Commute Elasticity and Google Timeline
The new Google Timeline dashboard and feature in Google Maps is very cool. In addition to answering useful things like "Where did we stay for that long weekend in New Hampshire last fall?" or "When was the last time I went to that restaurant?" or the general appeal of the automatic private journal, it makes it amazingly easy to measure commute elasticity.
By "commute elasticity", I mean how much better or worse a commute gets for every minute earlier or later you leave. The downloadable data from Location History is a bit more raw, but it's easy enough to go through Timeline and collect a few dozen departure times and arrival times into a spreadsheet. The UI makes it easy to go through a day at a time and grab your departure time from Home/Work and your arrival at Work/Home. From here, it's straightforward to see what the correlation is between departure time and commute length. I did this analysis 7 years ago for my morning commute with much more manual data collection (http://matthew.gray.org/2008/05/moved-to-reading-commute-analysis.html). The effort now to do it for my evening commute (which I had never gotten around to previously) was much easier.
In particular, my evening commute is a linear additional 15-seconds-in-traffic for every minute earlier I leave during the ranges that I am likely to leave. In the mornings, per what I learned 7 years ago, I avoid leaving before about 9:15am, but the new data suggests that after that there's a mild reduction in traffic as the morning goes on: about 6 seconds less in traffic for every minute later I leave. The variance of commute time compared to the forecast is also easy to compute so now I have a nice formula to give me 90% chance of making it home by a particular time: Minimal commute time + 7 minutes* + (Arrival minutes before 7:30pm)/4. Seven minutes is 1.25 SD of te commute off forecast, which should give me my desired 90%.
Also, the map of the world, US, and neighborhood showing all the places you've been, by itself is pretty cool.
https://www.google.com/maps/timeline
The new Google Timeline dashboard and feature in Google Maps is very cool. In addition to answering useful things like "Where did we stay for that long weekend in New Hampshire last fall?" or "When was the last time I went to that restaurant?" or the general appeal of the automatic private journal, it makes it amazingly easy to measure commute elasticity.
By "commute elasticity", I mean how much better or worse a commute gets for every minute earlier or later you leave. The downloadable data from Location History is a bit more raw, but it's easy enough to go through Timeline and collect a few dozen departure times and arrival times into a spreadsheet. The UI makes it easy to go through a day at a time and grab your departure time from Home/Work and your arrival at Work/Home. From here, it's straightforward to see what the correlation is between departure time and commute length. I did this analysis 7 years ago for my morning commute with much more manual data collection (http://matthew.gray.org/2008/05/moved-to-reading-commute-analysis.html). The effort now to do it for my evening commute (which I had never gotten around to previously) was much easier.
In particular, my evening commute is a linear additional 15-seconds-in-traffic for every minute earlier I leave during the ranges that I am likely to leave. In the mornings, per what I learned 7 years ago, I avoid leaving before about 9:15am, but the new data suggests that after that there's a mild reduction in traffic as the morning goes on: about 6 seconds less in traffic for every minute later I leave. The variance of commute time compared to the forecast is also easy to compute so now I have a nice formula to give me 90% chance of making it home by a particular time: Minimal commute time + 7 minutes* + (Arrival minutes before 7:30pm)/4. Seven minutes is 1.25 SD of te commute off forecast, which should give me my desired 90%.
Also, the map of the world, US, and neighborhood showing all the places you've been, by itself is pretty cool.
https://www.google.com/maps/timeline
And BTW I still wait for the day, when Google Now, after I biked to office in the morning, does not tell me about my car ride home :-)
Sometime in the past year or so, it stopped giving driving directions for extremely short distances and switched to walking directions, but it's far from flawless.
Well, not exactly flaws, but more things it could infer and do, if someone wanted it to.