[leegraham]

Maybe I misunderstood the article, but my impression is that the ‘ 4% (1 in 25 users)‘ is just to make the percentages more understandable and the author didn’t only send the email to 100/25/12 users.

[gk1]

That's correct. I'm editing it to say "1 of every 25 users" to make this less confusing. The actual sample size was quite large.