WHOOP and the gamification of health
Plus, how wearables may shift the future of healthcare.
Welcome to Day to Data!
Did a friend send you this newsletter? Be sure to subscribe here to get a weekly post covering the technology behind products we use every day and interesting applications of data science for folks with no tech background.
I’ve been running for years now and there’s always a new company coming forward, promising to provide the best insights on your training and how you can improve your performance. Today, let’s talk about WHOOP.
Founded in 2012, WHOOP aims to be your personal digital fitness and health coach. The company offers a wrist-worn wearable that provides in-depth analytics around sleep, respiratory health, training, and more. The wearable, that features no screen, is built for data collection. The lack of screen is intentional, as the battery that would traditionally go towards powering a screen is allocated towards improving the device’s data collection. The streamlining of the product mimics the company’s goals with the device.
In a conversation with the SVP of Data Science and Research at WHOOP, Emily Capodilupo sits down to talk about the evolution of data at WHOOP and how data has informed many of the features they’ve built out for users. WHOOP makes for a great case study around building out a platform that is based in data and analytics, but requires signing on thousands of users and time for the platform to become its most valuable. Using data science to inform your product roadmap can improve adoption down the road as past behavior can help uncover future success.
Building data-driven features for users
Capodilupo says the work that excites her most is when she gets to ask “what stuff that nobody’s thought about is our data capable of providing insight to?”. With WHOOP’s incredibly vast datasets, there are plenty of problems to explore. At this point in the process, I will present you with a Michelangelo quote you probably didn’t expect in a data science blog:
“The sculpture is already complete within the marble block, before I start my work. It is already there, I just have to chisel away the superfluous material.”
In a way, you can think of data science like this. The features to build and patterns to detect are within the data, but as Capodilupo says, her team has to think about what the data is capable of becoming that they haven’t even thought about. In another interview, she mentions that at the origin, WHOOP was focused on preventing overtraining. After digging into the data they saw from customers that they planned to use for evaluating athletic performance, they saw that their algorithms could also support areas like cognitive and emotional performance as well. The trove of data they saw coming in from their users allowed them to see where new features could be built out. Her data teams are leaning into the story the data is telling, rather than massaging the data until it exposes a solution to the problem at hand.
The power of wearables
WHOOP requires the use of their hardware, a wearable data collection device with no screen that can be worn on the wrist, bicep, or other spots on the user. The wearable is capable of being charged wirelessly, which enables WHOOP to get 24/7 data collections, unlike many devices like a Garmin watch, which require the user to remove the watch for several hours to charge, losing valuable data. The value of data of this breadth is garnering the attention of the healthcare industry. In the video interview above, Capodilupo talks about how a wearable like WHOOP differs from a traditional medical device, and its ability to create quite the complete view of a patient’s data history is something that healthcare partners are interested in. From assisting in the early detection of COVID-19 to discovering an abnormal heart rate, comprehensive patient data can help expand precision health. Let’s look at some research.
Research using WHOOP’s hardware
Clinical research has been done to validate the performance of a WHOOP strap compared to other medical devices, and also to evaluate the potential of integrating WHOOP and its data into existing patient processes.
Some studies include:
A validation study of the WHOOP strap against polysomnography to assess sleep
Findings: WHOOP is a good tool to measure sleep in field settings (i.e. at home) when polysomnography (a sleep study) is not feasible
Note: this was completed by the researchers at WHOOP in collaboration with the Central Queensland University, Appleton Institute for Behavioral Science in Adelaide, South Australia
Findings: Sleep and heart rate variability data collection “may potentially be a noninvasive method for monitoring cognitive changes related to pre-clinical AD [Alzheimer’s Dementia]”
Using data to unlock human potential
WHOOP has become a market leader among athletes, including Lebron James and Michael Phelps. The team is continuing to find new opportunities for how health data can improve athletic performance, as well as cognitive and emotional performance. The company’s success hinges on their data teams finding new ways to provide user value while also improving the accuracy of their existing features. I’m eager to see how wearables become, or don’t become, a part of the patient experience in the future, and if WHOOP plays a part in this potential evolution. The key to living a healthy life may just be hidden in the data.
I think that the precise measurement of multiple metrics as to the impact of exercise on both health and longevity may become one of the greatest blessings of "big data." The more data the merrier, if only the analysis is through and unbiased. The world could use that. We have plenty of calories for input and plenty of possible metrics on caloric output for the human body; the data just need to be conveniently and better measured, than analyzed well.
And while 'data' has apparently become universally accepted as a singular noun, I think that remembering and using 'data' as a plural noun greatly reinforces the idea that the data can consist of billions of discrete, analyzable items rather than view data as an indiscrete blob of a collective noun. Plus this exchange helps to remind people of the significant difference between "discrete" and "discreet". :)