I ran my updated cookie-ability + JS-ability recording code for a few days and got some results.

With a combination of PHP (if they visited more than one page) and JS (if they have JS), I recorded whether the visitor had actually picked up the cookie or not. I then collapsed the records of visitors who did not pick up the cookie (or I couldn’t verify their cookie-ness because they didn’t have JS and only visited one page) by IP + user agent. This gave me a total of 2882 unique visitors to my site in the past couple days.

Here are some quick questions I wanted to answer and what they mean for ShrimpTest:

Were there many users who had JS but had cookies disabled?: No: only 25/2882 < 1%

Were there many users who had cookies but no JS?: Yes: 154/2882 = 5.3%.

The take-away:

For calculating ShrimpTest statistics, good approximation of “unique visitors” would be to record using PHP that certain visitors have actually retained the cookie, then collapsing the other visitors by IP + user agent. Using JS to confirm cookie-ability is not necessary, and using only JS to assign cookies would miss a decent number of users.