Sunday, August 02, 2009

Fair and balanced: battle of the credit giants

Fair Isaac Corp. v. Experian Information Solutions Inc., 2009 WL 2252583 (D. Minn.)

There are three major credit bureaus in the US: TransUnion, Experian, and Equifax. They have 100% of the market for aggregated consumer credit data, which they use to generate credit reports sold to lenders. Because credit data reporting is voluntary, and individual lenders have different agreements, each bureau often reports somewhat different aggregated credit data for the same consumer.

Fair Isaac developed an algorithmic credit scoring model, which lenders use to evaluate individual consumers. These FICO cores quickly came to dominate the credit scoring market, representing from 74% to 94% of various segments. But Fair Isaac typically doesn’t sell FICO scores directly to lenders or consumers because it doesn’t have direct access to the aggregated credit data. Instead, it has scoring agreements with the credit bureaus, allowing them to sell FICO scores to lenders and consumers, typically as a part of a bundle with a credit report containing the underlying aggregate data. Fair Isaac has developed variants for each bureau, and it gets a royalty on each sale.

Fair Isaac’s royalties represent a significant component of the bureaus’ cost of doing business, so the bureaus developed their own in-house credit scoring models. But they were largely unable to convince lenders to switch away from FICO. Lenders prefer the tri-bureau nature of FICO scores, which account for the differences in aggregate credit data among bureaus. The credit bureaus are unwilling to share their data, so they can’t make their in-house scores into tri-bureau scores.

The bureaus met to discuss developing a tri-bureau score in 2003-2004, allegedly with the goal of eliminating FICO scores from the market. The problem was that if one bureau acted alone in trying to switch, there’d be a huge risk of losing customers who wanted continued access to FICO scores to the other bureaus. Thus, a consultant recommended that the bureaus “act in concert.” (Given that Fair Isaac’s antitrust claims were dismissed, I have to wonder what it would take to have a smoking gun for antitrust purposes these days.) Anyway, in 2006 the bureaus announced the VantageScore scoring model, based on all the bureaus’ data, to which all three bureaus had 1/3 ownership and a license. The bureaus have claimed superior predictive power for VantageScore compared to FICO, because VantageScore has unmatched access to all three bureaus’ data sets, uses a single algorithm rather than three modified algorithms, and uses “characteristic leveling” so that the results are consistent no matter which bureau’s data set is used.

Fair Isaac sued for antitrust and Lanham Act violations. The court granted summary judgment to defendants on the antitrust claims because Fair Isaac lacked standing, proving to me that antitrust law is even worse off in this area than false advertising law, and holding that Fair Isaac hadn’t shown sufficient injury because VantageScore only has a couple of percentage points of market share.

Of interest to me, Fair Isaac alleged that the VantageScore 501-990 scoring range is confusingly similar to the trademarked FICO scoring range of 300-850. (Why start above zero? Is it just too mean to give someone a score of 0? There’s probably a behavioral psych reason why this increases access to credit.) Fair Isaac also alleged false advertising.

Trademark: Fair Isaac has several registrations for the 300-850 range for credit scoring and related services. Defendants argued that the mark wasn’t distinctive, attempting to rebut the presumption of distinctiveness conferred by registration. First, the court concluded that the mark was descriptive. Though there are cases (wrongly) finding suggestiveness when the mark alone isn’t enough to identify the goods/services, the court went with the (correct) majority view that descriptiveness is assessed from the perspective of a person who knows the context—the goods/services at issue. In context, the 300-850 range is descriptive, even though it doesn’t precisely map to the actual ranges Fair Isaac provides to any of the bureaus (397-871, 368-839, and 407-829). Fair Isaac employees agreed that the 300-850 range was approximately the range of actual credit scores. Fair Isaac also uses the range descriptively: “A FICO score is a 3-digit number ranging from 300-850 that represents your credit rating,” etc. Look at those ranges—so even someone with terrible credit will get a score around 100 points above the announced minimum. Again, it’s hard to resist the conclusion that something is going on with anchoring and other perceptual quirks, and that the something has to do with loose credit.

Fair Isaac argued, indeed, that the ranges were arbitrary and that any other range could have been selected. But that’s not the key to descriptiveness of a number that encompasses all three actual scoring ranges. Nor is the Court persuaded that the term 300-850 is more than merely descriptive simply because the range of 300-850 applicable to FICO scores was chosen arbitrarily and any other range of numbers could just have easily been selected. “[T]he pivotal question is whether the term was chosen arbitrarily in light of the nature, ingredients, qualities, and characteristics of the product.” The court didn’t cite the extensive PTO precedent that numbers are descriptive, but still got to the result: Fair Isaac needed to show secondary meaning. (Now, registration of a descriptive term should result in a presumption of secondary meaning, at least if the registration was granted on the basis of secondary meaning; all the Fair Isaac registrations were filed as ITUs but granted as use-based marks, and I haven’t looked up whether Fair Isaac submitted evidence of secondary meaning.)

Defendants argued that Fair Isaac didn’t promote the mark as a brand, and that their survey showed only 2% of respondents associated the mark with a single source. Fair Isaac argued that secondary meaning could be inferred from defendants’ copying of similar three-digit ranges, and from evidence of confusion: “customers complained when they realized the credit score they had purchased was Trans Union's or Experian's in-house score or a VantageScore credit score rather than a FICO score.” Intentional copying and confusion are evidence of secondary meaning; thus, the court found genuine issues of material fact on the issue, meaning the court didn’t at this point need to consider Fair Isaac’s alternate licensee estoppel argument that the bureaus couldn’t be allowed to challenge the validity of the marks they license.

Fair Isaac also alleged infringement based on defendants’ purchases of keywords containing Fair Isaac trademarks; defendants argued that there was no infringement because the resulting ad text didn’t include Fair Isaac trademarks. But the caselaw, unfortunately, doesn’t support this argument. There can still be likely confusion; it’s an issue for a factfinder, and Fair Isaac has an expert who opined that confusion was likely.

False advertising: defendants said some allegedly false things about (1) whether an appreciable number of lenders actually use in-house credit scores and VantageScore credit scores in making lending decisions and (2) VantageScore’s predictive ability compared to other scores on the market. Fair Isaac argued that less than one percent of lenders used Trans Union’s in-house score, nobody used Experian’s in-house score, and no lender used VantageScore when the statements at issue were made. The court, however, determined that the statements didn’t convey an implied message that an appreciable number of lenders used in-house scores or VantageScore in making lending decisions. The statements were things like: “Most lenders would view your creditworthiness as very poor,” “Know where you stand no matter which credit bureau your lender checks,” “the same type of score that lenders see,” and “Most lenders offer their ‘good’ rates to consumers in this category.” These simply convey that the score is “indicative” of how lenders assess creditworthiness, and at a minimum are capable of more than one reasonable interpretation, thus not literally false or false by necessary implication.

As to VantageScore’s predictiveness, the statements at issue included that VantageScore “allows credit grantors to evaluate consumer creditworthiness with significantly greater precision,” is “more predictive than what’s in the market,” is “the most accurate scoring algorithm attainable,” and is based on the “most up-to-date information available.” Fair Isaac submitted evidence that FICO scores sometimes outperformed VantageScore in head-to-head testing, and that more often and not there is no significant difference in predictive power. The court found that these were vague, subjective representations of superiority: puffery.

Thus, the false advertising claims failed as a matter of law. (Did Fair Isaac not ask for an opportunity to prove implied falsity of the first set of claims?)

No comments: