Oct 3, 2005

Advertising

So, you want to get into Internet Advertising? Well, here’s my advice in a nutshell:

Don’t.

It’s not that there is a lack of opportunity out there: the number of fools that believe internet advertising is effective is only exceeded by those with page space to sell. And the algorithms required to run a successful campaign are simple enough that a first year programming student could understand them.

No, the problem is data. You need it, lots of it. All of the good Bayesian algorithms out there require a lot of data, more data than you could possibly imagine. You need to know what kind of phrases people are looking for, how they correlate to each other, how behavior changes with the results displayed.

In short, you need to run a search engine for a few years.

Now you see the problem. The two big internet advertising companies are Google and Yahoo. The two most popular search engines are Google and Yahoo. Unless you can get 10% of the search engine traffic, the data you collect will be seriously skewed and virtually useless.

The problem here is that to do a proper internet advertising campaign you need to address the internet populace as a whole: you cannot provide a niche service and collect general results.

The first is to bite the bullet and buy the data you need. This is the approach Google and Yahoo would like you to take, especially in the light of recent announcements. Google will sell you the information you need wholesale.

The second is to surreptitiously collect the data you need from public APIs. This is the approach a lot of shady companies take (read: EvilCo.net). In fact, the companies tend to be proud of the fact, with projects named “Parasite,” “Predator,” “Leech,” &c.

The third is to intercept the queries before they get to the search engines. If you can get some popular websites to forward that information to you at the same time it’s sent to the search engine, you can build up a statistical database of your own. This is the approach meta-engines use.

The end result is that while it may currently be an advertising company, in the long run Google is going to be a statistics wholesaler.