A lawyer bringing a gender discrimination class action against Microsoft Corp. faced tough questions Monday from a federal appeals court panel struggling to understand the process by which the software giant determined its pay and promotions.

Anne Shaver, a partner at San Francisco's Lieff Cabraser Heimann & Bernstein, tried to convince the U.S. Court of Appeals for the Ninth Circuit to reinstate a nationwide class action alleging Microsoft systematically discriminated against more than 8,600 female engineers and information technology professionals, in both pay and promotions. The panelists peppered her with questions about how Microsoft evaluated its employees, and Judge Johnnie Rawlinson of the U.S. Court of Appeals for the Ninth Circuit repeatedly asked her to identify the policy at issue.

"But under Dukes, we have to find what the policy is because the policy has to operate across the board," said Rawlinson, referring to the 2011 U.S. Supreme Court decision Walmart v. Dukes. "I don't feel there is a policy that's been pointed out."

The panel continued to have questions after Lynne Hermle began to argue for Microsoft. Rawlinson even asked her not to use the word "clearly" again.

"Please don't say 'clearly,' because very little is clear to me in this case," she said.

Hermle, a partner in the Menlo Park office of Orrick Herrington & Sutcliffe, who was asking the panel to affirm a federal judge's 2018 decision not to certify the class, insisted that plaintiffs attorneys had shifted their legal theories during the course of the case to identify a policy that did not exist.

"The reason there is so much confusion about it is there is not a single piece of evidence to support it," she told the panel. "There was no evidence this was a common practice."

The case is one of several gender bias cases that targeted the technology industry. Other companies that have faced scrutiny include Google Inc., Oracle Corp., Twitter and Uber Technologies.

In his decision for Microsoft, U.S. District Judge James Robart of the Western District of Washington found there were no uniform policies or job descriptions that tied the proposed class together.

The Ninth Circuit agreed to take up an interlocutory appeal of that decision, which more than 30 civil rights, labor and women's rights groups pushed to reverse in an amicus brief.

After Microsoft filed its opening brief, six organizations provided supporting amicus briefs, including the U.S. Chamber of Commerce and the Washington Legal Foundation. Among other things, they wrote that a class action was not appropriate in the case.

Shaver, on Monday, told the panel, "What's at stake in this case is the ability of employees to challenge employment practices that discriminate against them through a class action."

She acknowledged that Microsoft's employment practices involved some discretion but explained how the company puts employees into "peer groups," based on similar work, then separates them into "pay bands" or "stock levels" to determine their pay. That process, in which women are "locked into" lower pay bands, systematically discriminates based on gender, she said.

"The result is a woman might perform equally or better than her male peer but can't be paid the same as him because he's in a higher pay band," she said.

She also leaned on statistics that showed "significant" disparities in pay and promotions between men and women at Microsoft.

"When the claim is the most senior levels of the company knew and failed to remedy it, aggregate statistics are relevant," she said. "This was a company with very serious problems, and leadership knew about it."

U.S. District Judge Leslie Kobayashi, sitting by distinction from the District of Hawaii, questioned whether Microsoft's most senior executives, who could have been just "rubber stamping" the decisions of the lower level managers, were aware of potential discrimination in the pay practices.

Addressing Hermle, Ninth Circuit Judge Richard Paez asked whether Robart had applied the wrong standard in rejecting the plaintiffs' statistical evidence.

Hermle said he did not.

"What the plaintiffs chose to do was apply the statistics at the EVP level where decisions are clearly not being made," Hermle responded.