Hacker News new | past | comments | ask | show | jobs | submit login

How would you possibly A/B Test SEO improvements?

By rolling it out to different segments of your data. Like a single city or category.

How do you control for natural variations in the popularity of different segments, then?

Say that you're Yelp, and you make a change to the Boston section of your website. Unbeknownst to you, a competitor has sprung up for local Bostonian reviews, and is rapidly gaining market share. You see your Boston pages drop in rankings, but it may not be because of the change you made. It may be because a bunch of Google's other signals have picked up that this other competitor is suddenly much more popular, and are adjusting ranking accordingly.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact