Compete cannot measure search market share
This will be the last month, until Compete changes its reporting structure, that I attempt to measure search market share using Compete’s data. Their inability to drill down to the sub-domain level renders a lot of their data virtually useless for competitive analysis. In September 2008 Search Market Share per Quantcast, I found it necessary to include 6 search destinations I had not previously measured. The new services include MySpace’s search tool, Publisher’s Clearing House’s search tool, and CNN’s search tool.
Quantcast estimates that only a small fraction (in the area of 10%) of MSN.com’s traffic is Web search traffic. Compete makes no distinction between MSN’s search and non-search traffic. Less than half of Yahoo!’s overall traffic is related to search, according to Quantcast’s estimates. Using Compete’s public data to rank search engines is about as misleading as using queries performed to rank search engines. You end up with meaningless gibberish because you’re not talking about real search traffic with either metric.
Simply measuring the number of visitors to a search destination is only slightly better because we don’t know why those people are visiting the sites. The best possible metric for measuring search market share would be the referrals that search services provide to other sites (but then you have to deal with the sticky issue of bounces — when does a short visit become a bounce?).
Compete estimates that MySpace received around 56.5 million visitors in September. Quantcast estimates 68.3 million visitors for the same period. However, Quantcast only estimates 23.9 million visitors to MySpace’s search tool — about 1/3 of MySpace’s estimated visitor count. If most people who visit MySpace, MSN, and Yahoo! are NOT performing Web search at those sites, does it make sense to include them in the monthly search market share reports?
I find it hard to justify including a growing number of such services just because they have millions of search-related visitors. The rankings based on Compete’s numbers are nonsense. But to exclude those services would not be realistic, either. As I noted in the September 2008 Quantcast report, I have increased the required minimum estimated visitors to 1,000,000 in order for a service to be included in the Algorithmic Search report.
MySpace, with an estimated 23.9 million search-related visitors, draws more search-related traffic than MSN, AOL, and Dogpile combined (and you could throw in several other small services and still not match MySpace for search traffic).
I could, of course, adjust the Compete data artificially by applying the Quantcast ratios to the Compete data. For example, Quantcast’s estimated search traffic for MySpace (at 23.9 million) is about 35% of MySpace’s overall traffic (visitors move around MySpace, using more than one sub-domain in many cases). So I could multiple Compete’s estimate of 56 million visitors and create my own estimate of 19.6 million search related visitors. This kind of normalization is used in many applications.
However, Quantcast knows nothing about Compete’s database and analytics. I might as well make up a random percentage (say, 42%) and apply that with equal validity to Compete’s data. There is no meaning in using Quantcast as a benchmark if I cannot show there is a correlation between Quantcast’s breakdown and how Compete would break down the data.
So there is really no way to use Compete’s data to estimate search traffic based on visitors without making serious compromises that invalidate the tabulated results. For what it’s worth, Compete estimates small increases in traffic to all the major search services for September — but it’s impossible to see just how much more search traffic Microsoft has grabbed.
I suspect that both Quantcast and Compete have the resources to provide some estimates for search referrals. I would like to see them publish rankings for the top 20 search services based on estimated referrals — or to make that search referral data available (aggregated by search source, rather than detailed by destination).
I did spend several hours over the weekend working on a report for September 2008 Search Market Share per Compete’s data, but at the end of the process I just felt like I had wasted my time. I won’t be publishing that data, but I hope that Compete improves its free services to include more detailed information by sub-domain, as well as search referral data by search source.
www.seo-theory.com
published @ October 14, 2008