On the other hand, if you actually bother to do the math, using the normal approximation you get that a 95% confidence interval around p=0.1 with n=3000 has an upper bound of 0.111, which is well-separated from p'=0.12.

(in fairness, the confidence intervals do overlap each other)