The Knewton Blog

Our monthly newsletter features edtech and product updates, with a healthy dose of fun Knerd news.

Knerding Out: How Data Can Optimize Commute Time

Posted in Knerds on May 21, 2014 by

Commutes are the worst. They hurt job satiesfaction, and they cause stress and illness, depression, and even divorce. That’s before we even get to lost productivity — every minute we shave off commutes in America is worth about $20 billion annually.

When Knewton was evaluating potential office locations, the issue of commute time was one of the most important decision criteria. As a data company, we decided to examine this criteria in a data-driven way.

Thanks to the MTA (who makes its data available for developers and the public) and OpenTripPlanner, an open-source project, we were able to do that in a couple of short(-ish) steps.

First, we took a look at where Knewton employees tend to live. We’re a fairly spread-out bunch. There are large contingents in Lower Manhattan and Downtown Brooklyn, but plenty of Knerds also represent Williamsburg, Greenpoint, Long Island City, Astoria, the Upper East and West Sides, and New Jersey.

 https://s3.amazonaws.com/uploads.hipchat.com/15040/413071/FSWQYFjycfqymrQ/wherewelive3.png

Knerd density map. Knerds live in the areas highlighted red and yellow, with the red areas representing the highest concentrations (maps drawn with qGIS).

Next, we made a map to visualize how easy it is for Knerds to get around various parts of the city. Since every Knerd has a different commute to each point in the city, we compromised here by looking at the medians for each point.

https://s3.amazonaws.com/uploads.hipchat.com/15040/413071/JhtpD32ytu4XhmI/commute_times.png

Knerd commute time map

The deep red areas represent parts of the city where the median Knerd commute time is reasonable (under 45 minutes). The lighter red and yellow areas have commute times closer to an hour, while blue represents commute times up to 2 hours. The unshaded regions were not included in the calculation (which we performed with OpenTripPlanner).

This visualization helped rule out neighborhoods like DUMBO, which have plenty of promising office space but would cost too much in commute time. Much of lower Manhattan, however, is totally viable (along with a few other areas, e.g., Downtown Brooklyn and Long Island City).

Having narrowed the research a bit, we then took a closer look at a few of the top contenders. These potential locations included buildings in the Financial District, SoHo, Union Square (our current location), and Flatiron.

For these targeted locations, we looked at the full distribution of commute times for all the Knerds (since we didn’t want to pick a location which worked fine for most Knerds but left a few with terrible commutes). What we found was that three of the top contenders were very similar in their distribution of commute times, but the fourth (in the Financial District) had a distribution that tended to be longer to a very obvious degree. While the Financial District location would still yield a median commute of less than 45 minutes (just barely), the other locations actually give us about a 30 minute median commute (and tails that extend less far) — which means we wouldn’t be rewarding the majority at the cost of punishing a few if we went with that location.

Distributions of Knerd commute times for our candidate office locations

 

 Zack joined Knewton following years of training and experience in computational modeling, statistics, and neuroscience. He has published highly-cited work in bioinformatics as well as in top neuroscience journals, and has presented at multiple conferences in topics dealing with the analysis and modeling of complex biological phenomena. Most recently before joining Knewton, he was completing a Ph.D. at Weill Cornell Medical College, where he was advised by Jonathan Victor and Sheila Nirenberg while studying the brain’s encoding of visual information. During his Ph.D. research, Zack also found time to consult on small projects he found interesting, including a visualization for the European Academy of Bolzano of how new train routes in the Alps region will affect Europe’s commuters.