Home > Questions and Answers > General Questions

HELP needed with Statistical Significance - Split Test Calculations (3)


01-13-2014 12:59 PM #1 iAmAttila (Veteran Member)
HELP needed with Statistical Significance - Split Test Calculations

Hello,

Wondering if someone could help me understand why banner2 wins - been trying to wrap my head around this, but having troubles understanding.

I have banner1 - 239,070 impressions, 689 clicks, 3 leads
I have banner2 - 140,710 impressions, 435 clicks, 1 lead

The split test calculator says that banner2 wins by 73.45% Confidence ..

-it obviously does not factor in total leads generated- but what if it did?

or better yet, how could I factor in total leads generated into the formula when calculating the stat?


01-13-2014 07:22 PM #2 andyvon (AMC Alumnus)

In your example, banner2 doesn't really "win", it's just a bit more likely to be the better performer of the 2.

As mentioned in this thread you should aim for at least 90 % confidence.


01-14-2014 06:04 PM #3 caurmen (Administrator)

Also, you'll find that the calculator produces weird results at low numbers of conversions. I'd recommend getting some more conversions on both banners before running them through the Bayesian split test calculator.


Home > Questions and Answers > General Questions