNN | Imported from supported devices |
| Oura Ring | RMSSD | Whole-night integration |
| Whoop | RMSSD | Last sleep cycle of the night |
| Garmin | RMSSD | Overnight + on-demand |
| Polar H10 (chest strap) | Both | User-configurable window |
| Welltory (camera PPG) | Both | 60s seated measurement |
Which one should you track?
Whichever your primary device reports natively. Don’t try to translate between them. The right approach is to pick a device, learn what its number looks like for you on good days and bad days, and watch your own seven-day rolling average.
RMSSD is more responsive to short-term shifts in vagal activity, which makes it the metric of choice for HRV biofeedback research and for short post-session measurements. SDNN integrates more sources and is slightly more stable across measurement-window sizes, which is why long-term clinical studies have historically leaned on it.
Both metrics correlate strongly with autonomic flexibility, with recovery, with stress, with sleep. The training-effect signal you care about — your number going up over weeks of practice — shows up in both. Pick one, ignore the other.
Why HRV Breathe uses SDNN
We read HRV from Apple Health. Apple Health stores HRV asHKQuantityTypeIdentifierHeartRateVariabilitySDNN — that’s the only HRV type Apple exposes through its HealthKit API. Apple Watch samples it natively in SDNN. Third-party devices like Oura, Whoop, and Garmin write their readings into the same SDNN slot when they sync to Apple Health, with a translation step from their native RMSSD.
For our purposes, the choice was made by the platform. SDNN is the metric Apple gives us. The before/after delta on the HRV Breathe completion screen is an SDNN delta. The trend chart is an SDNN trend. We’d use RMSSD if it were available; in practice the within-session signal we surface looks the same in either metric.
Higher-order metrics — pNN50, LF/HF, SD1/SD2
The HRV literature has half a dozen more metrics — pNN50, SDANN, frequency-domain LF and HF power, the SD1/SD2 ratio from Poincaré analysis. They’re all related; they all measure slightly different things; and for non-clinical practical use they add complexity without adding insight beyond SDNN and RMSSD.
If you’re a researcher or a clinician working with clinical-grade data and a specific hypothesis, those metrics matter. If you’re an adult with a wearable trying to get your numbers up through behavior change, they don’t. Pick the metric your device already reports, follow your trend, and ignore the rest.