matt comments on What if the front page… - Less Wrong

38 Post author: matt 14 March 2012 06:47AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (35)

You are viewing a single comment's thread.

Comment author: matt 14 March 2012 06:42:49AM 1 point [-]

No no no. No. You have it all wrong. I've upvoted this comment and replied to it with a detailed description of what you should do instead.
(Or I've upvoted this comment and a child of this comment I think is a better idea.)

Comment author: John_Maxwell_IV 14 March 2012 11:06:09PM 5 points [-]

Before experimenting with the Less Wrong homepage, we should start using A/B testing techniques to see how to best interest folks in Less Wrong. Here is a nifty framework:

http://genetify.com/demo/

Ideally, any user would be allowed to suggest a change to the Less Wrong introductory copy and see it tested out. Over time, we could learn what copy best interests new users, and experiment with new home page designs based on what we learn from that.

Comment author: matt 16 March 2012 12:15:52AM *  2 points [-]

A/B testing is something I'd love to do… but I don't have the time to do more than support someone else managing the project.

Stuff I can make happen:

  • support for multiple wiki pages mounted at lesswrong.com urls
  • support for the front page to run js that will select which variant page to display

Stuff someone else (with high site karma or vouched for by someone with high site karma) would have to commit to do:

  • generate page variants
  • decide on the goal condition we're optimising for
  • manage the ongoing test (requirements determined by the A/B framework we use)
  • anything else I didn't think of - you'd be the project manager.

(Without having reviewed Genetify I'd prefer to use Google's free A/B testing system, but would be happy to convinced to follow the project manager's decision.)

Comment author: Modig 15 March 2012 10:47:37AM 1 point [-]

Upvoted for A/B split-testing! Don't guess when you can test.