The production of records from Sprint after the close of discovery gave defendants the missing evidentiary foundation for their consent-predominance argument. Defendant searched its accountholder database for the names of subscribers and users provided by Sprint. Certain common names matched with thousands of Yahoo! user accounts, and even limiting the results to less common names (using a somewhat arbitrary measure), defendant can now show that tens of thousands of potential class members may have consented. For example, defendant found that 47,672 unique names from the Sprint data matched five or fewer Yahoo! accounts (indicating that, with further inquiry into the Yahoo! accounts, a more precise match between Sprint user and uTos acceptance is possible). [281-1] ¶ 18. Defendant also searched its accountholder database for the associated telephone numbers provided by Sprint. It found that in 65,061 cases, a number that received a Welcome Message matched a number that was provided to defendant in registering a Yahoo! account. [281-1] ¶ 27. Excluding large group plans (e.g., a corporate account with thousands of users) results in 47,109 instances, according to defendant. See  at 9. The association of thousands of potential class members to a Yahoo! account (and therefore, in turn, an accepted uTos) is a concrete showing that individual consent issues will predominate the case moving forward. To drill down on whether these class members accepted the uTOS, and thereby provided consent, the facts surrounding each person associated with a matching number would have to be explored. This could not be done in a single proceeding through records and an expert witness because the requisite identity-match may be dependent on class member testimony. Defendant now has evidence sufficient to justify an individual consent inquiry for a significant percentage of the class (perhaps between 20 to 25%, maybe more). See  at 8–9. Plaintiff says that the evidence of consent identified by defendant is inconclusive and does not prove that a significant percentage of the class members consented. While it is true that more would need to be done to prove consent, defendant does not need to prove consent to decertify the class. It just needs to show that proving consent requires individualized analysis such that the class does not meet the predominance requirement. It has done that much. When the class was certified, it appeared that determination of consent was susceptible to class-wide proof based on the information then available to the parties—by analyzing user information in defendant’s records to identify individuals who had agreed to the uTOS, and deferring further inquiry until a claims administration process. But the information disclosed by Sprint gives defendant an opportunity to present its defense on an individualized basis at a trial, and I will not exclude that evidence. Plaintiff requested defendant’s evidence of prior express consent during discovery, and defendant never produced this data. In fact, defendant did not pursue evidence of consent in discovery, and I denied its request to reopen discovery to pursue consent evidence. . The Sprint data came into the litigants’ hands as a result of plaintiff’s efforts to identify class members and issue notice. Defendant took that data and, using custom scripts, conducted an analysis that was not readily available during fact discovery. [281-1] ¶ 10. Parties have a continuing obligation to supplement discovery responses, and perhaps defendant should have conducted and disclosed its analysis sooner. But excluding the analysis would be too harsh a sanction. Defendant did not seek out information related to associated telephone numbers in discovery, and, as plaintiff points out, neither party requested it from Sprint. This development is not attributable to some discovery misconduct by defendant. Ultimately, the truth-seeking function of litigation would be better served by testing the evidence, not pretending it doesn’t exist. Plaintiff proposes that the class be redefined to exclude class members who might have consented. Jettisoning people who consented would not create an impermissible fail-safe class (because whether the PC2SMS system is a prohibited automatic telephone dialing system is still a contested issue unresolved by the class definition), but plaintiff does not propose a new class definition. Without a specific definition that has been tested against all the Rule 23 requirements, I decline to adjust the class further. Moreover, counsel’s idea to exclude potential class members based on defendant’s investigation invites concerns over adequacy of representation and loyalty to the class. Decertification, not redefinition, is the appropriate step in light of defendant’s showing that individualized consent inquiries will predominate.