Best SEO Blog|Launch your site.

Back when Media Metrics first started measuring search market share they counted estimated unique visitors to each search engine. comScore bought Media Metrics in June 2002 and in April 2003 comScore announced its decision to begin tracking queries performed.

At the time, they justified their decision this way:

Previously, marketers gauged search activity using counts of “unique visitors” to search engines sites, or data reported by search providers themselves. However, these approaches are an incomplete measure of true search activity, and may thus lead to less effective business decisions and investments by search providers, product developers, marketers and advertising agencies. For example, many Internet users visit a portal or search engine without actually performing a search (e.g. some people use a search engine site as their home page). In fact, comScore research revealed that on average, only 64 percent of visitors to portals and search engines conduct a search at that portal in a given month.

Additionally, because search engines and portals employ varying standards in counting searches, qSearch methodology is a more accurate and objective than data provided directly by sites. Since the methodology eliminates confusion resulting from automated searches by “bots” and meta-search engines, as well as double-counting resulting from multiple submissions of the same query, qSearch allows for a more standardized comparison than previously available.

Their qSearch technology uses combined data sources — at the time representing about 1.5 million Internet users. At the time the argument seemed plausible, but through the past 5-1/2 years it has become obvious that query-based analysis of search market share is woefully inadequate. In fact, the very inaccuracies that comScore asserted it could filter out have in fact invaded the data that comScore, Compete, Hitwise, and Nielsen use to measure search market share.

Through the past two years I have documented across several Web sites a trend in the search market share data that reveals a significant increase in estimated Google search market share that is not reflected in data shared by Webmasters about their search referrals (nor is it reflected in the data I track for hundreds of Web sites).

Many, if not most, SEOs now report that Google is responsible for 80-90% of their search referrals. However, I have found that very simple factors influence the perception of reported referrals. The chief flaw in these SEO valuations is that most people in our industry now rely almost solely upon Google Analytics for their data. Google Analytics, however, is one of the most inaccurate search referral trackers on the market. I have compared actual server logs to Google Analytics data and found that Google completely overlooks the vast majority of search referrals (even undercounting its own referrals).

A typical Google Analytics search referral report may indicate a handful of referrals from all other search services combined and thousands from Google. I have consistently found through dozens of server log comparisons that Google substantially underrepresents the search referral traffic from Yahoo!, Microsoft, Ask, and other, “smaller” search engines. There is no need to explain the discrepancy — it’s obvious the Google Analytics data is not helping the search engine optimization community optimize for search as much as people believe.

Another factor that directly impacts apparent search referral distribution is third-party site search. For example, I found that my own Microsoft referrals more than doubled when I replaced my Google site search box with a Microsoft site search box — and at the same time my Google referrals only declined slightly. Google’s site search is clearly inadequate because of its failure to index the majority of content provided by web sites. Microsoft’s site search is far from perfect but they do index more text from more pages per site than Google.

Other factors that directly impact search referral analysis include: use of Javascript-based metrics programs, improperly configured server log analysis tools, an inexpert attention to only the top-most referral data, and SEO industry bias toward Google referrals that emphasizes optimizing only for Google.

There is a nearly complete disconnect between the SEO industry and what is actually happening in the world of search with respect to understanding search market share. There are, at this time, no accurate or reliable methodologies for measuring and reporting search market share. For the past several months I have been using data from Quantcast and Compete to estimate real search market share based on estimated unique U.S. visitors.

There are some drawbacks to this approach but comScore’s objections are no longer valid. For example, not only has the search industry all but abandoned the “portal” concept, both Google and Yahoo! have reported that millions of people use their search engines as navigation tools. The increased use of toolbars and default browser search settings make it easier than ever for people to integrate search engines into their daily Web surfing.

Unique visitors, of course, only tells part of the story. There is still some merit in looking at number of queries performed, but the problem with that approach is that people now habitually perform rank checks throughout the day. Automated tools may or may not be filtered by comScore and its competitors but hand-checks cannot be because they are conducted through browsers. My team and I conduct more than 1,000 hand-checks per week. We also frequently discuss hand-checks with our clients, who surf along with us on the phone.

Beyond the SEO industry’s hand-checking we now have the phenomenon of “ego searching”, where people “Google” and “Yahoo!” themselves, friends, and families.

Local search has also altered the search landscape since 2003. People now regularly use search engines to perform address and telephone number lookups — actions that don’t require click-throughs when they find what they are looking for.

As search engines transform into information providers the need for clicking through to deeper Web sites diminishes. In 2006, Jakob Nielsen accused search engines of leeching traffic from others sites. Since then it has become commonplace for major search engines to provide news headlines, stock prices, product prices, and other consumer information once regarded as “commercial content” directly in search results.

Today you can read your email, follow 100 blogs’ RSS feeds, browse the news, engage in discussions on news groups, and perform other tasks without leaving Google. The Google search box connects you directly with dozens of Google services (should mail.google.com really rank first for “email”? should books.google.com really rank first for “books”?).

The more generic queries Google optimizes for itself, the less competitive other Web sites become, the less need for people to click through Google’s search results. Although I said above that the search engine has all but abandoned the portal concept, the truth is that Google has actually found a way to virtually portalize itself without looking like a portal. Hence, the very problem that comScore thought it was solving — counting people who used search services for something other than search — has complicated the metric that comScore championed.

Search is no longer simply search. It’s evolved to include extensive query-based navigation and Google, along with Yahoo! and Microsoft to a much lesser extent, has engineered an infrastructure that trivializes the Web. Google’s Supplemental Results Index, which encompasses as much as 80% of all the content that Google has indexed, ensures that other Web sites don’t rank well for queries. Google’s Universal Search ensures that people now see results from Google News, Google Books, and Google Video search — all services that had low participation prior to Universal Search.

In effect, the search engines are pre-emptively lifting query-based navigation from other Web sites and inserting their own content into the navigational search results. This transformation of the search experience thus indicates that much of the estimated search market share is being withheld from the market. A significant percentage of Google’s assumed 70% search market share is never shared with the rest of the Web.

For these reasons and others I began a few months ago to estimate real search market share on the SEO Theory blog to help people in the search optimization industry realize they were missing out on many opportunities for legitimate, converting traffic. Simply ignoring Microsoft’s search engine (which now drives more search traffic than Yahoo!) ensures that most optimizers are ignoring as much as 15-20% of all search referrals.

One of the best kept secrets in search today is that Microsoft is catching up to Google. While people look at the obsolete query-based metrics, searchers are using other search services besides Google.

If you were to compile August 2008 search market share data per Compete’s publicly available data, you would find that Google only received 24.12% of all the traffic distributed among the top 23 search engines (based on estimated unique U.S. visitors).

Yahoo! captured 23.62% of that traffic. Microsoft’s combined traffic for MSN and Live amounts to 28.94%, more than Google’s share of traffic. Of course, the data Compete makes available includes a lot of non-search traffic (especially for Yahoo! and MSN). Nonetheless, Microsoft has grown into a search powerhouse that the SEO industry fails to recognize.

You can count the number of queries performed all you want, but if you’re not optimizing for the entire search engine industry your actual referrals will be much lower than your potential referrals.

It’s time for us to develop a better search market share metric. We should be looking at number of visitors who leave the search engines through queries, how many times they use search engines per month, how many queries they use versus how many queries they actually click out of, and more. Query-based market share estimates are unsophisticated and uninformative.

If you’re not measuring search market share beyond number of queries performed, you’ve sacrificed your competitive edge.