Super insightfull post, thanks for sharing your approach to SEO experiments!

When comparing the average difference before and after the test, how exactly do you calculate that your test gives you significant results (let’s say at 0.95)?