Eatery: Diet Tracking with Pics and Opinion

For this week’s module on diet tracking, I’ve decided to look at Eatery, an iOS app developed by Massive Health prior to their purchase by Jawbone in 2013. Eatery provides users with a tool to track, visualize and receive feedback about their diet via a simple camera and rating interface. Users snap a picture of their meal and assign a single rating of ‘healthiness’. Other people within a user’s network can review recorded meals and assign their own ratings of that meal’s ‘healthiness’. The app was built upon principles of Social Learning Theory, and designed to leverage differences in perceptions of ‘healthiness’ to prompt thought and learning about diet. The simplicity of the inputs and interfaces allowed Eatery to capture an impressive corpus of data, with 7.5+ million ratings from 50+ countries during the app’s first five months.

A striking feature of Eatery is that it relies upon subjective measures of a meal’s ‘healthiness’. By distilling information about a meal down to this single, simple metric, Eatery allows itself to avoid issues of numeracy and nutrient content and cut straight to differences in perception that may be more effective in challenging and changing behavior. Whether the user is operating in isolation and viewing their own ‘healthiness’ ratings without comparative inputs, or receiving anonymous ratings feedback from others in their network, they are presented with a representation of their diet that offers insights into patterns that they may not have otherwise understood. The underlying assumption applied in the app’s design is that highly subjective but relatable ‘healthiness’ will be more impactful than a devised and unfamiliar index of objective nutritional measures.

An interesting element of the Eatery story is that, after gathering submissions for six months, Massive Health released their data set to researchers and published a series of statistic-rich infographics. An interesting element of these infographics was an emphasis on the power of statistical methods to pull insights out of large, noisey data sets. Their materials invoke Halvey and Norvig’s “Unreasonable Effectiveness of Data” in praise of large data sets¹, while citing Sunstein on the power of averages and the wisdom of the crowd.² While I might have also liked to see some validation of user opinion versus nutritional science, or information on user outcomes, I found Eatery’s analysis to be refreshingly concrete, and a welcome compliment to the subjectivity of the data gathering process. They showed strong correlations between the ‘healthiness’ ratings of users’ meals and independent measures of obesity for several large US cities, and were able to show social network effects of dietary ‘healthiness’ that mirrored prior scientific findings. Their data set is now open to third-party research, and I look forward to seeing whether other investigators are able to expand, validate or challenge that first round of analysis.

For the purposes of this class, I think that Eatery can provide two particularly useful lessons. The first is an example of how to focus on relatable methods and measures. Taking a photo and getting opinions may not provide rigorous documentation or analysis of diet, but it offers a low barrier to entry and relatable feedback. A second important lesson is the value of analytics and evaluation. Eatery may have been dismissed as narcissistic foodie-ware if it hadn’t been backed up by a series of reflections and representations of data that communicate the value of their methods. This suggests that planning for scalable data collection and investing in analysis and communications are important considerations for the design of tools for well-being.