[network] Analysing the Google business model

How is Google making a profit? What are the costs and revenues if you look at each individual search result? I had a detailed look at the 2004 annual report to figure out these answers. Google’s total costs in 2004 were $ 2549 million, of which $ 1228 are traffic acquisition costs, e.g. paid out through AdSense (annual report pages 19 and 27) To calculate the content delivery cost per search result, it is fair to exclude the latter, leaving $ 1321 million.

It is surprisingly hard to find out the other relevant key performance indicators, such as number of search results served. The most recent numbers are end of 2002, when Google allegedly served 150 million queries per day with 15.000 servers. In the mean time the server farm has increased 10 fold, as has total spending. The revenues have gone up with a factor of 4. On this basis my guestimate would be that Google in 2004 served 750 million queries a day, which is still less that one per day for each active Internet user.

This implies a content delivery cost of 0.5 cents per search result, including serving up the ads and spidering the web. The pure technology cost seems to be about half of this. This is considerably higher than the 10 microdollars for serving a web page, which is what Jim Gray of Microsoft arrives at.

Google states that 99% of the 2004 revenue ($3189 million) is from ads. This implies a revenue of 1.1 cent per search result. To double check this, let us assume that a typical search result has 4 ads, and that the average click through on ads is 1% (based on my own data). An average click through will cost around $ 0.27, which seems to be in the ballpark.

I am puzzled by the difference between the growth in revenue and the growth in technology cost. It also looks like Google is way oversized in infrastructure, in relation to the current services provided. It gives me the impression that they are betting several hundred million dollars on the emergence of profitable new services. For an outline of what those could be, have a look at the labs.

It would be interesting to make a more detailed analysis of the cost structure, both on the technology side, where hardware and network seem to represent only a fraction of the cost, and on the content side, where Google is more and more relying on partners (i.e. AdSense) to boost revenue at considerably lower margins. It would allow us to figure out what the real profit growth potential for Google is, and therefore what a realistic stock price is. I don’t know yet, so I don’t have any stock, yet.

2 Comments on “[network] Analysing the Google business model

5 February 2007 at 07:42

>Mark Mahaney, has some assumptions on the numbers in this analysis that differ by a factor of two or so.

In the International Herald Tribune of Monday feb 5, 2007, Mark Mahaney is quoted as estimating 5 cents revenue per google search, considerably higher that my 1.1 cents.

bernard lunn
27 March 2008 at 16:07

>I am doing some research on this. Do you have any updated research or know of anybody who has done this recently?

Leave a Reply