It should be no surprise that courts in the Northern District of California–home of silicon valley and the tech-saviest judges in the nation– tend to be data nuts. While some courts may gloss over the technical aspects of TCPA class cases, you’re likely to get a closer-than-average look at your class data sets and arguments up in the Bay Area. A perfect example of this phenomenon is the Hon. William Alsup’s big new TCPA certification denial in Revtich v. Citibank, Case No. C17-06907, Doc. no. 150 (N.D. Cal. April 28, 2019).
In Revtich, the Plaintiff sought to certify a class definition that did not turn on any specific data-tracking elements of Defendant’s system. (This, btw, appears to be the new trend out there–keep an eye on it.) But although they ran, they couldn’t hide from the infirmities in their data set.
The definition proposed in Revtich included, inter alia, all persons called “where such person was not listed in Defendant’s records as the intended recipient of the calls.” To identify class members, therefore, the Plaintiff must somehow prove that an individual picking up the phone was not the person Defendant intended to reach.
Sounds tricky no? Yes.
To start with, although Citibank does track “wrong number” dispositions these entries are commonly entered in error by agents and debtors often claim to be receiving calls on “wrong” numbers merely to have calls stop. Plaintiff proposed to find the “true” wrong numbers by filtering out numbers that were once marked as ‘wrong’ but were later determined to be valid, and sending the rest to TransUnion to see if Citibank’s customer was the same as the subscriber identified in TransUnion’s ever-so-accurate data. If not, Plaintiff says, the number is owned by a class member. Hmmm.
Two rather obvious flaws with this process jump out at you. First, who says TransUnion’s data is accurate? Second, just because the name in Citibank’s record isn’t on the subscriber list doesn’t mean that its a wrong number–maybe Citibank’s customer is on a family/shared/work plan?
Well the experts tested the data in Revtich and the findings were rather interesting. First, of a 20,000 number sample set the Plaintiff’s expert was only able to identify 176 phone numbers within the definition. But then Citibank dug into those 176 numbers and found that–unsurprisingly– many were actually related to a customer even if the susbcriber information didn’t match.
For instance in one circumstance presented to the court a father’s number was provided by a son in connection with one account and then by the son in connection with a different account. The number was subsequently identified as a “bad” number to reach the father, but remained perfectly valid to reach the son. Nonetheless Citibank subsequently called the number seeking the son but reached the father–which would place the father within the class definition although the call was consented to be the son. (The court does not specifically address the issue of whether the son could consent for a call answered by the father but it is necessarily implied that the answer is yes–and the answer IS yes assuming the son regularly used the phone.)
Unpackaging the father/son story further, the Court distills the issue– phone numbers can be associated with multiple accounts owned by different people. So just because a number is marked “wrong” on one account does not mean it is “wrong” for calls placed to a different person or on a different account. Citibank’s expert found “many” such instances within the Plaintiff’s identified expert set.
The largest issue, however, involved Plaintiff’s expert’s handling of the TransUnion data itself. The TU data included a “last seen” date suggesting a date that a number may have switched hands. The Plaintiff’s expert did not consider that data very important and did not consider it. But Defendant’s expert demonstrated that using this data would have eliminated 120 of the 176 numbers Plaintiff’s expert had identified as ‘wrong’ numbers. Indeed, Defendant’s expert opined that all 120 of those numbers actually were associated with the account holder Citibank was trying to reach at the time the call was made– that’s an incredible 68% error rate in Plaintiff’s methodology!
There is also a missing data issue. Citibank did not begin tracking historical data for consent codes until November, 2017 (and one wonders why they began tracking that data at all). So for numbers called for the last time before that date the only phone quality that would be available would be the most recent. But as phone qualities change all the time compiling an accurate historical look at the code present on each number at the time of each call would require resort to a file-by file-review–which, of course, destroys commonality.
Setting all of this aside, one fact almost certainly doomed Plaintiff’s certification bid from the start– he would not have been identified using his own expert’s methodology! This is because although the parties agree he received a wrong number phone call, his account was never coded with the wrong number code that Plaintiff sought to use for purposes of identifying class members. And that demonstrates why the Plaintiff’s methodology is most flawed–the data does not track the class definition.
Importantly, the Court refused to let the Plaintiff shift the consent narrative from an issue of predominance to one of ascertainability (which has no teeth in the Ninth Circuit.) As the Court puts it:
The problem here is not identifying the individuals who fall within plaintiff’s proposed class. Rather, the problem is that adjudicating the claims of those who do fall within plaintiff’s proposed class would devolve into the tedious resolution of individualized issues based on individualized evidence.
That last point is critical for folks seeking to defeat certification in the Ninth Circuit– do not get bogged down on ascertainability. Difficulty finding class members alone is rarely an issue defeating certification. Focus instead on why the claims of class members–once they are found–cannot be properly adjudicated using general proof.