Polar has launched a website intended to sell its sports/fitness/physiological algorithms to other companies. With this new venture, they aim to enable other sports/fitness etc. Companies can leverage Polar’s algorithms. This can be used in areas such as sleep, activity tracking, training load and recovery. So, if you want to make your own watch called a banana watch, you can use Polar’s health/sport algorithm to do it (don’t make a sports tracker called “Pear”, it exists ).
This license load is also common in other Finnish companies. Firstbeat has done this for many watch and wearable makers. Of course, Garmin acquired Firstbeat about 2.5 years ago, but the current Firstbeat site doesn’t exactly give the impression that they want the business (the privacy/legal/accessibility page says e-there I have an email at , so I think I can use that). Nonetheless, I checked with Firstbeat/Garmin and they confirm that they do in fact license the Firstbeat algorithms, including new business opportunities.
In any case, all this recent change clearly opens the door for companies like Polar to step in and fill that void with their own (and similar) slate of algorithms.
For a brief background and explanation of this segment, historically we have seen many companies license the Firstbeat algorithm in the wearable space. For example, Suunto, Amazfit, Huawei, Mont Blanc, Casio, and many other companies have been using such algorithms for years (past CES featured a huge web listing all devices and algorithms). There was even a huge display full of them along with the site). ). These can be used for simple components such as VO2Max estimation and calories burned. Or it can be more complex, such as training status, race prediction, etc. Companies that are less focused on sports tech devices typically go for basic and moderate algorithms, while those with a focus on sports tech aim for more complex algorithms.
Generally speaking, companies license these algorithms on a per-quantity basis and typically apply discounts for bundled sets of algorithms (eg, a set designed to work together). These sales volumes are reported on a catch-up basis when selling devices. Sure, that could lead to scenarios where competitors are afraid to report sales numbers (like Garmin buying his Firstbeat or now using Polar’s algorithm). Practically speaking, most companies have a rough idea of how many units their competitors are selling.
Finally, this space doesn’t have many players. There are players like Valencell and LifeQ, but they are also focused on providing bundled sensors and algorithms. For example, buy an optical HR sensor package to get different algorithms. Still, there are standalone products as well. For example, Suunto recently moved their optical HR sensor to LifeQ (from Valencell). However, all of these companies tend to offer a lot of business-to-business services, and at times it can seem like they threw spaghetti at the wall. / We see business opportunities in the medical field and try to fill those holes.
Anyway, with that background out of the way, let’s talk about Polar’s products.
Polar Algorithm:
So what does Polar offer? According to their site, there are currently 25 different algorithms available, and according to Polar, these algorithms are divided into the following categories (explained below: reference).
– Activities: 24/7 activity tracking lets you know how your customers are moving in their daily lives
– training: Monitoring training gives customers detailed insights that allow them to take action and work to improve
– performance: Knowing your key performance indicators, your customers can focus on how to improve their performance
– Sleep: Improve customer sleep insights to ensure they are better rested and fully charged each day
– recovery: Guide your customers to optimal recovery and prepare them for their goals for the day
– Wellness: Helping clients regain balance in their lives through holistic health and rhythm
As you can see, these are very general categories without many details. You can imagine that sleep includes not only sleep time but also some of Polar’s different sleep metrics for ANS, HRV and sleep stages. It may also contain a SleepWise component. Hard to say. Below these categories are details on next steps and how that process works.
This all makes sense and is fairly straightforward for an algorithm-focused company.As you can see, we don’t just give you a snippet of code (or a compiled binary) and call it done. Rather, it is also a verification that the algorithm is working as expected. Some of these are not very obvious at first glance. Sure, it’s easy to test when you go to bed and when you wake up. But having a test case for training load/recovery is much more complicated and it’s important to make sure the activity/workout data pipeline is feeding the algorithm correctly.
With that in mind, I reached out to Polar for a little more information on step 1 above, specifically a description of the functionality. These high-level descriptions are certainly not confidential. Unfortunately, Polar says they have nothing more to share.
For context, competitor Firstbeat lists not only algorithms on their website, but also detailed instructions and user interface examples (they may not list contact methods, but at least I’m pretty specific in listing what I’m selling). detail). Here’s a snippet of his one of the 26 algorithms Firstbeat currently offers:
Of course, I’m not sure if this is the full Firstbeat catalog. Some are quasi-new, but others are missing his Firstbeat driven features/algorithms. It may be because they haven’t updated their website or Garmin has decided not to license them. Both are driven by his Firstbeat team.
Now, all that aside, one thing I often hear from people (regardless of company) is that all of these algorithms are “junk” or “useless.” Basically, I classify these complaints into three basic categories.
A) A legitimate case for most people where the algorithm is kind of terrible and very shortsighted (Whoop Recovery score is a good example of this)
B) Legitimate scenarios where the algorithm has edges/use cases that don’t work well for a small population (e.g. Polar SleepWise doesn’t take naps into account)
C) User dislikes or does not understand the data output or applies it incorrectly (example Garmin’s previous training/status is misleading/misleading)
It turns out that most of Polar’s algorithms tend to do what they’re supposed to do well. If you actually allow them and are willing to understand them. It doesn’t matter how good your algorithm is if it’s easy to misunderstand or misuse. Also note that in this context, we are setting aside the underlying data stream (such as optical HR sensor or GPS/accelerometer data, if applicable). This is also important in cases such as previous generation Whoop devices where the optical HR sensor was so bad that the final state strain/recovery data was useless. for people).
The link from Polar’s Algorithm Licensing page is the Polar Research Center. I never saw it until recently. This is very cool and has lots of links to both internal and external research. Even the internal research whitepaper is far superior to what Garmin (or Whoop, Oura) publishes. Granted, they’re mostly trying to expose only “good” data, but at least they’re exposing the details of how features work beyond the marketing paragraphs.
In fact, my guess here is that most of the algorithms they license probably match the neat little list shown in the “white papers” section of this site almost exactly.
That list linked above most closely matches what I would expect Polar to offer/license, and is most competitive with what Garmin/Firstbeat offers.
from now on:
I think this strategy makes a lot of sense for Polar. There are a number of companies that have leveraged Firstbeat over the years, but since Firstbeat is no longer independent, it makes sense that these companies wouldn’t want to give his Garmin any revenue or sales numbers. (even if the number of units sold is mostly known). ). Polar has a solid slate of algorithms that compete very well in the market, making them very attractive to companies like SRAM/Hammerhead or Wahoo that have no physiological capabilities at all. You may like it, but the reality of the market clearly shows that you need at least some of these features to compete in the same price range.
But it is also attractive to companies that are direct competitors. Companies like Amazfit, Suunto could see Polar as a better option either in terms of price or competition. After all, if I were Suunto, I’d rather pay Polar than Garmin. Sure, it’s a bucket drop, but in most cases it’s better to have a smaller company. Moreover, both are from Finland. Sauna credit perhaps?
It will be interesting to see where this goes and how long it will take for Polar to get its first public customer. I wouldn’t have spun up this business unit/product for as long. I’m sure time will tell.
Thank you for reading!
.