Earlier this week, the Sixth Circuit released an interesting opinion addressing the use of representative evidence in “collective actions” brought under the Fair Labor Standards Act. As discussed below, the Court held that uniform testimony from dozens of individual employees can establish liability without the need for statistical evidence. At the same time, the decision yields some important questions regarding the use of statistical sampling in future cases.

In Pierce v. Wyndham Vacation Resorts, Inc., a group of over 150 sales employees at four Wyndham resorts in Tennessee alleged the company failed to compensate them for overtime. Plaintiffs fell into three different groups: “front-line” salespeople who sold ownership interests in Wyndham timeshares to new customers; “in-house” salespeople who sold upgraded interests to existing owners; and “discovery” salespeople who sold “trial packages” to prospective owners. All three groups received commissions and a minimum-wage draw (deducted from any commissions paid). Plaintiffs, however, alleged the company systematically required salespeople to underreport their hours and altered timesheets to avoid paying any overtime.

In an opinion by Judge Sutton, the Sixth Circuit affirmed the district court’s decision to allow two types of sales employees (front-line and in-house salespeople) to proceed as a collective action, and to use “representative evidence” to prove their claims against Wyndham. Although they sold to different customers, the Court held that collective treatment was appropriate because they performed “identical tasks.” Both groups arrived at the same time each morning to attend a mandatory sales meeting, and participated in the same types of sales activities (tours, party weekends, etc.). And Wyndham subjected both groups to a “common policy of not paying them overtime, even when they worked over 40 hours per week.”

With respect to the third type of sales employees (discovery salespeople), however, the Sixth Circuit held that the trial court erred by including them in the collective action. Although discovery salespeople were subject to the same illegal overtime policy, they were not similarly situated to the other two groups because their hours were not necessarily the same – they sold different products to different customers via different sales activities. As the panel majority explained:

A common policy cannot overcome the factual differences between the discovery employees and the other salespeople (what they sold and when they started work), which goes to determining the heart of the claim (the total hours worked each week).

Because the district court included the discovery employees in its collective damages calculation, the panel majority reversed in part and remanded for a recalculation of damages. In a separate opinion, Judge White explained her position that the discovery salespeople were similarly situated to the others.

The Court’s discussion of the “common policy” necessary to sustain a collective action under the FLSA is worth a read, as it illustrates the type of evidence required to demonstrate that employees are similarly situated. Plaintiffs presented a wealth of evidence regarding the roles and responsibilities of front-line and in-house salespeople, but only one discovery salesperson testified at trial (and his testimony failed to establish that discovery salespeople performed the same tasks as the other two groups).

What caught our eye, however, was the Court’s discussion of plaintiffs’ use of “representative evidence” to establish liability and damages – an issue we recently had the opportunity to brief in the class action context (although the Court ultimately did not have to reach that issue). From that perspective, a few things from the Court’s opinion in Pierce are worth noting:

1. The Court relied on consistent testimony from dozens of witnesses regarding an “across-the-board” policy to establish liability.

As its name suggests, “representative evidence” allows a group of plaintiffs to establish liability and damages for testifying and non-testifying employees alike through the testimony of individual plaintiffs, who testify as “representatives.” The Supreme Court approved the use of such evidence in Tyson Foods, Inc. v. Bouaphakeo, 136 S. Ct. 1036, 1046 (2016), holding that plaintiffs may rely on representative evidence to establish class-wide liability when “the evidence is reliable in proving or disproving the elements of the relevant cause of action.” The Sixth Circuit followed suit in Monroe v. FTS USA, LLC, 860 F.3d 389, 408 (2017), holding that “representative testimony” may be used to “establish a pattern of violations that include similarly situated employees who did not testify.”

Importantly, however, neither Tyson Foods nor Monroe establish a bright-line rule permitting the use of representative testimony in every collective action. Rather, the question arising from those cases is whether the evidence offered by plaintiffs is “reliable” and “representative” of the group as a whole, so that it can be fairly extrapolated to non-testifying group members.

To make this showing, plaintiffs often rely on statistical evidence. In Tyson Foods, for example, plaintiffs relied on an expert study examining the average amount of (uncompensated) time required for various types of employees to put on and remove protective equipment (“donning and doffing”). The parties debated the statistical validity of the expert’s conclusions, and whether they established a common injury.

