> They even have graphs with historical wait times so you can time it for a low point.

I no longer trust those measurements.

Needing to visit the DMV for whatever paperwork (maybe in-person license renewal), I checked the DMV website. One of the local offices had a very short wait time compared to all of the others (maybe 5 minutes versus 1hr+). I went to the office with the short wait time displayed on the website only to find out they had a ridiculously long line outside the office to get the ticket, which starts the timer. They were gaming the per-customer timer by metering the rate at which customers outside the building could enter the building to get the ticket.

I thought it was both terrible and creative at the same time. Decent example of Goodhart's Law[1].

[1] https://en.wikipedia.org/wiki/Goodhart%27s_law

I don't trust the estimates online either, but I do trust the relative data within each DMV. So I think the graphs that tell you number of people per hour are relatively correct and can be used to predict how busy a DMV will be.