Smarter Eating, Smarter Fitness: AI supports your goals

We eat and exercise and hope to stay fit. MyFitnessPal takes the guesswork out of the equation so you can get the results you want.

Physical fitness is it an outcome of a process that is dependent on many variables- some which we could control, and others which we do not. Variables which at first glance seem to be under our direct control include the food and levels of physical activity. However, a quick examination of these two levers reveals the illusion of control. We seldom remember the quantities we eat and often forget the exact calorie count. We make spur of the moment decisions without accounting for the nutritional impact of a cheat day on our overall wellbeing. We have no way of objectively knowing if it’s better to do 20 press ups today and then rest or do 25 continuously.

 

Founded in 2005 by Albert Lee and Mike Lee1, MyFitnessPal is a website and mobile app platform that attempts to solve the fitness challenge by tracking both diet and exercise patterns of users. It synchronizes with wearable technology products and other tracking apps to allow all exercise information to be uploaded from multiple sources.2 MyFitnessPal analyzes exercise and dietary patterns and converts the data into meaningful information which enables users to make better decisions about their overall health and fitness. In 2015, MyFitnessPal was acquired by Under Armour3 and added to its Connected Fitness lineup, an umbrella through which Under Armour is able to reach 160 million users3. MyFitnessPal added 80 million4 users into the Under Armour fold, along with new technology-based opportunities for the sports apparel manufacturer.

The role of machine learning within MyFitnessPal has evolved as the company has grown from a startup. Initially, MyFitnessPal crowdsourced food information, relying on individuals to input data about their diet into the system5. Machine learning about the individual user allowed the platform to find patterns in the user’s dietary habits- thus, it would start to maintain a record for frequently eaten food and recently eaten food to speed up future data entry. MyFitnessPal made it easy to input data into the platform by pre-loading a lot of information about nutritional value available in food, e.g. a sandwich at Starbucks. 6 After lowering some of the barriers to regular input of dietary information, the next machine learning challenge the MyFitnessPal took on was process of creating “verified foods”. Based on having a large number of entries of the nutritional value of a certain type of food, MyFitnessPal could more accurately predict the actual calorific contribution and eliminate the effect of human error in data entry. According to Chul Lee, the unit’s head of data engineering and science, MyFitnessPal also ran a food categorization project as part of training its AI7.

Most recently, under Under Armour, is leveraging data from MyFitnessPal through a partnership with IBM Watson to ensure that users of this app platform can access research-based recommendations on how to improve their sleep, exercise and nutrition2. The machine learning megatrend is significant for MyFitnessPal because it allows the platform to become what it was named for: a true fitness companion that is better at supporting a user’s fitness levels and health.

MyFitnessPal has shown that it can use data science to help users to consistently take the guesswork out of their dietary process. Having overcome some of the challenges associated with the fallibility of human memory, and the amount of analysis need to convert data from food labels into meaningful information, MyFitnessPal platform can go further to support users by integrating with other suppliers in the health journey. The online platform retains your information longer than any personal trainer could and could even serve as a bridge when you move between personal trainers. One could envision a world where MyFitnessPal discovered your dietary patterns and integrated with e-commerce platforms like Amazon to ensure that the foods you ate often were ordered and delivered. Using your body type and composition alongside data from their thousands of other users, MyFitnessPal could predict the likelihood of success of fitness trends e.g. ketogenic diets, HIIT or mindfulness. The exercise arm of MyFitnessPal could integrate with health insurance platforms and allow users who engaged in more health-seeking behaviors to receive reductions in premiums or rebates as a tangible reward to incentive self-care.

 

As we think through the possible future applications of machine learning in the digital fitness space, two key questions remain that need to be considered. The world of personal fitness relies on trust. First, how might user information, gathered over extensive periods of time, remain secure as different companies share this information in a bid to improve predictive power? Secondly, how might supervised and unsupervised learning process change as we extend AI-supported fitness products from humans to their pets? (772 words)

Footnotes

  1. Crunchbase database, “MyFitnessPal” https://www.crunchbase.com/organization/myfitnesspal#section-overview, Accessed November 13, 2018.
  2. MyFitnessPal Company, “Guide: How to properly sync among MFP/Fitbit/Garmin Connect/Strava” https://community.myfitnesspal.com/en/discussion/10374823/guide-how-to-properly-sync-among-mfp-fitbit-garmin-connect-strava, Accessed November 13, 2018.
  3. Nanette Byrnes, “AI hits the mainstream” https://www.technologyreview.com/s/600986/ai-hits-the-mainstream/, March 28, 2016, Accessed 13th November 2018
  4. Databricks, “Customer Case Study MyFitnessPal” http://cdn2.hubspot.net/hubfs/438089/case-studies/Databricks_Case_Study_MyFitnessPal-07012015.pdf, 2016, Accessed November 13, 2018.
  5. Dow Jones Institutional News; “Spark, a Tool at Big Data’s Cutting Edge, Helps Under Armour Perform Faster Analytics” https://search-proquest-com.ezp-prod1.hul.harvard.edu/businesspremium/docview/2065409796/68F3E2DD5B3847F8PQ/1?accountid=1131103 June 2015, Accessed November 13, 2018.
  6. MyFitnessPalApp, “Introduction to MyFitnessPal.” YouTube, published Aug 7, 2012,

https://www.youtube.com/watch?v=fu9RKqlmD1Q, Accessed 13th November 2018

  1. Data Science Weekly, “MyFitnessPal – Data Science to improve Health & Fitness: Chul Lee Interview,” https://www.datascienceweekly.org/data-scientist-interviews/myfitnesspal-data-science-improve-health-fitness-chul-lee-interview, Accessed November 13, 2018.
  2. Neil Versel, “Under Armour buys MyFitnessPal, Endomondo for $560 million, ” February 05, 2015 https://www.aiin.healthcare/topics/business-intelligence/under-armour-buys-myfitnesspal-endomondo-560-million, Accessed November 13, 2018.

