Penn State researchers did a study and found that web searchers who evaluated identical search-engine results preferred Yahoo first and Google second even though more claimed to use Google regularly. These results, they say, provide evidence that branding matters as much on the Internet as off and that there is "carry over" branding in effect too.

Here is their method - they copied identical Google results pages (“camping Mexico,” “laser removal,” “manufactured home” and “techno music”) and attributed them to four different search engines -- Google, MSN Live Search, Yahoo! and an in-house engine they created for the study.

Then the researchers showed the faked results to 32 participants who were asked to evaluate the engines’ performance in returning relevant results.

Despite the results being identical in content and presentation, participants indicated that Yahoo! and Google outperformed MSN Live Search and the in-house search engine.

“Given that there was no difference in the results, all of the search engines should have had the exact same score,” said Jim Jansen, assistant professor in the College of Information Sciences and Technology (IST) and lead researcher. “Some emotional branding is having an effect here.”

Participants ranked results from Yahoo! more relevant across the four queries.

Given that many of the participants said they used Google to search, Jansen said he was surprised that Yahoo! came out on top. Its total scores were 15 percent above the average for the four queries while Google’s total scores were just 0.7 percent above the average. Future research will consider whether participants “carried over” satisfaction with other products when ranking search engines, Jansen said.

The researchers were motivated to understand why Web users gravitate toward a handful of search engines when there are about 4,000 search engines that have similar technologies and similar interfaces. The performance -- defined as the ratio of relevant documents to the total number returned at some point in the results listing -- of those search engines also is practically the same.

To determine each engine's “performance,” participants rated the returned results on a three-point scale: very relevant, somewhat relevant, and not relevant. After averaging the scores, the researcher determined an average -- about 36 percent of all results were judged relevant to the query.

The researchers then looked at each engine’s “score” to determine whether it fell above or below the average.

AI2RS, the search engine created in-house with no brand-name recognition, fared the worst. The researchers calculated its average precision rating as 10 percent below the average although AI2RS had the highest score when the query was “laser removal.”

Jansen and co-author, Mimi Zhang, an IST graduate student, detailed the study in a paper, “The Effect of Brand Awareness on the Evaluation of Search Engine Results,” at the recent Computer/Human Interaction 2007 Conference in San Jose, Calif.

Source: Penn State