Diversity reports don't always tell the full story of what representation looks like at tech companies.
The earliest versions of tech industry diversity reports created a binary narrative about diversity and inclusion , according to Bo Young Lee, Uber's chief diversity officer. The reports, which tech companies began consistently releasing in 2014, communicated that "diversity is about gender representation, and it's about race and ethnicity, and it's about nothing else," Lee told Protocol.
"And these diversity reports kind of boiled everything down to these two sets of identities," she said. "When in fact, just knowing the representation of women or just knowing the representation of underrepresented people of color, you don't really actually get a sense of how subgroups within there are thriving."
Reports in recent years have begun to detail additional data points, including, for example, information about intersectionality, which considers overlapping identities that may contribute to discrimination or some other form of disadvantage. But the industry still lacks a standard, comprehensive method of reporting and presenting employee demographic data.
Data pet peeves
A good diversity report needs rich and clean data, according to Bernard Coleman , the chief diversity and engagement officer at human resources startup Gusto. That means companies need to be equipped with a reporting system that makes it easy to collect that data and a strategy to encourage employees to self-identify across a variety of categories.
"Maybe there's a lot of data maintenance to make sure that you truly understand your folks," Coleman said. "We've done a self-ID campaign to really understand who are you and what do you need to be successful. I think that's really important. A lot of times, I've seen folks just don't have clean data."
Lee agreed with Coleman's sentiment. She said it's "shocking" that many companies "don't actually have very good validated workforce data."
Sometimes, Lee said, the ambition to report as much data as possible can undermine the integrity of the report. Many companies, for example, have begun to report sexual orientation and gender identity information.
"But I would guarantee you almost no company probably has a statistically significant amount of information to say that the sexual orientation and gender identity [information] is an accurate reflection of their workforce," Lee said.
In addition to the standard race and gender data, Uber collects self-reported employee data around sexual orientation, gender identity, veteran status, caregiver status, disability and socioeconomic status during childhood. But it has yet to report that additional data because Lee wants at least 80% of Uber's workforce to respond to those questions.
"Otherwise, we don't really know if it's representative," she said.
Companies also vary in their methods of presenting data. At Twitter, the company previously reported 100% of workforce data but in 2017 began reporting undisclosed responses. That change resulted in Twitter appearing significantly more diverse than it once did. In 2016, for example, Twitter was 57% white. In 2017, when Twitter reported the demographics of only 80% of its workforce, Twitter was 44.3% white.
In 2018, Square began grouping Black, Latinx, Native American, Pacific Islander and employees of two or more races under an "underrepresented minority" umbrella when detailing the demographics of its tech, business and leadership teams. At Intel, the company grouped white and Asian-American employees together in a "majority population" bucket from 2017 through 2019.
"As an Asian American, I think it is very insulting," Lee said.
The limitations of the EEO-1
A diversity reporting standard does exist, but both Lee and Coleman say it's subpar. The Equal Employment Opportunity Commission requires companies with 100 or more employees to annually report demographic data via an EEO-1 filing. But this standard has its flaws.
One limitation with EEO-1 reports, Lee said, is that it's only U.S.-based data. Additionally, the EEOC asks companies to input data in a way that is "fundamentally different from the way corporations actually do it," she said.
EEO-1 reports require companies to report the raw numbers across 10 different types of roles, as well as male, female and six races. Lee argues that it doesn't allow for a deeper analysis of a company's workforce.
"So you're seeing pure gender-based data, you're seeing pure race- [and] ethnicity-based data [in the EEO-1]," she said. "So again, there's a limitation there from a reporting compliance perspective."
Coleman said the EEO-1 does serve a purpose but recognizes that there is room for a better standard. He also thinks a diversity report standard would make it easier for more companies to participate.
"A lot of folks I think sit on the sidelines because they don't want to get it wrong," Coleman said. "And I think most organizations don't want to misrepresent what their data is. [...] That would take a lot of that guesswork out, if we all had a standard we could go by that was deeper and richer than the EEO. Because the EEO is kind of limiting and I would argue decades behind."
