I purchased my Oura Generation 2 ring in April 2021 and upgraded to the Generation 3 ring last week. As expected, it wanted to transfer my data to the new ring and restore the Gen 2 to factory default. Instead, I registered it on a second phone, allowing me to simultaneously track my statistics with both rings.
For those of you with the ring, you’ll be aware that some values are reported relative to a baseline, which Oura needs 1-2 weeks to establish. Additionally, some of the scores include 7 day averages, which will also mean the two aren’t directly comparable this week. However, there’s plenty that can be compared, such as heart rate, heart rate variability, sleep statistics and respiratory rate.
I’ve been typing up 16 values from both rings in a spreadsheet each day. This allows me to filter by date and value to compare the two. Let’s begin with the main score, Readiness.
For both Lowest Heart Rate and Average Heart Rate, the two values match on six of the seven days. On the seventh day it differs by 1 bpm – pretty good!
For Heart Rate Variability (HRV), the new Gen 3 gives on average a lower score. As you can see they differ by no more than 3ms, or around 5%.
The Maximum HRV had some bigger differences, but I’ve always considered the graph to look quite noisy. For that reason I don’t read too much into my maximum values, and here they differ by up to 22ms.
The final Readiness value that’s directly comparable is respiratory rate, which is generally within 0.2 (one breath different every 5 minutes). The values I can’t yet compare are body temperature and overall readiness score. The latter is because HRV balance, sleep balance and activity balance are all greyed out while it awaits more data.
I’ll quickly tackle Activity now and then finish with a comparison of sleep metrics.
Now the Oura ring doesn’t claim to be the leading wearable for fitness tracking, so expectations are low. That being said, calorie burn is at least within 10% of each other*. I considered trying different activities to see where it differs most. However, the next software update will introduce heart rate tracking during workouts, so it seems redundant to attempt that now.
*I should note that 30th November includes 109 calories manually entered as a workout.
Inactivity is defined as sitting, standing or otherwise being inactive in the last 24 hours, excluding resting and sleep. I do have periods at my desk where I’m only monitoring something on screen, and both rings can classify this as rest. It’s within 10%, but the absolute difference being as much as 45-50 minutes feels like a lot.
Okay, on to arguably the most important metrics reported by the Oura ring.
I’ve always felt the Oura Gen 2 ring was excellent at determining my initial sleep time and eventual wake up time. The discrepancy here in total sleep time is therefore due to the interpretation of night waking. I have 14 month old twins, whose cots are either side of our double bed, so night waking is almost guaranteed. Here the newer Gen 3 ring reports on average 10 minutes longer total sleep than its predecessor. The sleep algorithm is constantly being updated so you would expect the Gen 3 to be the more accurate of the two, but I can’t personally comment.
Deep sleep is thought of as the physically restorative, taking place mainly in the first half of the night. While two nights report fewer minutes than the Gen 2, on average the new generation 3 ring reports 20 minutes more deep sleep.
By contrast, REM sleep (rapid eye movement) is the mentally restorative sleep, generally in the second half of the night. In the first 7 days of wearing both rings the Gen 3 ring assigns an additional 16 minutes per night to REM sleep. While I haven’t recorded it here, the third category of sleep is Light sleep, and the three stages add up to total sleep time. What that means is, with 10 minutes extra total sleep and 36 minutes more REM and deep sleep, the Gen 3 ring reports on average 26 minutes less light sleep than the Gen 2 per night.
Sleep latency is the number of minutes it takes you to first fall asleep, having got into bed. This is one I would expect the rings to perhaps be close to each other, but overall one that I feel it underreports.
Scenario: I go to bed at 10.10pm, but after lying awake for a while I reach over to check the time on my phone. It says 10.35pm. I then roll over onto my other side, and drift off in 8 minutes. The true latency here should be from 10.10pm to 10.43pm, i.e. 33 minutes, but it will generally reset from my movements at 10.35pm, and report a latency of 8 minutes. That’s likely what happened on the 29th November, where the difference is 10 minutes, and otherwise the two are pretty similar.
- Sleep efficiency is the proportion of time in bed that it considers to be sleep time, which is within 1-4% so far in week one.
- Restfulness measures tossing and turning, getting up to go to the loo etc.,
- Timing is based on an optimal window (in my case 9.30-10.30pm), and measures consistency of bed time.
These three, and the above four values, are the contributors to my overall sleep score.
Unsurprisingly, given the Gen 3 reports longer total sleep, deep sleep and REM sleep, my sleep score is better in the Gen 3. As the sleep score feeds through to the readiness score, this may have an impact on how much activity it will recommend for the coming day. 29th November was the biggest difference in total sleep time, with the Gen 3 reporting an extra 29 minutes of sleep.
Both average and lowest heart rate continue to be either aligned or differ by a single beat per minute. Respiratory rate differs by less than 0.2 breaths/min on average while body temperature was within 0.1 degrees celsius, and so I won’t discuss these further.
Based on the second week, HRV tends to be lower with the new Gen 3 ring, by an average of ~3ms. Maximum HRV differences ranged from -27 (Gen 3 being lower) to +14. The new ring finally had enough baseline data to report comparable numbers in the main Readiness score category, seen below.
As seen in week 1, the higher sleep scores are feeding through and boosting my readiness score, though not consistently. If I had to pick out something different for the 6th Dec I’d say it was the increased number of wakeful periods I had the previous night, so that may be where the two rings most disagree.
Calorie burn has continued to be within 5% and I didn’t do any new activities so there’s nothing really to report. I also got the email through about the delay to workout heart rate data until 2022 which is a shame.
I’m starting to see a more consistent disagreement over how much time I spend inactive. In a typical 8 hour workday I try to stand for at least two 60 minute periods, but I’m not sure where the difference comes from. The Gen 3 is registered to my wife’s phone so I can ask to sync it during my lunch break and see whether the difference is evenly distributed throughout the day.
Total sleep time is generally longer with the new ring, note that the 7th Dec is misleading as I confirmed a 20 minute nap on my Gen 2 but forgot to do the same on Gen 3. Deep sleep in week 2 was on average 25 minutes longer (1:38 vs 2:03) and REM sleep by 10 minutes (0:58 vs 1:08), meaning my sleep score was often 4 or 5 higher with the Gen 3 ring. Latency was very well aligned in week 2, with only one day differing by more than a minute.
It’s unfortunate that the software update is now scheduled for early 2022. Before then I will do my best to narrow down the differences in interpretation for inactivity time and report back here.