http://chrisjpowers.com
@chrisjpowers
Multivariate Testing
Control White Text Black Text
Control Control Variant 3 Variant 6
Blue BG Variant 1 Variant 4 Variant 7
Green BG Variant 2 Variant 5 Variant 8
Slide 13
Slide 13 text
http://chrisjpowers.com
@chrisjpowers
Great Expectations
Slide 14
Slide 14 text
http://chrisjpowers.com
@chrisjpowers
Big Lifts
14%
62%
http://chrisjpowers.com
@chrisjpowers
Statistical Significance
“If my p-value is less than 0.05, then I
have a 95% chance that my reading is
correct.”
Slide 35
Slide 35 text
http://chrisjpowers.com
@chrisjpowers
Slide 36
Slide 36 text
http://chrisjpowers.com
@chrisjpowers
Statistical Significance
“If my p-value is less than 0.05, then I
have a 95% chance that my reading is
correct.”
Incorrect
Slide 37
Slide 37 text
http://chrisjpowers.com
@chrisjpowers
Statistical Significance
“If my treatment has no actual effect
(null hypothesis) then there’s a 5%
chance I will measure a p-value less
than 0.05.” correct
http://chrisjpowers.com
@chrisjpowers
You May Need These Lifts
Slide 51
Slide 51 text
http://chrisjpowers.com
@chrisjpowers
Living with Small Samples
Slide 52
Slide 52 text
http://chrisjpowers.com
@chrisjpowers
Slide 53
Slide 53 text
http://chrisjpowers.com
@chrisjpowers
Slide 54
Slide 54 text
http://chrisjpowers.com
@chrisjpowers
Slide 55
Slide 55 text
http://chrisjpowers.com
@chrisjpowers
Slide 56
Slide 56 text
http://chrisjpowers.com
@chrisjpowers
Slide 57
Slide 57 text
http://chrisjpowers.com
@chrisjpowers
Run experiments at 50/50. Only amateurs
run experiments at 1%. - Ron Kohavi
Slide 58
Slide 58 text
http://chrisjpowers.com
@chrisjpowers
The Peeking Problem
Slide 59
Slide 59 text
http://chrisjpowers.com
@chrisjpowers
Slide 60
Slide 60 text
http://chrisjpowers.com
@chrisjpowers
Peeking Problem
Peeking regularly and stopping
experiments when significance is
achieved raises the 5% false
positive rate to… 30%