Our data goes back to 2012, but as a practical matter doesn't really start until 2013 since not everything was hooked up to fedmsg.

Meanwhile, I divide users into three groups for visualization:

• new users, active less than one year
• intermediate, which was 1-2 years but I just changed it to 1-3
• old school, which is more than that.

Currently, the graphs start at January 2015, because with a threshold of 2 years for "old school" that meant that a large group of people who had been around longer than the start of the data show up categorized as old school on the graph. But now I changed that to 3 years, the beginning of the graph looks funny.

I can think of several possible solutions:

1. Just start at 2016
2. Start at 2014 and add a note as to why the data looks like that
3. Special-case names which show up in the first year to mark a certain cohort of people as old-school with a lower threshold
4. Instead of extracting first use from this data, use user creation time from the account system.

I'm not sure if 4 is available to the general public, so we may want to avoid that, even though it seems like the neatest solution.