The Shape of My Commute : Wrapup

I've stopped accumulating data about my commute (for a while, anyway). Here's the summary of what I did and what I learned.

I first wrote a program using Selenium to extract estimated trip times from Google Maps. I doubt this will work with their new end-user interface, but I got the data that I was interested in. I accumulated the estimated trip times from 4 a.m. to 8 p.m. over the course of a few days. This graph shows the average: 
I added a line for my "tolerance". My hope was to stay under 45 minutes each way. One minor interesting point: the estimated trip time is not symmetrical. You can see the "bump" at noon, when I switched the commute direction.

I then manually collected estimated times and actual times:
I also put my "tolerance line" at 45 minutes. Here the reality sinks in. Google Maps underestimates the commute time (during commute hours) by 5% - 25%. I've read a number of posts that complain that Maps overestimates trip time, but those seem to apply mostly to longer, non-rush-hour trips.