A little late but 20 samples is statistically insignificant. I'm not saying this to discredit your findings. I'm saying this as someone who has studied lots of statistics. It's just not enough to suggest that the rate has changed.
(Sloppily) running through the calculations using some fairly benign approximations says you need well over 100 samples to even have a 95% chance of disproving that it is wrong. It's a Bernulli Distribution and I used a Normal Approximation.
Replies
Sush he knows his "math"