The US Supreme Court has endorsed (6-2) the use of a statistical sample of employees in order to prove class-wide liability in an overtime case. Tyson Foods v. Bouaphakeo (US Supreme Court 03/22/2016). [Read syllabus here.] The trick in any class action is to establish that questions of law or fact common to class members predominate over any questions affecting only individual members. The problem in the Tyson case was that the employer had not recorded how much time employees spent donning and doffing protective gear. Certainly the amount of time was not uniform.
So the plaintiffs hired an industrial relations expert who sampled some employees, calculated an average for all employees, and added that to the individual time sheets.
The Court opined that if the plaintiffs had proceeded with individual lawsuits they would have been able to use statistical evidence. Basically, it comes down to whether the statistical method is reliable, and that is mainly a question for the district court to resolve. You're not going to get a blanket declaration from the Supreme Court.
As for the problem that some employees suffered no loss at all, the district court can sort that out when the jury award of $2.9 million is disbursed.
[For recent decisions and pending employment law cases, see Supreme Court Watch.]