This episode of the
Combining Fitbit’s health expertise with Google’s long record of technical achievement makes for a winning heart rate tracking mix. DeCarlos discusses the distinct talents and strengths behind Fitbit and Google that enable the creation of this exceptional health and fitness device – and what makes heart rate tracking tick on the all-new
Collecting data is a crucial part of understanding how to make heart rate tracking accurate, precise, and tailored to an individual. Learn how Google developed its heart rate tracking data, sensors, and hardware to increase the accuracy of tracking regardless of your activity level.
Hear the full episode on the
Transcript
Rachid (00:09.538) Decarlos, welcome back, your second time on the podcast. What I remember from last time is that you aren't just some computer scientist, but one that's been an athlete too. Are you still getting your heart rate up these days?
DeCarlos Love (00:22.863) Yeah, thanks to one of the benefits of the job, you have a lot of motivation to stay active. And so I would say this year is probably one of the more active years for me personally. Ran a 5K, been doing 5K training, which is a new thing for me because my background is football and sprinting more sprint short term type of workouts, but I've shifted my training this year and started to think a little bit more about 5Ks.
Rachid (01:23.422) Now, for those who haven't heard the previous episode with you, could you just please explain your role at Google when it comes to heart rate tracking?
DeCarlos Love (01:31.663) Yeah, so I lead a team that works on our fitness experiences, as well as some of our algorithms on the Pixel Watch. And so my team's role in that effort is really, how do we set the KPIs on what makes a good heart rate? How do we make sure that our heart rate performs well for our users and what people will actually do with it? And then we're working hand-in-hand with our algo engineer - engineers across the different stacks, whether that's on our device side or our system software side and really trying to help bring that heart rate all the way through from the low level to the actual user experience to our research, our testing, our validation. Just kind of being that coach, that quarterback with all of those great and very talented folks that we get a chance to work with.
Rachid (02:27.758) So I'm wondering what would you say is the essence of heart rate tracking?
DeCarlos Love (02:34.323) That's a good question. I think the essence of heart rate tracking is understanding how people will use the devices where heart rate is going to be applied. And I think that forms the very initial essence of what we need to think about when it comes to heart rate. Are we designing something for people that are going to be sitting down at their desk every day? Are we designing something for people that wanna walk a little bit? Or are we designing for people that are gonna be you know, getting that heart rate up, doing those intense and vigorous movements and activities. And I think that having that understanding really forms the essence, you know, from how we think about planning our research or validation, bringing people in the lab, our data collection, how we think about training our models, all of those things kind of form that from a top level. I think the other part of it is really gets down into, you know, the hardware, which is something that I think we're excited about when we think about Pixel Watch 2 this year.
Rachid (03:35.126) Any smartwatch now can do hardware tracking, right? So what sets Pixel Watch apart from the rest you feel?
DeCarlos Love (03:47.755) I think that the Pixel Watch is definitely set apart because of where we have the opportunity to combine the expertise of Fitbit along with the expertise of Google. I think those two things really form a unique combination. You have several years of experience of building health and fitness products, building wearables within a Fitbit organization, working on algorithms, being one of the first to bring heart rate to scale on the wrist, and then you combine that with the deep understanding of data, the deep expertise in machine learning and AI. And it's like those two things combined together are a really unique aspect of how we think about developing heart rate experiences and wearables in general.
Rachid (04:36.866) So last time we spoke, the original Pixel Watch was just out, and I'm wondering what the biggest difference is between that watch and the brand new Pixel Watch 2 in terms of how they track your heart rate.
DeCarlos Love (04:50.287) Yeah, so when we think about going from, Pixel Watch 1 to Pixel Watch 2, heart rate is at the center of it. We know and understand that people are gravitated towards wearables because they care about and want to understand more about their health and fitness. And heart rate sits at the center of that for us. And so we landed a really solid foundation with the Pixel Watch 1, the first time that we, you know, employed machine learning at a full capacity.
And so when we came to Pixel Watch 2, we still wanted to understand like, what are ways that we can improve on that heart rate experience? And so we approached it really about from the ground up and literally from scratch. And we took the idea of several different prototypes, you know, multiple iterations, and even some things that were further out on our roadmap. We needed to play some well-calculated bets to get us to where we are right now.
And the core of that is really thinking about how do we update and change the actual sensor and the hardware. On PixelWatch 1, we really started from a single LED photodiode path. But PixelWatch 2, we took that a lot further and thought about how can we increase the data that we have coming in, the surface area that we can cover on the wrist. In doing so, we went with a multi-path approach.
So our sensor, instead of just having a one-to-one mapping of LEDs and photodiodes, we have multi-pass there. We actually are using 10x the amount of inputs in terms of LED and photodiode pass that are going into this algorithm. And so when you look at the...
Rachid (06:33.43) So heart rate tracking, that green blinking light that sort of shoots against your skin. And then there's another sensor that looks at how that light is reflected. And that sort of tells you how the heart is...pumping the blood through the veins basically. I mean, that's probably... That will not get me into college, if I'd explain it that way, but... Is that sort of like the basics of it?
