Inside the science of contextual matching

As an AdSense Product Manager for ads quality, I'm part of a group that works to make the ads on your site relevant for users. One of the ways we do this is through contextual targeting, which matches ads to the content of your webpages. With contextual targeting, our system parses and analyzes the content of your page, conducts an auction among the relevant ads, then serves the winning ads -- all in the blink of an eye! Pretty simple, right?

Actually, matching relevant ads to the content of your pages is quite a scientific challenge involving semantic analysis and machine learning. It's a challenge that keeps me and my team busy. Our machines are very good at the matching process, but there are still a few cases where their definition of relevance differs from our human definition of relevance. In these few cases, the system might end up serving ads that don't seem immediately relevant to users. We understand that increased ad relevance contributes to a positive experience for users, publishers, and advertisers, so we're continuously working on ways to improve the relevance and quality of ads that appear on your sites.

I'm excited to let you know that this week, we'll be rolling out a series of enhancements to AdSense's contextual targeting capabilities, which will more accurately match relevant ads to webpages. You won't need to update your AdSense account or ad code, as these changes will be applied automatically. In many ways these types of improvements are par for the course at Google - but I wanted to let you know about the kinds of improvements we've been focusing on. Please keep in mind that these changes won't affect how other types of ads are matched to your sites; for instance, you'll continue to see placement-targeted ads when advertisers bid to appear specifically on your pages.

This is by no means the end of our efforts to perfect the system, but we believe that with these changes users will start to see even more relevant ads, advertisers will generate more attractive returns by finding the right users, and publishers should make more money over time.

As always, we're interested in your feedback on these improvements!

Post a Comment