A number of recent cases involving keyword infringement have involved likelihood of confusion surveys. This post looks at two of those surveys to highlight some of the issues that may arise with confusion surveys in keyword infringement matters.
One very recent decision that involved such a survey is Rosetta Stone v. Google (Fourth Circuit, 2012). You can link to the opinion in this matter in many places, including here.
The search results on Google can be organized into two categories: organic results and sponsored links. While organic results are displayed according to algorithms used by Google, sponsored links are displayed based upon a system where advertisers bid for the best placement.
In 2004, Google began allowing companies to bid on trademarks as keywords. The case of Rosetta Stone v. Google focused on Google’s practice of allowing competing language software companies to bid upon the keyword term “Rosetta Stone”. For more information about the survey and the arguments to include or exclude it, see Rosetta Stone’s opposition to Google’s motion to exclude the survey.
The keyword infringement survey used in Rosetta Stone v. Google was designed to measure the likelihood of confusion by recruiting respondents who (among other criteria) were interested in learning a foreign language, had heard of Rosetta Stone as a provider of foreign language products, use Google, and planned to use the Internet to look for information about learning a foreign language.
During the interview, respondents were asked to enter the term “Rosetta Stone” into a computer, and press enter. They were then shown either a page of test results, which were an actual page of Google search results, or a control page, which did not include any sponsored links. Both types of respondents were then asked questions to measure different types of confusion, including topics such as:
- which links sell Rosetta Stone language software,
- which links are a Rosetta Stone company website, and
- which links are endorsed by Rosetta Stone.
Respondents were counted as confused if they identified certain links as belonging to Rosetta Stone or as endorsed by Rosetta Stone.
As is typically the case with likelihood of confusion surveys, this survey did not measure actual purchase behavior, but rather measured the likelihood of confusion after respondents viewed the test or control page of search results generated by the survey. The survey was designed to measure “initial interest confusion,” and respondents could not actually click on the links they saw.
Another survey, also involving Google, is GEICO v. Google (Eastern District of Virginia, 2005). This case focused on Google’s practice of allowing GEICO competitors to bid for sponsored links using GEICO as a keyword. The survey employed a similar methodology as the survey in Rosetta Stone v. Google. In this case, respondents were qualified as to consider purchasing or renewing auto insurance, and intending to use the Internet to search for auto insurance. As with the other survey, respondents saw a screen shot from a search using the term GEICO. In this case, however, the control screen shot included sponsored links, but to Nike.
This survey was criticized by the court, among other reasons, for using an inappropriate control, for the order in which questions were asked, and for using Google material that did not match Google’s then-current practices.
The standard practices for conducting infringement surveys are still under development. These and other cases raise a number of questions. One such question involves the issue of initial interest confusion. With products that are sold at retail, initial interest confusion can become quite important. As described in an analogy used in an early case on keyword infringement, Brookfield Communications v West Coast Entertainment (Ninth Circuit, 1999), a consumer who is confused by an advertisement may end up physically in a store. For example, a confusing ad for a store or product may attract consumers, who may then end up buying something other than what they intended to save the time involved in to going to another store. In other words, the consumers do not need to be confused to alter their purchase patterns.
It is also true, however, that online consumers can easily click out of one website and head to another website. In the cases cited here, courts ruled that initial interest confusion matters. It remains to be seen whether surveys will measure confusion beyond initial interest confusion.
A second issue that these cases raise is the importance of the visual stimuli used in research. Both of these cases used search result pages as stimuli, and the presentation of both test and control stimuli were central to the evaluation of the surveys in both cases. Selecting the right stimuli is particularly important because the Internet is a fast-changing medium, where both ads and search results evolve over time.
With additional time, observers should expect additional perspectives on what is keyword infringement, and how to measure likelihood of confusion in a keyword infringement survey.
Dr. Bruce Isaacson
MMR Strategy Group