DeCarlos Love (07:13.327) Yeah, so we use a PPG sensor. And within there, you have a couple of key components. You have your LEDs, which shine the light. In our case, we're shining the green light from an LED. And then we have photo diodes that help us pick up the reflection of that light from the blood and helps us understand the heart rate from that reflection. And so that's super gross simplification. But the idea of shining light, understanding the reflection of that and using that to understand what the heart rate is. And so in the world of PixelWatch 1, we really had one pathway, one light shining, one picking the light reflection back up. And the challenge with that is that sometimes the wearable is going to be on different positions on a user's wrist. And so if you think about when you're running or when you're rowing or you're doing some type of HIIT exercise, like your wrist, your arm, your hands are in different positions. And sometimes that watch will tilt, it will rock. The different vibrations may cause different reactions in those scenarios. And so when you only have one shot at getting that LED right, like you're not maximizing the opportunity, you're not maximizing the coverage on the wrist. And so when you examine the actual hardware of Pixel Watch 2, you'll see that we have an array of those LEDs. And so there's a lot more light that can actually shine when that is actually in motion. And so that gives us an opportunity to cover more area on the wrist and gives us a better shot at getting more data into our algorithms. And that's what our design for Pixel Watch 2 really shines. No pun intended there, but because we have those, you know.
DeCarlos Love (09:07.619) Add more data to our algorithms, which presents another set of challenges, right? It's like now you have a lot more data coming in. Now you need to figure out what to do with that. And that's really where we then went into another phase of changes and evolutions in terms of how we make our algorithms smart enough to handle that vast amount of data that's coming in from the new sensor.
Rachid (09:32.286) You’ve got to sort of figure out which of the data is to be trusted maybe and which isn't because there is a chance that one of those paths is blocked maybe by, I don't know, maybe sweat or hair on my arm, for example.
DeCarlos Love (09:47.003) Exactly. Each of those sensors could be giving something a little bit different depending on their position on the wrist or where they are and what's happening. Some people may have tattoos or some people, light might be shining in a different way. And so the different sensors may give off different readings. And the role of the improved algorithm is really to help make sense of that. And that's a lot of things that we did. A lot of our explorations on the algo side, we're really about how do we figure out the right way and the right combination to use of the new sensors and the new LED photo diodes paths that we have coming in.
Rachid (10:28.21) I'm looking at my Pixel Watch 2 right now and I see one blinking LED. Is that because I'm not in active training?
DeCarlos Love (10:35.287) Yeah, so if you turn that into an exercise and you actually start an exercise where the movement kicks in, you'll see the full path actually come through. So you'll see all the lights start to shine. But why you're seeing one right now is you're sitting down, you're not in an exercise. And this is actually another part of the challenge that we have to think about is, yeah, we have more LEDs, we can shine more light. But that also could mean that there's more power being consumed. And this is where we've had to do things like tune our models so that we can be more power efficient.
Rachid (12:05.846) That's really clever. I was just, you know, while you were talking, trying this, because I only saw just one LED, so I went into a crossfit workout. But then indeed, yes, all the lights start to come up. So it's funny how these smart things you don't see it because it's at the bottom of your watch while wearing it.
DeCarlos Love (12:39.779) Yeah, I think you actually hit on a really important part, right? Are things that are going on that people don't see. And that's actually, you know, a good thing for us is that there's a lot of interesting technology. There's a ton of, you know, smart choices that are made in design architecture in order to make the heart rate even more accurate. But the most important part is that these things should kind of blend into the background and not really force a user to understand, know what is going on, and really just make it so that the experience is a part that comes through. The ID on the device is really something that we're proud of. And so we want that to be the thing that the user focuses on. And then also as they interact with the device, the different experiences from the software and our UX are the things to focus on. And...Yeah, trust and be assured that there's a ton of things going on inside of that watch that may not always come through, but they're important in making that experience the best it can be.
Rachid (13:44.758) You mentioned the lab before. Now, let's say I would have been a subject in your lab maybe six months ago. What would you have me do to make a better PixelWatch 2?
DeCarlos Love (15:05.603) Yeah, so when we think about collecting data, we really kind of think about it in three key different types of data collection. I think one of them is definitely more of our structured protocol that we do in the lab. In that case, we'll have people do a variety of different exercises and movements, everything from running to cycling to some HIIT stuff, but several different exercises that we think about.
How does that correspond to what our users are doing and what we learn from the data as the top activities combined with some of the activities that we know are challenging for collecting heart rate. And so that one is the lab. I think two, we look at it as more in a real world scenario. And how do we make sure that we're collecting data and we can compare that data to ground truth, but we do so more in...you know, it's not a control. It's people, we're monitoring people's activities and we have access to, you know, thousands of people in this format, but we still have a good idea of how they're performing. And then I'd say the third one is just a little bit more, just like a little bit more dog food free for all. And, you know, each of those are really important for helping us get a gauge on the accuracy, the quality of our heart rate tracking, as well as our other algorithms. But each one of those plays a different role. And so if you were coming into the lab, we would give you, you know, the forms to sign and let you know that it's going to require a good amount of physical activity. And so be prepared, because it's important that we capture that. And it can take a lot, you know, several times several different trips into the lab, in order for us to continue to improve our data collection.
