Sunday, August 16, 2009
I disagree somewhat with Hal Varian's (Google's Chief Economist) criticism of the theory that combining Yahoo and Microsoft in search will lead to improvements based on scale, data, etc.
Statistically, for sure, scale is already high enough that more won't apparently lead to significant improvements in either search performance or ad program performance due to the impact of data on improving relevancy. That's on paper, in the lab.
Off paper, in the real world lab:
I believe that Varian's assumption is mainly wrong because he's giving his competitors credit for having more consistent share across all major segments than they actually do. Aggregate numbers look impressive, but the information is less consistent as you drill down. Doubling or tripling the available information in any given segment, especially small ones, is bound to be helpful.
- Yahoo's reported 20% share is fiction. Globally, it's much lower. In the US, the real number is actually lower.
- Data is highly granular in a number of ways. So to start, Yahoo and Microsoft have different search shares in every language and every country in the world, and different search shares in sub-regions of the world. In many, one or the other currently hold share of 1% or less. By bringing both up well over 1% and closer to say, 3%, you get a significant increase in useful data.
- Even the tools that Microsoft provides for advertisers will improve markedly with a doubling or tripling of available data across all major markets, because usable data also comes in the form of highly granular data about keywords. Google doesn't have every last useful tool for researching keyword and consumer behavior: Microsoft has and will develop some really useful ones. Currently, as an advertiser trying to use the tools, you get "insufficient data".
- And though this may stray somewhat from the subject of how to improve a search engine's relevancy... what about something super real-world and practical: running an ad rotation test for a group of keywords and trying to select a winning ad from a field of eight? Isn't that search marketing? Right now, no one is testing very much on any platform other than Google. I suspect they'll be more likely to try tests specific to the Microhoo audience now, rather than just porting all of their consumer feedback driven campaigns over to the Yahoo and Microsoft platforms. The current way is just guessing: really testing in the actual auction you're buying the media in, is more precise.
- By "now," of course I mean when the Microsoft-Yahoo platform consolidation is complete in around a year's time.
To double predictive accuracy, Varian suggests you need "four times as big a sample". Well depending on whether you're looking at it from the standpoint of Microsoft or Yahoo for any given teeny tiny segment, the number of instances where one of them now has "four times as big a sample" is going to be very high. Doubling predictive accuracy on teeny tiny segments - either as a search advertiser or a researcher looking into search trends - is our bread and butter out here. We'll take the "bogus" scale of the Microhoo deal any day.
P.S. I loved Varian's other insights, including the interesting note on the emergence of the "micro-multinational" type of growth company. Though I might have to take a run, at some point, at the recurring Google theme about "communication costs basically going to zero." The costs for collaborative tools have gone close to zero. But...
Labels: google, hal varian, microhoo, microsoft, yahoo
View Posts by Category
Andrew's book, Winning Results With Google AdWords, (McGraw-Hill, 2nd ed.), is still helping tens of thousands of advertisers cut through the noise and set a solid course for campaign ROI.
And for a glowing review of the pioneering 1st ed. of the book, check out this review, by none other than Google's Matt Cutts.
Posts from 2002 to 2010