Trademarks, Triggers, and Online Search
Google policy change allowing more ads on TMs. Navigational search: allowing 3d party ads may divert consumer attention to 3d party websites, making search harder—bad for TM owners and consumers? (RT: how do you know that it’s bad for consumers, and not consumers exploring alternatives? The TM owner’s site is usally first in the organic results, so it’s right there if they want it.)
But for nonnavigational search, allowing 3d party ads may be good for TM owners if it increases the amount of useful information.
Empirical data from France and Germany: categorizing over 140,000 different searches. Idea: if someone enters only TM as search term, more likely that this is a navigational search. TM + some other word, likelihood that this is nonnavigational is higher. Number of words also matters (e.g., iPhone + model number)—lower is more likely navigational. Change over time—if someone consults the search engine again, how do they change their search? Likelihood that it’s nonnavigational is higher if they add more words and vice versa. Results: Likelihood of visiting TM owner’s site within 10 minutes increased (+14.7%, 80% of sample) for nonnavigational searches after Google’s policy change, whereas for navigational searches it decreased (-9.2%).
Shift focus of debate from consumer confusion to consumer behavior. A full allocation of property rights to TM owner is not necessarily optimal for TM owner.
Discussant: Amanda F. Myers, USPTO
Dynamic nature of TM—making consumers use TM in new or previously unexplored ways. Owners and nonowners can deploy them. Empirical work can turn a legal argument on its head. (RT: I don’t get it. The legal argument (at least in the US) is about confusion. Evidence of consumer behavior is not evidence of confusion. It’s not even evidence of dilution, in the sense of changing the signal of the mark. As with cognitive models of dilution, empirical work can give TM owners new means of defining what “harms” them, but agreeing that this is a harm is a normative move beyond the empirical data.)
Query whether French and German attitudes towards brands differ.
Bechtold: information on actual purchase behavior isn’t observable from data because of privacy concerns. (Cut off part of the URL.) So we can’t consistently observe whether someone goes to a shopping cart. Duration/time on site—we are exploring this more, but the data are noisy and we aren’t clear why.
Q: why not observe the order of visit and only judge the first visit after the search?
A: it’s on our list to check.
Q: could still classify websites that allow purchase and websites that don’t (e.g. Blackberry). Might see differential impact. Another comment: might see that someone goes to Amazon to buy the thing—that’s not a diverted sale from the TM owner.
A: that’s a manual classification issue.
Q: concern would be that if someone searches on Samsung, then goes to buy some other phone elsewhere, then back to Samsung to buy an accessory/peripheral. (What kind of problem or difference does that represent? I think Bechtold’s answer was something like that—that this is probably too granular.)
Q: effect of autocomplete?
A: can’t observe that, can only observe url.
Q: what about searches with multiple TMs? Amazon + Blackberry.
A: we used the first brand named.
Q: might be able to get category information—someone who goes to BMW then to Google Maps might be looking at a dealer, whereas someone who goes to another car company is still continuing the search.
A: how can we identify navigational search if a TM has multiple meanings including non TM meaning.
Q: we’re thinking about classifying Apple etc.—though in Germany/France, more plausible it’s used as a TM. Often marks we studied didn’t have generic meanings in the languages, but we are trying to figure out if we have a sufficient number of marks to control for this.
David Franklyn, University of San Francisco School of Law & David Hyman, University of Illinois College of Law, Trademarks as Keywords
Study in US, given over 100 keyword lawsuits worldwide, lots of articles & CLE. Judicial “casual empiricism,” initially focusing on diversion rather than confusion. Assumption was that search terms reflected search goals. Assumed consumer knowledge of labeling and architecture. E.g., 9th Cir. assumed consumers could distinguish paid from organic/algorithmic results, and were therefore less likely to be diverted. (My thoughts here: Reasonable consumer is a legal construct, not a completely empirical creature. Many theories make this a feature, not a bug: you should learn that some uses of TMs are not authorized, because of the needs of competition/free speech.) Assumption that use of term in ad text was more likely to confuse/divert. So are consumers confused? Are they even diverted, if they are using TM as an entry point into a general search on the category at issue?
Wanted to determine searchers’ knowledge of search engine labels/architecture: algorithmic results are unlabeled, and you’re supposed to be able to figure out by negative implication that they’re not paid. Google changed from “sponsored links,” which they studied, to “ads.”
Research questions: how often are TMs bought as keywords, and who is buying them? Why do consumers use TMs as search terms? Are consumers actually confused, and if so is it TM confusion, consumer protection confusion (unawareness of what’s advertising), or some other type of confusion? Asked respondents about actual results pages.
Vendor of TM products and competing products: paid 27%, unpaid 3%; TM owner: 13%/44%; vendor of TM products only, 6%/3%; collateral information and sales opportunity vendor, 24%/3%, vendor of competing products only, 6%/3%. When TM owners are present, they are often at the top—30% of the top paid links.
What do consumers look for and expect to find when they click on a link? Products bearing that brand name only, 65% looking for the product/45% expect to find that brand name only. Products bearing that brand name and similar competing brand names, 34% looking and 39% expecting. Complicates the account that people are only interested in the TM and anything else is diversion. (RT: Many TM owners would want to argue that this is still diversion—taking away the “earned” benefits of being “top of mind.”)
