The Technium

The Value of Search, 2


The other day I attempted to determine an approximate cost/value of a single search. I  calculated how much Google spends to support search and then divided that by the number of searches to get a cost per search. And I took Google’s total market value determined by its stock price and divide that by the number of searches to get a value per search.

My cost calculation drew no objections but a number of smart people wrote in criticizing my use of market cap valuation as a surrogate for actual value of a product or service. That’s a legitimate concern because what a market cap is primarily measuring is potential future earnings. I dabbled with market cap valuation because I lacked an alternative method. A few correspondents wrote in with alternative methods. 

The three suggestions for increasing the validity of my calculations were: 

1) Use a conservative multiple of earnings as the total value.

2) Use only the earnings from search divided by number of searches.

3) Determine how much time an average search saves, multiplied by an hourly wage.

Brad Templeton suggested that if I had to use a valuation (and he did not think I should) it would be better to take an established “multiple” of the earnings, rather than the stock market cap, which obviously fluctuates day to day. An ordinary business might be valued at 5 times their earnings (P/E 5). At today’s stock prices Google’s market cap is about 50 times earnings (P/E 50). That 10X inflation is due to very high (if not irrational) expectations of future growth in Googl’s revenue. It we dial back Google’s valuation to a more reasonable and ordinary P/E 5, we reduce its total value by 10X. Thus we can reduced the value per search by 10X, which yields a value of 2.7 cents per search.

Brad suggested an interesting way to confirm this approximate value of 2.7 cents per search. The return to Google per thousand searches is $27. Let’s call that a $27 RPM (return per thousand). If Google routinely pays out $10 or $12 per thousand impressions (CPM) to blogs for their Adsense program, we could expect they earn $13-15 CPM. So search RPM falls in the same rough neighborhood as CPM for page impressions – which would be reasonable because advertisers have the choice of advertising on AdWords (search) or AdSense (page impressions), and the fees might be competitive. This would suggests 2.7 cents/search is reasonable. 

The second method is to rely on earnings. Google’s revenue last quarter was $3.9B, or $15.6B for 12 months. If 66% of that is search revenue, that means Google earns $10B in search. If it serves up 444 billion searches, each one generates 2.3 cents. 

The third method is different, and in my mind the most promising. It was sent to me by a Google Insider.  This method removes the most troubling factor in trying to determine the value of search – the fact that its economic value is amplified by its association with advertising. My correspondent says, “I think market cap can be used to estimate the value of a google AD, but not of a SEARCH.  Because if Google had a different monetization method, its market cap would be different, but that wouldn’t change the value of search.”

He/she gave me permission to post it without attribution, noting that it was a back of the envelope musing and “not official company policy.” 

Interesting article.  I was trying to compute the same metric myself. I have more accurate counts of the number of searches per day (which unfortunately I can’t share), but your estimates are good enough for back-of-the-envelope. 

Here’s another way to figure: use Comscore’s 1.2 billion searches/day. Let’s say 

1/4 of all searches are really easy ones (like “american airlines”) that save the user maybe 30 seconds; 

1/4 are a little hard and save maybe 5 minutes; 

1/4 are just wasting time, and 

1/4 are hard ones that lead to substantial savings –  like diagnosing your serious disease, or choosing the right college, or  the right vacation destination.   

Suppose it takes 10 searches on average to get one of these “hard” answers, but when you get it, you’ve saved maybe 3 hours. That averages out to 6 minutes saved/search.  Figure average income of $25000/year, or $12.50/hr. So we get a value of $1.25/search by this metric. 

If we tally that up for 444 billion searches it equals $547B/year, or 4% of US GDP. That certainly sounds too high.  Maybe people are doing more trivial things, and their big searches are only saving an hour. That would save us 3 minutes/search. And maybe we should value people’s time not by what they get paid per hour for working, but by dividing their income by their total waking hours. Doing that we get $0.20/search.   

So your numbers (27 cents) come out in the same ballpark as mine. 

Let’s add a little slop and say the value of a search is somewhere between $.15 and $1.50, which, if you believe Comscore numbers, works out to somewhere between 1/2% and 5% of US GDP.  Note that at $.27 your figure comes says search is worth about 1% of US GDP. 

I’d like to add: This is only the value of search. And not just Google search but all search by all search engines. However search is instrumental to, but not encompassing of, the entire value of the web. If we were to add the value of the web itself (entertainment, publishing, PR, communication), I would expect it to be quite a bit larger portion of the GDP. 

My conclusion from all these guesstimates is this: the cost of a search is about .3 of a cent. The value of a search is between 3 cents and 30 cents per search. It’s a good business to be in.




Comments
  • http://dakec.blogspot.com dake

    I think adsense payout figures of $10-$12 RPM is quite high, may be I couldnt figure out how one could earn from Google Adsense.

    But article is really very interesting.
    Good business to be in search.

  • http://bernardlunn.wordpress.com bernard lunn

    Kevin, I found this via a Google search (of course) for “cost per search” but I was looking for the cost to the engine and not to the end user. Trying to build a basic model of the search biz. Revenue Per Search is well established. So I was looking for the cost side of the picture ie all those server farms amortized across all the searches. Have you looked into that? Bernard

    • Kevin Kelly

      No Bernard, I have not looked at the search biz model. Many have though.