Advertising makes the web go ‘round. As annoying as it can be, we wouldn’t have access to today’s internet without it. Nor, for that matter, would I be writing this. Engineers and advertising savants have been working at how to make web advertising profitable for the past 20 years, and at this point, the advertising market seems to have a model that works—judging by the amount of advertising I see on the web today, at least.
The infrastructure that advertising networks use to deliver those banner ads, pop-over, pop-under, or sidebar ads is really something. Companies have been working on getting these networks in place over the past decade, and they’ve sunk barrels of cash into them. They are widely distributed, support near-real time auctioning of content, and perform pattern matching at a lightning fast pace. So, how does all this work?
Well, we can think of this as operating in three phases. In the first phase, the network pulls information on the persona behind the browser. In the second, that persona is matched against pre-existing personas, and the match determines the ads sent to the user. Finally, if the generated persona doesn’t match a preconfigured one, the network accepts bids from advertisers and delivers the ads from the advertiser with the winning bid.
The first step, pulling the browser persona, uses tracking cookies of various kinds and a large database of pre-collected data. Advertising networks use a variety of algorithms to do this. They may not know exactly who you are (they might though, depending on your social media preferences), but they can collect an enormous amount of data on your browsing and internet buying habits. Keep in mind, these networks are used by millions of sites. That gives them the ability to plant tracking cookies and evaluate those cookies at any of the sites that contract with them. They can also track your shopping over time, as we all usually use the same browser to shop with. Even if we don’t, our ISP DHCP settings likely give us the same IP address for long periods of time, and when they change, the tracking cookies are likely still stored in our browsers and can be correlated with the new IP address. This way, advertising networks can capture shopping and browsing habits over long periods of time, as well as collect your specific browsing habits for a particular session. All of this information is compiled into your persona, and that’s what they match advertising against.
The second step is pretty straightforward—the advertising network will take your persona and match it against specific profiles. These profiles are associated with pre-packaged groups of advertising content. This content is then presented to the user via the advertising network. If this match occurs, the advertising delivery process is wrapped up.
Things get really interesting when your persona doesn’t match a pre-packaged persona. When this happens, the advertising network begins the auction step. Here, a generated persona is presented to a group of pre-registered advertisers. If those advertisers are interested in advertising to that persona, they enter a bid. After a very brief period of time (in the order of milliseconds), the advertising network selects the auction winner, and that winner’s advertising is presented to the persona. Not only is this very fast, it happens hundreds of thousands of times a second, worldwide. In cases where users are not tracked or do not have a persona to match in the previous step, that user’s viewing usually goes immediately to auction.
Advertisers really want to increase clicks on submitted advertising, and the best way to do that is to present you with advertising directed to your interests. They also would much rather process advertising delivery via step two, as it’s faster, cheaper, and less complex than step three. Nevertheless, they’ll always need to be able to send advertising to users they can’t track or match, so they’ll always need something like step three... and so far, auctions have proven to be the best way to match advertisers to users.