Asked them to ID source of links—paid, unpaid, Google’s special marketing team (control), don’t know. Less than half were right about the paid links being paid, and only 51% were right about the algorithmic results being unpaid. People who were younger did better, but not impressive.
Are consumers paying attention to labels? Asked which of these labels they’d seen—sponsored links, sponsored results, ads, commercial ads—49% answered “sponsored results” as a time that Google hadn’t used that in a year, while 46% said “ads,” 33% claimed to have seen the unused control “commercial ads.” (RT: I don’t find this surprising or disturbing in any way. They mean the same thing. People translate to synonyms all the time. It’s a standard result in the memory research—if you show people synonyms/words with similar meanings they think they’ve seen them in previous word lists. There are a significant minority who didn’t remember seeing anything, and that’s an important number.)
Paid results—whether this was a competitor or not. About half of people correctly ID’d the reason a link was present was that it was paid; 15% thought it was Google’s special marketing team (the control); 14% thought that there was an authorization relationship. Not sure was 21% (over and above the people who were obviously guessing).
Was it fair/appropriate for the paid link to appear against Mercedes? Direct competitor: yes, 35%, no, 39-40%; Gorgeous Luxury Vehicles (which might be selling the Mercedes too): yes, 44% and no 32%. 25%: don’t know.
Consumer ignorance, obliviousness, and indifference is pervasive—consumers aren’t going to police this. (RT: Which also suggests something about major components of brand value ….) Armchair empiricism is wrong. Consumer goals and expectations are heterogenous. Consumer knowledge of search architecture and labeling is limited.
Hard to make a strong case for TM likely confusion (source, sponsorship, affiliation), but plenty of other types of confusion. 14% is low for a TM case—also note that there isn’t a TM control, that is a link that should clearly be deemed unaffiliated. Competitors don’t seem to buy TMs all that often, and consumers don’t seem to have a big preference to click on them, so it’s hard to see a substantial amount of confusion. Big question: does anything here translate to consumer harm? Rational ignorance is often ok; people have other things to do than understand Google’s architecture.
Future of TM: whether free riding rationales should make more of a difference in TM law? FTC has communicated on the issue of search result presentation.
Discussant: Carolina Castaldi, Eindhoven University of Technology
Bringing the consumer into the debate. Results: maybe no worries, since consumers don’t seem to have expectations to disrupt/may even appreciate alternatives. Key variable: type of product consumer is looking for. Study here: Mercedes—durable good, high risk purchase, effort in searching is likely to be higher.
A: we did test some other products, though not in the study where we asked people to look at actual search pages. Defense of looking for the real consumer as opposed to the construct of the consumer in the law. (I’m actually a big fan of this, despite my cautionary words above—which boil down to looking at the tradeoffs; we always need to ask “compared to what?” when we try to set a standard, because averting confusion of any kind has costs, and the authors are well aware that we need to ask whether we can do anything to reduce confusion below a certain level as well as what the costs are.) Reminder that eBay tested whether it needed to buy its own marks on Google, and it didn’t—for famous brands, it doesn’t make economic sense to defensively purchase a TM to keep competitors away. Followup paper will explore the economics of defensive purchasing. Not a TM infringement issue, but has arisen because of the dynamics of search.
We didn’t come out convinced that a prohibition on sale/purchase of TM as keywords was justified, nor did we feel that the current TM regime mapped well onto the consumer protection issues. FTC requires clear and conspicuous differentiation between paid and unpaid results, and it hasn’t been effective or enforced. Supermajorities of respondents said the difference wasn’t clear, and said they’d click less on paid ads if they knew. This is not a TM problem. Using likely confusion doctrine, through initial interest confusion, to capture various forms of consumer protection/anti-free riding ideas—and that isn’t a good idea.
Suing competitors v. suing Google? Compare to memo from tobacco lawyers: aggressive posture on discovery—the way we win is not by spending all of RJR’s money but by making the other son of a bitch spend all of his. This may be Google’s strategy. We are working to unseal the record on Rosetta Stone. American Airlines succeeded in backing Google down very early on, but that’s about it. There’s also something going on with Starbucks. Hard to see a value proposition in suing Google, or most competitors. (The authors disagree on this.)
Q: how many people thought that the pages were ad-free?
A: appreciable number of people get both the two paid regions wrong. Surveys in the UK as well. In our new surveys, you select an answer while looking at the relevant part of the page, and 30% are still not saying it’s an ad. Our hypothesis is now that you can’t fix it, because of learned behavior over where we think the relevant regions are. In the beginning there were no ads on the top, and they gradually put them on the top, then gradually increased the number of ads on the top, then gradually changed the shade of the ad box shading, and gradually shrunk the size and decreased the contrast on the “ad” disclosure. And then they added in Google Shopping, which nobody understands.
In reaction to Bechtold’s comments: search is more dynamic than what we could test. Ultimate test would be simulation, asking people to perform a search/decision task. They go back and forth and see something interesting and go off. We tried to ask people how they would behave, but self-reported knowledge isn’t as good as observational. Ben Edelman has tried to simulate, with mixed results. Goals are ambiguous, but (lack of) knowledge of search page is robust; whether that leads to ultimate confusion in a TM sense is very open to question.