Mobile transportation apps are reshaping the data available to transportation planners and policymakers when it comes to active transportation. We’ve previously looked at Strava, an app which tracks mostly recreational trips via walking, running, and biking. Another new app, Human, goes beyond Strava to track all user trips, which has transportation planners excited about the possibilities for better data.

Human comes closer to filling the frustrating void in transportation data for walking and bicycling. As Streetsblog proclaims: “The most popular bicycle transportation measurement system in the country (the US Census) is hopelessly skewed toward a niche activity: going to work.” There is currently no data available that comes close to measuring the mode share of walking and biking in cities. As such, it is easy to construct narratives such as “nobody walks in LA” or that investments in walking and biking are “non-essential” and that street space and/or funding should instead be directed elsewhere (typically toward driving). The reality, of course, is far different, but insufficient data exists to measure the magnitude of walking and biking for all trips in our transportation networks, or which streets experience the most pedestrian and bicycle activity.

Human partially fills a void in user-generated transportation data. By tracking all user-trips, its data is less skewed compared to apps like Strava and more reflective of a typical mix of trips. By average trip duration, 37 percent of trips on Human are walking, 10 percent are biking, 4 percent are running, and 50 percent are motorized (which may include driving and riding transit). The app makes use of the M7 chip, which lets users track their movements even when their iPhones are asleep.

Human, like Strava, has its limitations. A look into the data (below) suggests that Human oversamples data from a young, white, well-off user base and gathers little data from poor, minority, or elderly populations. As a result, the data is highly skewed and not an entirely accurate reflection of citywide transportation patterns. However, it still provides some useful insights.

Let’s dig into the data to see how it works.


Human generates maps for walking, bicycling, running, and motorized travel in 30 cities across the world. You can dig into data for cities ranging from Amsterdam to Bangkok to Rio de Janeiro to Washington D.C.


Human shows most of the bicycle activity occurring in the northeast corner of the city, which tends to be the youngest, whitest, and highest-income area (also the epicenter of the city’s tech boom and likely the biggest user base for Human). While it is true that more bicycling may occur in areas like the Mission than in the Excelsior or Sunset districts, it appears that these areas (and particular populations within these areas) are oversampled compared to users in other parts of the city.

Nevertheless, the data does have some interesting things to say about San Francisco’s bike network. While areas with high-profile bike investments clearly stand out, including Market Street, the Wiggle, the Embarcadero, other routes with little bike infrastructure are also apparent. There appears to be high ridership for east-west routes in the Mission, including 14th, 15th, 16th, and 17th Streets, and their connection to Townsend Street. While this feature may be a consequence of over-sampling tech workers heading to Caltrain, it does reinforce the need to plan for a citywide network as opposed to a downtown-focused system.

Second, let’s check out the walking map for Los Angeles:


This map clearly shows how Los Angeles has a number of boulevards, neighborhoods, and activity centers that experience high levels of pedestrian activity – Santa Monica, Venice, Beverly Hills, Mid-City, Hollywood, Downtown, and Pasadena are all apparent. Building upon these successes is a key objective of Mayor Garcetti’s Great Streets Initiative. However, this map greatly understates the walking patterns of the region’s lower income and minority populations: dense neighborhoods with high pedestrian concentrations like Westlake, Boyle Heights, or Leimert Park do not even show up on the map. It’s clear Human’s data has its limitations due to its user base.

Mode Share Comparisons

Human provides city-by-city mode share data which, again, is interesting, but can be somewhat of an apples-and-oranges comparison. The data for American cities is below:


It’s unclear how many conclusions can be made from this data given both the skewed sample of the data as well as the possible inconsistencies of city/region boundaries. For example, Human maps almost the entire Los Angeles basin, and presumably calculates its mode share data from the entire area rather than the City of Los Angeles. An area as large and varied as greater Los Angeles will have more motorized trips for longer durations compared to the City of San Francisco, which is shown to only include the city proper (an area less than 1/30th the size of greater Los Angeles). This may create some inconsistencies in comparing ‘cities’ to ‘cities and their suburbs.’ For these reasons, specific comparisons between cities appear difficult.

To add to the list of caveats: Human counts mode share by percentage of time spent traveling by each mode, as opposed to counting trip types. So if you take eight walking trips that are 10 minutes long and two motorized trips that are 40 minutes long, then your walking mode share is 50 percent, not 80 percent. This calculation skews the data compared to trip-counting mode share calculations that are typically provided.

Despite these issues, the general trends of the mode share data are about what you’d expect. Boston, New York City, San Francisco and Washington D.C. experience the most walking. More sprawling cities like Austin, Houston, Los Angeles, and Miami have less walking – but this data may be skewed by geography, demographics, and methodology as noted above. There’s not a whole lot of variance in biking and running, although San Francisco has a surprisingly high bike mode share (11 percent) compared to the official SFMTA estimate (3.5 percent) (PDF). This difference in mode share is again likely a result of skewed data.

Portland presents a perplexing case. A tech-savvy city widely considered the best in the U.S. for bicycling, Portlanders spend a paltry 2 percent of their time traveling on a bike (the city auditor’s office estimates 4 percent of all trips are by bike). In fact, Portland more closely resembles more sprawling cities like Houston. Portland might be slightly overrated when it comes to active transportation, but this data is clearly out of whack.

Human also packages each mode share value into international comparisons:


It’s no surprise that Amsterdam and Copenhagen are atop the cycling cities list – official trip count mode share figures for each city are around 38 percent and 36 percent, respectively. An estimated 15 percent of trips in Berlin are by bike as well. The presence of Rio de Janeiro near the top is surprising, however – anecdotal evidence suggests this figure might be highly overstated by a small or skewed sample.


The walking list aligns with what we think of great walking cities: Tokyo, New York City, London, and Paris are all near the top. Again Los Angeles appears to get unfairly penalized near the bottom.


Human erroneously labels motorized travel as “Car Cities” even though motorized travel may include cars, buses, and trains. What this list is really showing is spread out cities with long trip durations that can’t be made via walking or biking.


Human is a great tool that provides valuable data to add to the conversation for transportation planning and policymaking. But it is imperative to recognize its limitations: it measures travel patterns of a specific demographic and counts trips via duration, which yields very different data than just counting trips. It will be interesting to see how its data develops as the app becomes more popular, or if these limitations only grow more amplified. Like Strava, the U.S. Census, and other data sources, Human adds another piece to the transportation data puzzle; however, it is necessary to be very critical in reading its data before making conclusions for transportation planning or policymaking.