"Data Scale" and the Yahoo - Microsoft Search Partnership
Internet Search Insiders have understand the power of Data for some time now. The more data you have, the better the insight. Google's big chief scientist Peter Norvig has talked about the use of large amounts of data to get consumer insights and to improve algorithms. Now the need to get more data to improve search and advertising algorithms has resurfaced in the Microsoft/Yahoo deal again.
The relevant part from a NY Times article: (http://bits.blogs.nytimes.com/2009/07/30/behind-the-microsoft-yahoo-deal-the-internet-economics-of-scale/)
------
There is another dimension of scale behind the deal, according to Murthy Nukala, chief executive of Adchemy, a Silicon Valley startup that uses statistical models, advanced data mining and machine learning to help target online advertising. The pursuit of "data scale" in search, Mr. Nukala said, fueled the Microsoft partnership. The data-scale benefit, he added, comes from more than just generating more search traffic, though sheer volume is crucial.
A key challenge in search, he explained, is estimating the likelihood that a given user will click on a particular ad from a certain advertiser. In the search world, this problem is called "pCTR estimation," for probability of click-through rate. The ranking algorithms for search advertising, Mr. Nukala explained, incorporate not only the price per click an advertiser is willing the pay, but also the estimated click-through rate (calculated by applying clever algorithms and machine learning to vast quantities of query data).
"It is well understood," Mr Nukala said, "that as pCTR estimates improve, the quality of ranking is better, which leads to higher revenue per search."It is also important to understand, he added, that the click-through estimates do not improve merely proportionately as search traffic increases but by something more like an exponential multiplier. Presumably, that is the sort of thing Mr. Ballmer had in mind when he said "scale drives knowledge." Mr. Nukala concluded, "I believe that 'data scale' drove the strategic imperative and the structure of the deal."
------
Make a comment