How do you measure HRV?

Your device times the gap between each pair of heartbeats, then sums hundreds of gaps up in one number. Apple Watch uses a calculation called SDNN. Oura, WHOOP and Garmin use RMSSD. Measure at night, every night, on one device.

The two calculations answer different questions about the same beats, so a watch and a ring can give very different numbers for the same night. Neither is wrong, and there’s no way to convert one into the other.

Why do my ring and my watch give different numbers?

It’s one of the most common questions. You wear an Oura ring and an Apple Watch on the same night and get two numbers that aren’t close. One of them must be broken.

Usually neither is. They differ for three reasons, and each one alone can change the number a lot.

1. Different math. Apple Watch reports SDNN. Oura, WHOOP and Garmin report RMSSD.
2. Different stretches of time. Apple Watch takes short samples. Oura averages five-minute samples across the night. WHOOP weights its average toward deep sleep and later in the night. Garmin averages the whole sleep.
3. Different sensors and processing. Each company filters out bad beats its own way, and not all of them say how.

A simple way to picture it

Shoe sizes. A 9 in one brand can fit like a 10 in another. Your size in one brand is very useful for buying that brand again, and close to useless for comparing your feet with a friend’s in another brand.

HRV is the same. Your number on your device is a good guide to your own nights. It’s a poor guide to anyone else’s, or to your own on a different device.

Twelve gaps between beats, two numbers

Here are twelve resting gaps between beats, in milliseconds: 980, 1010, 1040, 1060, 1030, 990, 960, 950, 980, 1020, 1050, 1040. The average gap is about 1,009 ms, a heart rate of about 59.5 beats a minute. The gaps rise and fall with breathing, which is normal.

1. SDNN asks how widely the gaps spread. It takes all twelve, finds their average and works out the standard deviation around it. Here SDNN is 36.5 ms.
2. RMSSD asks how much each gap differs from the one before. It takes the eleven differences (30, 30, 20, minus 30, minus 40, minus 30, minus 10, 30, 40, 30, minus 10), squares them, averages them and takes the square root. Here RMSSD is 28.9 ms.
3. Now shuffle the same twelve gaps. Put them in this order: 980, 1060, 950, 1050, 960, 1040, 990, 1030, 1010, 1020, 1040, 980. The spread hasn’t changed, so SDNN is still 36.5 ms.
4. RMSSD more than doubles. The neighbouring gaps now jump back and forth, and RMSSD is 68.5 ms.

That’s why no formula turns one into the other. SDNN doesn’t care about the order of the beats. RMSSD is built from the order. A number on your screen has already thrown the order away.

The two also behave differently with time. SDNN picks up slow drifts across a recording, so a longer recording gives it more to find. RMSSD mostly follows the fast, breath-by-breath changes the vagus nerve makes, which show up in seconds.

Which device measures what?

Is night or morning the better time?

For most people, night. You’re lying down, still and asleep, so one night is a fair match for the next. Oura measures HRV only during sleep, it says, “to ensure the most reliable and interpretable data.” Garmin uses the whole night and calls that an advantage over methods that measure only part of it.

A morning check is the other good option, and it’s how the training trials we cite measured HRV, every morning. The rules that make it work:

• Same time, same position, every day. Researchers testing the Apple Watch had people lie still for five minutes each morning on waking.
• Breathe normally. Slow breathing raises HRV while you do it, so a breathing session isn’t a resting check.
• A few minutes is enough for RMSSD. In 3,387 adults, RMSSD from two minutes agreed almost perfectly with five, and even ten seconds gave a usable value. RMSSD held up better than SDNN at every length tested.
• Keep it apart. A morning number and a night number are two different measurements, even from the same device.

Daytime samples are the least useful for tracking. In a lab test of the Apple Watch, agreement with ECG was best at rest and worse with talking and walking, and an afternoon sample at your desk isn’t comparable with a night.

How accurate are watches and rings?

Good enough to follow your own trend, and less accurate than an ECG for any single number.

A chest strap measures the electrical signal of each beat. A watch or ring shines light into the skin and times the pulse as it arrives. Garmin calls its own measurement “pulse rate variability” for that reason. In 103 college athletes, the pulse-based RMSSD averaged 81 ms against 104 ms from ECG on the same recordings, and caught only 17 to 57 percent of the changes the ECG flagged, 1.8 to 5.5 days late.

These studies tested accuracy against an ECG, which is a different question from whether a device tracks your own changes. A device that runs low by a steady amount can still show your trend, though the athlete study above found the pulse-based number missed many short changes. That’s one more reason to judge a week, not a night.

Should I use the average or the highest number?

The average. Oura and Garmin both show your highest five-minute value of the night. It’s one window out of the whole night, so any single good or bad stretch moves it, and irregular beats can push it up.

Gyroscope’s old guide said both the average and the highest number were useful. We’d now use the average alone for tracking.

What changes the measurement without changing you?

• A new device, model or wear spot. Start a new history.
• A loose fit or poor contact. In the Oura sleep lab study, the number only matched ECG well when most beats in each window were clean.

A 2026 guideline paper from heart researchers lists the input signal, the length of the recording, where it’s taken, breathing and the method of analysis as things that all change HRV results.