4 Comments
User's avatar
Lindsay M. Pettingill's avatar

Good piece!

My sense is that the market generally prefers CEOs who are predictable, and this much of an about face is neither a good look nor particularly encouraging wrt leadership.

I'm intrigued by the counterfactual : if he had claimed it was an experiment, would he have gotten as much press? I would be willing to bet yes!

Expand full comment
Ronny Kohavi's avatar

"If you are shipping features without running a *controlled* experiment, you are really running an *inefficient* experiment, where analysis is done by looking at a time-series graph"

-- https://www.linkedin.com/posts/ronnyk_abtest-featureflags-experimentguide-activity-6897828685114675201-fJyX

Expand full comment
Ronny Kohavi's avatar

I love the statistics about failures in books, drugs, and social programs, but you're missing the one closest to home: for online controlled experiments, or A/B tests, the median success rate is 10%: https://bit.ly/ABTestingIntuitionBusters Table 2.

Expand full comment
Ryan Lucht's avatar

Indeed. I presumed I was writing for an audience familiar with the success rates of A/B testing programs but fair point - I should have a more optimistic presumption of who this might reach and should include this for clarity! I'll update the post.

Expand full comment