Rachid (16:52.066) Sounds like a lot of fun a day in a lab, so I might sign up one day. You mentioned dog fooding, which doesn't actually have anything to do literally with dog food. Maybe you can quickly explain what that is in Google language.
DeCarlos Love (17:11.319) Yeah, so it's something that is really inside of the Google culture of making sure that we as Googlers are testing our products. And so every team within Google runs their own form of a dog food. And it's an opportunity for us to really pressure test and go be those testers to give the feedback from an actual user's perspective, not just within the team.
If it just works for me, that might be okay, but if thousands of other people don't also think so, then it gives us informed feedback on what are the types of changes and things that we need to make.
Rachid (17:52.246) Yeah, it's definitely important to eat our own dog food. That's for sure. DeCarlos, I'm wondering, so you work on the algorithms, the hardware, you get the heart rate tracking right. It will tell me what my heart rate is while running or while not running, it will tell me my resting heart rate. But I guess it's also of essence to make sure it's communicated in a way to our users that it's useful information to them. So what does Pixel Watch 2 do together with Fitbit to actually make that useful information?
DeCarlos Love (18:28.835) There are a couple of key things that this year with PixelWatch 2, that we've been able to do. And so I think one is applying similar algorithm techniques, data collection to give us other features. And so we're introducing activity recognition on PixelWatch 2 this year, but we're leveraging some of the same email techniques that we've done from a heart rate perspective to introduce that feature. So that's one tangentially related aspect.
I think two is kind of what you alluded to, which is, hey, the heart rate is accurate. You have a brand new sensor from the ground up, but what are we going to do with that? The thing is that it gives us more confidence about delivering even more unique experiences than what we had on the Pixel Watch 1. We're introducing the heart rate zone coaching, but we're also changing the overall experience in the UX with the inside of our exercise app. And really a lot of that helps with the things like heart rate coaching so that we can have these interstitials that pop up on the watch to tell you if you should go a little bit harder or ease back a bit. But we're also being super intentional about everything from the font type, the font size, the spacing. And we want to make sure that we're to maximize the spacing. And so users will get a chance to understand that, every pixel was thought about on how we make this better with Pixel Watch 2 coming from Pixel Watch 1. And so really proud about in a similar way, the theme of like, hey, we started from scratch on designing our hardware, but even from a user experience standpoint, it's like we have even more confidence in that accuracy. How do we make the UI as best as can be? And so our UX team; You imagine walls of different designs plastered up, really sweated every pixel to make sure that the experience that the user was directly interacting with was going to be one that matched the idea of like, we wanna make this the best that it can be.
Rachid (23:08.318) Amazing and it has helped me slow down a little bit on the 5k to keep the heart rate in check. Definitely need some more training there. Now DeCarlos, we always end our episodes with a top tip for the road. So let's say someone recently bought a Pixel Watch 2, maybe like you and me, they're into 5k running. What kind of things would you recommend they try with Pixel Watch 2?
DeCarlos Love (27:15.815) Yeah, so obviously we talked about the improved heart rate and so that sits at the center of what we're doing. We touched on a little bit of, you know, more than a little bit, but we touched on the idea that heart rate coaching is something that's new with Pixel Watch 2, as well as our pace coaching and pace trainer features. I think another just from a general tip is that those things are going to help you with your improving your overall fitness.
But even for me, what it did was it helped me really just focus on the running and enjoying the running. As someone that came from competitive sports, what it allowed me to do is I don't have to always go as hard as I used to, or maybe I think I should, or learning when I was going too hard. And so it really helped me change my relationship with running and make it more enjoyable and really get me to focus on just the aspects of being outside and enjoying feeling the of your feet hitting the ground and getting the joy from those things. And so I just really kind of started to reshape my personal relationship with it. And so the tip is just like really lean into the experience, this experience that kind of helps nudge you, helps guide you. And it's really about making sure that you are focused on the thing that matters for you, whether it's you want to go as hard as you can, cause you want to improve that faster 5k, 10k time. But if you just want to enjoy, you know, being outside and enjoy feeling kind of the rush from the blood pumping from doing the exercise. The Pixel Watch 2 is really gonna be right there with you for that journey.
Rachid (28:54.122) Amazing. And that's just a great general tip. Just enjoy it and Pixel Watch 2 will take care of the rest.
DeCarlos Love (29:00.111) Yeah, and we hope everyone enjoys it and look forward to feedback and getting folks getting a chance to use the product.
Rachid (25:59.438) Thanks a lot and hope to have you on the part again soon.
DeCarlos Love (26:02.447) Absolutely. I'll come back anytime. Thank you.
Compared to previous Fitbit devices. Performance of heart rate tracking may be affected by physiology, location of device and your movements and activity.