Imagine you’re at a large cocktail party. You need to quickly identify those few people important enough to spend your limited time talking to, or about. There are a lot of signals you could use. Perhaps it’s a super-fancy party and people are announced as they come in the door, along with any titles they may hold. Maybe the really important people dress nicer. Or in another time or place, the really important people are the ones who can afford to be eccentric and show up wearing jeans and a hoodie.

You could ask, but each opinion might be a little different, and it would take a long time to gather enough opinions about each other person at a large party to gather value.

Then it occurs to you to watch the handshakes and introductions. When someone takes the time to introduce one person to another, they are signaling the value of each person. If many people introduce their friends to the same person, you begin to get a better picture. But then you realize that not all introductions are equal. If someone is getting a lot of introductions, then introduces her friend to another person, you might think that signal counts for a lot more.

The ultimate version of your vision is a map of the whole party that ranks everyone according to the introductions they receive, weighted by the rankings of those who perform the introductions. This map is basically how Google started ranking web pages.

What can you do online that is similar to introducing your companion to someone? If you have a website you can introduce your visitors to another website by linking to it. Google began as a system called “backrub,” which ranked all the websites it could index based on the link structure. This system was called PageRank, partly because Larry Page was one of the inventors.

Over time PageRank has been one of the primary targets of website owners seeking to increase search traffic.

You can read the original BackRub thesis here.