Rebecca Tushnet

Richard Craswell’s work on controls in surveys remains a must-read in this area: choosing a noninfringing control is often very important because it can determine the amount of net confusion. One particularly interesting point here: the “control” shows very high levels of confusion (38%), which would ordinarily seem like a lot. PNC would ordinarily have a strong incentive to argue that disclaimers don’t work. But that would run up against a strong preference for truthful speech, especially if the alternative seems to be that PNC can control what apps its customers can use. There’d be some obvious competition law problems with that as well as nominative fair use (what Plaid argued before PNC dropped the argument that the current screen, which is not entirely unlike the control, infringed). Craswell makes the argument that there is an implicit cost-benefit analysis in control selection: we’re asking what’s the least confusing option that’s worth it. If some confusion is irreducible but we still want the activity to continue, that’s the level of confusion we should accept. That could indeed justify a control with 38% confusion, but that may not be something many TM owners want to admit. And it calls into question the broad definition of “affiliation” confusion that courts have adopted–often by assuming that “affiliation” in the Lanham Act means whatever consumers think it means in response to survey questions, although it could reasonably be read more strongly or consumers could be educated/asked for their definitions thereof.

Anyway, after the survey showed the stimuli, it then asked: “which company provides this credentials screen,” “does the company that provides this credential screen have a business affiliation or connection with another company or companies,” “with which other company or companies does the company that provides this credentials screen have a business affiliation or business connection,” and “did…the company that provides this credentials screen receive approval or sponsorship from another company or companies?”

PNC’s expert Kivetz concluded that 79% of the participants in the test group were confused into believing that the company that provides that credentials screen either (1) was PNC (75%); (2) has a business affiliation or connection with PNC (5.9%); and (3) received approval or sponsorship from PNC (5.4%). (Id. at 59–60). In contrast, participants in the control group experienced a confusion rate of 38.1%. The net confusion rate (the difference between the two confusion rates) was 41.3%.

Plaid challenged the survey based on the control, the lack of an interactive process in the survey, and the failure to limit the survey to PNC customers. The sample issue didn’t merit exclusion; the survey included potential PNC customers by asking potential respondents whether they engage in online banking and geographically limiting the population to states in which PNC had a physical branch location. Anyway, any effect of the allegedly skewed sample was “speculative,” since the broader question was about “whether the average consumer who connects a bank account to cash payment and investment account fintech applications would be confused by the use of a given set of marks (PNC’s) on Plaid’s user interface(s) during the connection process.” So too with the static visuals—that increased survey completion, and if respondents had actual options they might have tried to reach a different bank’s credentials login screen. The survey explained to respondents what was happening; they could click on and read the Plaid privacy policy before going forward.

The control group argument gave the court more pause. The control screen “displayed a disclaimer that was not present in the Plaid Link user interface, at least at the later stages of a consumer’s interactions with Plaid Link.” This could have been “so visually different (and perhaps, so obviously affiliated with Plaid) from the PNC institutional and credentials login screens that it diluted the confusion results on that side of the survey.” But the control was also similar, though not identical, to the credentials screen Plaid actually uses today. This was a question of weight, not admissibility. The disclaimer left nearly 2/5 of respondents confused, which undermined the assertion that the control wrongly pushed people away from PNC.

However, the expert’s opinion on “tarnishment” was excluded since it relied on non-record evidence of “bad acts” by Plaid and he only speculatively linked that to PNC, rather than showing a basis relying on a reliable process or analysis.

Plaid’s consumer confusion expert, Dhar, used a “consumer journey approach,” designed to mimic what users would be seeing when deciding whether to buy or use a given product or service. He opined that users were unlikely to be confused by Plaid’s use of PNC’s marks, especially given consumers’ ultimate goal of connecting their bank accounts to a given fintech app. He also opined that any confusion wasn’t material, based on internal Plaid testing, “which purportedly shows that the effect of the usage of bank marks on the institutional login and credentials login screens had minimal impact on consumer conversion (that is, minimal effect on whether consumers entered their banking information into the fintech app via Plaid Link).”

The court declined to exclude Dhar’s testimony. He provided context that might assist a jury, opining that “by the time a given user encountered a Plaid Link screen within a fintech app, the decisions to (1) download the app, (2) use it, and (3) link a bank account were likely already made,” meaning that confusion was unlikely. “While an expert’s application of their own experience and of principles in the field may not be as empirically rigorous as an experiment or a survey, FRE 702 does not bar the admission of more ‘qualitative’ expert testimony.” His methodology was not novel or pseudoscience; the consumer journey approach is “well recognized” in the field of consumer behavior.  

He wouldn’t be permitted to testify on the ultimate likelihood of consumer confusion, but he could testify as an expert about how the considerations outlined in his report impact the applicable factors: “the care consumers take in using fintech apps and Plaid Link and/or the relationship of Plaid Link and PNC in the minds of consumers.”