 

 

Previous:

Uber: Future challenges in the era of AI and ML

Next:

Betabrand: Wearing Product Development on its Sleeve

7 thoughts on “Smarter Eating, Smarter Fitness: AI supports your goals

  1. Really enjoyed reading your thoughts about the potential of My Fitness Pal and how it could integrate with so many other apps to provide better insights and customized service to the user. I especially love your idea for the company to use collective learning from the cumulative data of other users to derive correlations between health benefits and user goals. In terms of data security, I think users would be willing to allow their data to be used if it could be done in an anonymous way and if they could also benefit from the shared learning. I would probably give users the ability to opt into sharing their data (and receiving the benefit of the collective intelligence) or keep their data private and have a more manual and less sophisticated user experience.

  2. This was a really well-written piece on applications of machine learning in the digital fitness space. I find the journey of MyFitnessPal fascinating as it has grown from a start up to one of the biggest players in the digital fitness space. In addition, its partnership with Watson is particularly interesting as it shows a clear interest in R&D in the machine learning arena. In response to your questions, I think that as long as there are appropriate measures taken when collecting user data to keep it anonymous and at the aggregate level, the data can be safely used for its predictive capabilities.

  3. Very interesting read Sophie!

    I agree that data security will be an increasingly controversial topic, as machine learning and other technological innovations develop. It seems the answer to the issue lies in technology. New developments like block-chain encryption can help solve the most immediate problems, and further down the line, more sophisticated methods of data storage will prevent its theft.

    Perhaps the more pressing question is how will citizens adapt to a landscape where privacy takes a new meaning. Today your private self is no longer who you are or what you do, but the parts of you that are not online. As consumers become increasingly aware of the scarcity of their private data I wonder how the internet landscape or our social structure will transform.

  4. I found it very interesting learning about how machine learning can be used for such routine activities as dieting and exercising. People often think about the used of this technologies in a highly academic environment, but the applications in our daily life are endless.
    Regarding the issue if data security, I believe it is a problem that eventually all companies will face – as we increase the availability of data online, both for storage purposes or sharing and crowdsourcing analysis and solutions, the risk of leakage and improper use of this information will increase, no matter how much we develop new security softwares and policies – given “hackers” will also develop new ways of breaching the security measures.
    Of course developing new encryption methods is crucial, but I believe companies will have to face the trade-off of taking this risk versus the benefits this data availability can bring to their business and their clients

  5. I had never realized that MyFitnessPal used machine learning in order to analyze caloric content of foods. I was always under the impression that all of this information was pre-loaded into the system. It is interesting to see the extent that AIs can be used even in industries that have simple mathematical concepts like diet and exercise, where in order to lose weight, calories in = calories out.

    While you bring up an interesting point about users having reservations in others gathering data about them, I believe that most would be ok with MyFitnessPal obtaining this data. As shown by the recent Facebook data breach scandal, it appears that as long as the use of this data doesn’t have an immediate harmful effect on the user, most users are willing to share this data. In and of itself, however, data on what type of food users consume is less dangerous, even if someone is able to breach security measures and trace it to individual users. However, data on where users eat, what time they eat, and the amount of money spend on what they eat, is significantly more critical in preserving the privacy of the end user, while this data is also most valuable to MyFitnessPal. The question remains will MyFitnessPal be content to only store data on amount of food and caloric intake of it’s users in order to improve their health, or will they venture out to collect as much data as possible about the user, which puts the privacy of the user at risk?

  6. This app poses a lot of interesting questions for the future, some concerning some exciting. You get to the crux of the main issue which is the use of such data for reasons other than your public health. If this were purchased by a health insurance firm for instance it could lead to higher premiums for people with health issues. Then again, it could also be an opt in program, like already exists for others as a way to try and reduce your premiums by living a healthy lifestyle. It’s also intriguing to think of the applications for large scale consumer goods from companies such as amazon

  7. Thank you for the article – it was extremely well written and really interesting to learn about how machine learning is being leveraged by MyFitnessPal.

    Your question regarding a tremendous amount of user data is a very relevant one given recent topical scandals with Facebook, Macy’s, LinkedIn, etc. My initial thought was regarding what type of data breach would affect the users. If this was a breach of their payment details for premium users, that would of course be detrimental, however, a breach of a user’s eating and exercise habits may not seem as harmful as that of other companies leaking information that can cause immediate issues to the user. That said, ruminating on your question, I do think they should prioritize protecting users information if the data can link to food purchases and or exercise equipment usage that can lead to unsolicited marketing or promotion from companies in the space.

Leave a comment