HRV Comparison: Eightsleep vs. Garmin

TL;DR

1. Eightsleep and Garmin were found to match HRV trends at times, but not consistently enough.
2. Eightsleep is measuring without any wearable sensors, hence the data are more difficult to acquire - but it seems that it compares well with wearables.
3. Garmin's measurement frequency is higher, resulting in greater volatility in its readings. But it can also pick up more sudden HRV changes that Eightsleep sometimes doesn't register.
4. Both devices' HRV readings have spikes and dips in the data. These changes are large and appear abrupt, likely due to the sensitivity in measuring HRV.
5. No obvious pattern as to which device picks up HRV changes first. Sometimes it's Eightsleep, other times it's Garmin.
6. Eightsleep overestimates readings when it registers an HRV increase (spike), while Garmin's readings are overestimated when it registers an HRV decrease (dip). Other times, HRV readings match in magnitude and their overall trend.

After comparing HRV readings during sleep between Oura and Polar, we now move our attention to Garmin and Eightsleep. One of our Terra team members wore a Garmin Epix 2 while sleeping on an Eightsleep Pod Pro 2 cover and tracked their HRV for a week.

Looking at the trends in general, we found that at times both readings did have comparable HRV trends that users can gain sleep and health insights. However, the readings are still too volatile when looking at the finer details, with unexplained spikes and dips occurring. HRV readings were also at times too low on the whole, but are only too high during spikes and not the entire night.

Let's take a look at some of the data.

The first thing to note is the difference in measurement frequency of the data. Eightsleep stores HRV readings every 15 minutes, while Garmin stores readings every 5 minutes. As a result, Eightsleep's data in the graph above is coarser and does not pick up changes occurring at shorter times. Garmin's data shows a greater amount of detail (and volatility) in their HRV trends.

For this night, the trends between the devices deviate quite a lot. However, from around 1:00AM to 6:00AM, a general increase in HRV can be seen that both wearables pick up. Garmin's HRV fluctuates more at smaller intervals due to the higher frequency of readings (for a list of different wearables' measurement frequencies, see this article).

There are also occasional spikes that occur in the data, especially for Eightsleep. This can be seen with the double spikes occurring at 7:00AM and 7:30AM for Eightsleep. Garmin picks up an increase and drop-off in HRV during this timeframe, but the reading does not abruptly spike as with Eightsleep and is lower in magnitude. This is likely due to Garmin's 5-minute measurement frequency picking up the trend more precisely. Since both devices record a change in HRV here, we can deduce that there likely was an increase in HRV around this time. However, a jump of nearly 60ms for Eightsleep seems too large.

Looking at the graph above, we again see Garmin's higher volatility as well as its dips in HRV readings. On the other hand, we see coarser data from Eightlseep as well as more abrupt spikes. For dips and spikes from both devices, the change in HRV seems too high.

The data here does not match well for the most part. This is another example where disagreeing HRV trends between devices make it difficult to conclude about the user's sleep quality and overall health.

There are interesting data points where again Eightsleep may not be picking up HRV changes as fast as Garmin since the Pod Pro 2 records every 15 minutes vs Garmin's 5-minute intervals. As a result, Garmin does a better job with sudden HRV increases, such as the spike at 3:00am. Garmin registers this increase at a lower magnitude and drops back down earlier than Eightsleep. There is also an increase in HRV from 4:00AM to 5:00am for both devices, but Garmin's data is periodically increasing while Eightsleep's increases linearly.

The devices both record an increase of around 25ms in HRV at around 6:00AM and a subsequent decrease. This change matches quite well, but Eightsleep estimates the spike at a slightly higher value than Garmin. Let's now take a look at a night with more agreeable readings.

We can see here that Eightsleep and Garmin's HRV trends match each other quite well. Garmin as always has greater volatility due to the higher measurement resolution. There is however some disagreement in the data that warrants further investigation. These disagreements are either bad measurements or due to Garmin picking up changes in trends in-between Eightsleep's 15 minute interval. For example, Garmin's spikes between 4:15AM and 4:30AM or 5:45AM and 6:00AM is not picked up by Eightsleep as it was not storing HRV readings in between these times.

These readings are the most agreeable HRV measurements between two devices that we have found so far for a night of sleep. This is a good sign that wearables can actually measure and match each other's HRV trends well, but greater accuracy is still needed. Data between two wearables need to consistently match, as is the case with heart rate measurements between wearables (for the most part). This was likely a night where the user had not moved a lot during sleep and both devices accurately recorded HRV. However, HRV readings need to agree more consistently night in, and night out.

We can better see from the graph above that on most nights, Eightsleep overestimates spikes in HRV (e.g. at 5:00AM) while Garmin overestimates dips in HRV. For Garmin, a dip can be seen in readings between 3:15AM and 3:30AM. While Eightsleep does not record a change in HRV in this time due to its 15-minute measurement window, Garmin records a dip of around 40ms, which seems too large. Garmin occasionally does overestimate spikes, such as the spike between 1:15AM and 1:30AM, but not as often as Eightsleep. For example, at 5:00AM, Garmin and Eightsleep both have an increasing HRV trend, but Garmin measures a much lower increase.

The volatility seen in HRV readings in these graphs is currently par for the course for HRV estimates from wearables, as we saw when comparing Oura and the Polar Unite last time. This article also showcases the importance of keeping in mind differences in measurement frequency between devices as well as the benefits of storing HRV readings more often to better visualize trends.

HRV measurements between wearable devices still disagree too often, lacking an accurate measuring standard. Being able to match HRV trends between wearables is important for us to be able to trust wearables' HRV readings. This would then allow users to meaningfully attribute the source of overall changes in their HRV, as well as spikes and dips, to actual health trends and lifestyle choices.