This is the second of three posts.
Most websites that make money do so by placing ads on their pages...and there’s a good chance that those ads are provided by Google. But it wasn’t always that way. Back when Larry and Sergey were at Stanford they wrote a paper called The Anatomy of a Large Scaled Hypertextual Web Search Engine http://infolab.stanford.edu/~backrub/google.html, where they expressed their skepticism and concerns about monetizing a search engine with advertising.
Having come from Netscape, I was quite familiar with ad models on the web. Banner ads at 2% average click-thru rates and text links at 0.1%. These could be sold by category, but likely were mostly run of site ads (they show up wherever there is a free space regardless of what page it is). This was the era of the “punch the monkey” banner ad. Basically in 1999 advertising on the web was so untargeted that click-thrus began to tail off. The novelty of clicking on banners was largely over. The solution that creative folks came up with was to make them little games or colorful seizure-inducing flashing ads that you would click on just to get rid of them. Yes, the click occurred, but no transaction was going to happen at the offending advertiser’s website. Ads that cost money but don’t produce a sale are bound to drop in rates (price)... and that was what was generally happening.
I was the website bizdev guy so I was on the inaugural ad design team. I was one of the few people sitting around the conference room table when Larry [Page] walked in and declared that “ads can be as relevant as search results” and then basically walked out. I scratched my head a bit and remember thinking “what does that mean?”. What it meant was that we were in no hurry to make money from our website. It meant that we weren’t going to have graphical banners on our site. It didn’t matter that all of the advertisers of the world were used to hiring an agency, having a small number of creative banners produced and then distributed to the pages of the web. It didn’t matter that not only were our ads going to be different, but it was going to be an entirely different sale and require changing buyer behavior and a whole new process. We were not going to consider how the entire industry worked, we were just supposed to build this thing as if the rest of the world and its established practices didn’t exist...as if we were operating in a vacuum.
Larry was right on. The higher qualified/targeted an ad is, the better it will perform and the higher the rates will be. Better=more revenue. What he realized was that you can’t get any more targeting potential than someone explicitly telling you what they are looking for. That is what you do when you go to a search engine. If you have an advertisement that answers that question, it actually adds value to the user. The ad becomes relevant to the search and is a reasonable place to click. Of course there was a problem. Graphical banner ads are not customizable and no agency is going to produce a multitude of banners to match against individual keyword searches. Our solution was to use text ads instead of graphical banners (which had the added bonus of being very fast at a time when internet connections were still relatively slow). We created a very simple online interface which would allow advertisers to create as many custom text ads as they wanted and experiment with which keyword phrases to advertise to. We wanted them to be able to learn what keywords to buy which resulted in the most clicks, not only because it would mean more revenue for us, but because more clicks meant the ads were relevant to the searcher. Changing user behavior, in this case the entire advertising world is not easy. What what we knew and ultimately figured out was that we needed to economically incentivize advertisers to want to do a better job advertising to our users. All advertisers wanted the top ad positions on the page, but we made it that they couldn’t simply buy there way to the top. What we did was said that their position amongst all the ads was relative not only to what they bid for their ad, but cross referenced against the click-thru rate for each specific ad against each specific search term used by the searcher. You could pay less for a better slot if you had a better and better targeted ad. We also encouraged the advertiser to link to the relevant page within their own site that related to the search, not just to their home page (i.e. if you searched Google for "wedding china", an ad for William Sonoma might appear. A click on the ad would lead the searcher to the wedding china page, not the williamsonoma.com homepage, which is what used to occur). This type of thinking made the whole experience better. And the rest as they say, is history.
My take-away is that:
Google completely disregarded the way the entire advertising industry worked and came up with its own solution as if it existed in a vacuum. And ultimately, it worked. Sometimes to find the right answer you have to look at a problem and work out a solution which completely disregards current standard practices.