The court also allowed Dhar’s materiality opinion.  “Plaid ran a series of internal tests and studies that addressed the impact of Plaid’s use of bank logos in its user interface, and the notion that the data were unreliable solely because the data came from Plaid is inaccurate.” He explained the internal testing in his expert report, with detailed descriptions, and he applied scientific principles to the data, including in his visualizations. “Plaid was experimenting with different institutional selection and credentials panes for years, seeking to measure conversion rate, i.e., whether users would be more inclined to enter their banking information depending on the presentation of the given user interface. These tests are squarely applicable to one of the ultimate merits issues in this case: whether consumers were more or less likely to enter their banking credentials when Plaid used PNC’s marks.”

The court allowed PNC’s damages expert’s disgorgement analysis, though not his “contributed capital” damages theory, which was based on the idea that Plaid’s use of PNC’s marks constituted a forced investment in Plaid by PNC. That latter had issues of fit and reliability.

“PNC never made an investment in Plaid, and the notion that the fact finder in this case should view a portion of Plaid’s enhanced valuation over time as directly and proportionally attributable to the connections that Plaid made between PNC consumers and fintech apps during a one-year span in Plaid’s early days in a straight-line fashion is the kind of speculative opinion, unmoored from scientific rigor, that courts are to exclude under FRE 702.”

The disgorgement opinion, though, was fine because, given the statutory burden-shifting, it was ok to assume that 100% of PNC conversions were attributable to Plaid’s use of PNC’s marks.

Plaid’s damages expert, like its consumer expert, relied on Plaid’s internal testing suggesting that PNC customers using Plaid Link to connect their bank account to a fintech app “would still have connected their PNC account…95 percent to 99 percent of the time,” regardless of whether PNC’s marks were displayed. It was ok for the expert to rely on studies she didn’t conduct, especially a study that perfectly fit a key question here.

Duelling marketing experts also mostly got in. PNC’s marketing expert opined that Plaid benefited from the usage of PNC’s marks and that Plaid’s usage of PNC’s marks harmed PNC’s brand. His report purported “to demonstrate how PNC built its brand, how it continues to invest in its brand, how valuable its brand is, how Plaid utilized PNC’s brand (and the brands of other banks), and how that usage impacted PNC’s brand.” The court excluded his opinion regarding the general risk of harm to PNC’s brand from Plaid’s use, but not the rest of it. (Given that 2019 is now several years in the past, presumably there’s also real-world data about whether the brand was harmed.)

It was ok to use a qualitative analysis of “bad press” that allegedly came about from Plaid’s screens’/CSRs’ criticism of PNC. This went to the claim that required evidence of damage to goodwill (that is, false advertising). But his opinion that the mere use of PNC’s marks, in and of itself, put PNC’s brand at risk wasn’t reliable; it was speculation rather than expert analysis.

Beyond that, it also appears to the Court to be nothing more than an argumentative truism, akin to saying that a person lending her car to another necessarily places all of the assets of the lender at risk in the event the loaned car becomes involved in an accident. Adding the patina of an expert opinion to such a truism does not aid the finder of fact and is therefore unnecessary, as that is an argument that PNC can make without relying on expert testimony. Under FRE 702, PNC has not met its burden in demonstrating the reliability of this particular opinion. This specific aspect of Dr. Carpenter’s testimony—that Plaid’s use of PNC’s marks inherently placed PNC’s brand at considerable risk—is therefore excluded.

Plaid’s marketing expert rebutted PNC’s experts. It was also ok for him to use qualitative analysis.

PNC also offered proposed expert testimony on Plaid’s cybersecurity in “seeking as part of its damages out-of-pocket costs incurred [by PNC] from fraud on PNC customers that used Plaid Link.”  But the expert was unable to link the use of the trademarks to that harm, as opposed to Plaid’s retention of customer authentication information. Here, it mattered that the record showed that “at least some meaningful portion of PNC customers would have used Plaid Link even without visibility of PNC marks.” And Plaid’s central causal contribution was allegedly storing PNC customer data and then, critically, “auto populating” the “challenge question” authentication credentials that a PNC customer previously entered into the Plaid Link screens. That just wasn’t sufficiently tied to the trademark claims. The experts were unable to quantify or differentiate the harms to PNC that were caused by the marginal customers who might have been driven by the use of the marks.

However, if Plaid relied on its own cybersecurity processes or questioned those of PNC, expressly or by implication, the court might allow an expert to opine on “the mechanics of how authentication credentials operate generally and any vulnerabilities such would foster.”

Also, PNC would be allowed to use lay witness testimony that it contended demonstrates that “Plaid’s true purpose in using PNC’s marks was not to ease consumer use in connecting to fintech apps but was instead to increase Plaid’s data repository of consumer banking information for its own purposes.” And PNC would be permitted via lay witnesses to state generally what motivated it to alter how it dealt with Plaid, and how/why its limitations on PNC customer access to fintech apps via Plaid Link came to be, since that’s relevant to intent. “Lay or expert testimony as to who caused/did not cause the 2019 Cybersecurity Event, and the details ‘under the hood’ and/or explanations of that Event, will not be permitted, as such would readily lead to substantial jury confusion in light of the actual claims/defenses in this case and would likely generate substantial undue prejudice that eclipses any probative value under FRE 403.”