> 1. Apparently everyone in Santa Cruz goes to bed early and gets up early.

Probably some selection bias at work here. It's possible the type of person who's buying and wearing personal fitness monitors is also the type of person who gets up early.

> Maybe I'm the only one that sets my alarm on the hour

Probably my OCD coming out but I only set my alarm with minutes beginning 0-4 and ending with 8. I'm sure everyone has their own habits. On the hour seems so numerically boring!

If it was selection bias like that, the same would apply to people in other cities, so it wouldn't explain the difference in the graph. My guess is people in Santa Cruz on average have a longer commute.

It's going to be a selection bias of:

1. People who are fit (or want to be)

2. People who are technical (or are interested in quantified life in general)

Technical people who commute from SC to Silicon Valley have far less bus options and routes than SF (SC has three each way for Google which ends at 9:25). That forces a lot of the number 2 people awake earlier than they otherwise would be.

I think SC is probably the same, if not average getting up later than other Bay Area places, due to the large student population and short walk/bike commute for those who work in the town.

> Maybe I'm the only one that sets my alarm on the hour though.

I assume most of the UP-wearers will use the "smart alarm" function on their wristband that will wake them at the "optimal moment" in their sleep cycle.

> 1. Apparently everyone in Santa Cruz goes to bed early and gets up early.

What I'm guessing is that people in areas of higher population density stay up later and get up later on average. The 25-50 and 50-75 mile bands seem to contain the big urban centers, while the 0-25 and 75-100 bands are more the suburbs, small cities and rural areas.

I would expect that most people set their alarms in 5-minute increments, and dial in what minute they choose to meet the needs of their commute and morning routine. This forces variations away from a "sharp" time spike. Also, alarm clocks are almost never right on time; I would expect that most alarm clocks are more than a minute off. This factor would smooth out the 5 minute steps I'd expect to see in the data. The alarm clock inaccuracy may decrease over time as people move to using their smartphone to replace yet another device.

Since they don't say whether each line represents the same number of people, I would guess that an outlier line is just a smaller sample.