In Pierce, however, the Sixth Circuit relied on something far more straight-forward: unanimous testimony from dozens of employees that Wyndham subjected all salespeople to a common, across-the-board policy of underreporting hours and altering timesheets to avoid paying overtime. The panel noted that 43 of the 145 sales employees properly included in the action testified at trial, and that those witnesses “consistently said that Wyndham required them to underreport their time or altered their recorded time.”

The absence of any countervailing evidence eliminated the need to rely on statistical evidence. As the Court explained, “the collective-action framework presumes that similarly situated employees are representative of each other and have the ability to proceed to trial collectively” (quoting Monroe). Wyndham failed to rebut this presumption with any evidence that individual salespeople were treated differently. Uncontested testimony regarding a common policy was thus sufficient to establish liability.

2. The absence of countervailing evidence allowed the Court to sidestep some interesting issues regarding the use of statistical evidence.

Although Wyndham did not present testimony that individual salespeople were treated differently from one another, it did challenge the “sample” of testifying salespeople offered by plaintiffs at trial. The Court’s response to this challenge is worth unpacking, as it sheds some light on potential issues that may arise in future cases in which a defendant does successfully rebut the presumption that the individual experiences of testifying class members were “representative” of the entire group.

First and most importantly, the Court held that a valid statistical study is not required in every case. As the Court explained, “Tyson did not discuss expert statistical studies because they are the only way a plaintiff may prove’ his claim by representative evidence under the Act.” Rather, a statistical study is merely one means of demonstrating that evidence can be reliably extrapolated to all members of the group. (As Pierce demonstrates, unanimous testimony from a large group of individual employees will also do the trick.)

Second, if Wyndham had successfully rebutted the presumption of similar treatment, this case would have presented some interesting issues regarding sample size and, specifically, the reliability of sampling relatively small plaintiff classes. The district court heard testimony from 43 of the 145 salespeople properly included in the collective action – a percentage (29.66%), the Court noted, that compared favorably to the “sample” accepted in Monroe (17 of 293, or 5.8%).

But the raw percentage of class members included in a “sample” is not a valid proxy for statistical reliability. Rather, the size of the sample needed to achieve statistical significance is a function of the size of the population under study – not the raw percentage sampled. Because the margin of error depends inversely on the square root of the sample size, large populations (such as that of the United States) are often represented by a sample size of 1,000 – representing far less than 1% of the population.

To reliably assess smaller populations (like the total number of Wyndham salespeople in this case), however, a much larger percentage of the population must be sampled. A population of 500, for example, may require a sample of 218 members (44% of the population) to arrive at statistically reliable results (at the conventional 95% confidence level). The size of an appropriate sample also depends on the degree of variability or differences that exist in the population under study. Several organizations provide helpful, easy-to-use calculators and tables online that allow one to determine whether a particular sample is within the realm of reason.

As explained above, the size of the sample was not determinative in this case. (And, admittedly, there is some questionable precedent out there that appears to accept comparing raw percentages of class members “sampled” as a measure of reliability. See Monroe, 860 F.3d at 410 (collecting cases).) But the Court’s discussion of sample size highlights an issue that could arise in future cases.

Third, Wyndham also challenged the “randomness” of plaintiffs’ sample. After an expert pulled a random sample of Wyndham sales employees, plaintiffs supplemented that sample with the testimony of six additional salespeople who were not part of the random sample. Wyndham argued the addition of these “hand-picked” employees undermined the reliability of plaintiffs’ sample.

Ordinarily, selectively adding members to a purportedly “random” sample would undermine the reliability of the evidence, introducing potential bias. But the Court’s response to Wyndham’s argument demonstrates exactly why these types of statistical issues did not trouble the Court in this case.

As the Court explained, Wyndham deposed all six newly-added salespeople, as well as every other salesperson included in the sample. If any salesperson in the original (or supplemented) sample experienced different treatment, Wyndham “could have called [them] to rebut the testifying employees.” But it “failed to do so,” leaving the Court with “no warrant to doubt the reliability of the witnesses.”

The Court’s response makes clear that it based its decision on the uniform testimony of dozens of sales employees who were subject to the same policy and experienced the same injury – not the statistical validity of plaintiffs’ sampling methods. Nonetheless, the Court’s discussion of representative evidence raises some important questions for future cases. If the defendant had effectively rebutted the allegation that all front-line and in-house salespeople were treated the same way, how would the Court determine whether the testimony of individual plaintiffs was truly representative of the group as a whole? What role should statistical evidence play in that inquiry? And in future cases, how will the Court address criticism of sampling methods, and the “conclusions” drawn from questionable samples?