Blog Post

Can We Avoid Fitness-Tracking?

Can We Avoid Fitness-Tracking?

A recent report released by market intelligence firm Tractica indicates:


Tractica forecasts that the overall wearables market will grow from 17.0 million device shipments in 2013 to 187.2 million units by 2015, representing a compound annual growth rate of 34%.



Given this sharp rise in sales, will wearables become as ubiquitous as smartphones have become, or will they remain forever a niche product? To put a fine point on the issue: Can we avoid wearing wearables?


Right now, the best way to avoid the kinds of fitness-data collection that Fitbit and Jawbone Up generate is to simply not wear one. As is well-documented, Fitbit collects their users’ data in aggregate [Fitbit Privacy Policy]. 

What Data May be Shared With 3rd Parties?....

Data That Does Not Identify You (De-identified Data)

Fitbit may share or sell aggregated, de-identified data that does not identify you, with partners and the public in a variety of ways, such as by providing research or reports about health and fitness or as part of our Premium membership. When we provide this information, we perform appropriate procedures so that the data does not identify you and we contractually prohibit recipients of the data from re-identifying it back to you.  [source:]

Those who are concerned about privacy or surveillance might take solace in the privacy policies of Fitbit, which claims to not sell any data that is personally identifiable. However, as was recently made clear from the proposed sale of Radio Shack user-data [Washington Post US News] there are all kinds of perfectly legal ways that a company like Fitbit could transfer your data to another corporate entity. In the event of a merger or sale, your data is then governed by a new company’s privacy policy, one to which you never explicitly agreed. 


If this is cause for concern, let me push the point further. With infographics like this one published on March 24, 2015, it is clear that Fitbit is interested in using demographic data to draw certain conclusions about its users:



From "What Jobs to Active People Have [Infographic]," posted March 24 2015, Fitbit Blog. Link to full post and infographic.


First of all, I take issue with the title of this graphic. “People” is obviously too strong. “Fitbit users” is more accurate. But the reason that their confusion of “people” with “Fitbit users” is worrisome is because this is exactly the kind of conflation inherent to sloppy conclusions based on limited data. The slice of the population represented in this infographic is merely those who use Fitbits, not the population as a whole. And indeed, the comments for this blog post are mostly critical, and indicate dissatisfaction that some obvious job-types are left out, and with them large segments of the population.




I do not know how they collected this data (My request for more information in a comment is currently “awaiting moderation”). And I do not know of any statistics on what kinds of jobs are most common among Fitbit users. However, it it quite possible that the kinds of jobs left out—food service workers, cleaning company employees, craftsmen, postal workers— are not well-represented in the Fitbit set.


This is my point: in generating these statistics, Fitbit is creating categories that easily enter into typical “Big Data” analysis. Data brokers, as has been reported by Julie Angwin in her book Dragnet Nation, create categories of people (i.e. “affluent baby boomer,” “bible lifestyle,” and conclusions based on those categories like “estimated household income” ). These categories can be used for targeted marketing and other more nefarious purposes like denying loans or employment discrimination, as demonstrated by Frank Pasquale


By generating statistics based on its actual users, Fitbit makes it possible to generate probabilistic conclusions of non-users. Even if an individual does not wear a Fitbit, the fact that there is (presumably accurate) data about average numbers of steps in some professions means that data broker companies can estimate the number of steps that non-Fitbit users take as well.


And even if data brokers do not specifically know what a specific individual does for wor, they may be able to estimate that as well based on other data points. Because correlational analysis determines likely outcomes based on statistical regularities found in large data sets, the more data points a data broker has, the easier they can estimate the other data points.


If infographics like this one are indicative of much larger data sets that Fitbit has, it is reasonable to conclude that data brokers or other interested partys could use Fitbit’s aggregate data to estimate specific information about an individual based on other demographic data. Thus, while Fitbit’s privacy policy strictly prohibits personally-identifiable information, this requirement is rendered toothless by data broker practices that are legal and quite common.


In a real sense then, fitness-tracking affects us all, whether or not we wear a Fitbit or other tracking device. And if Tractica's market analysis is correct that sales of wearables will see an annual 34% jump, we can expect even more aggregate data that tells us even more about both users and non-users alike. 




No comments