How Analytics Can Solve Silicon Valley’s Diversity Problem

It is clear there is still a problem, and data could be the answer


Google’s decision to fire one of its software engineers, James Damore, over his anti-diversity manifesto has drawn a heartening amount of criticism. Taken alongside Uber CEO Travis Kalanick’s recent resignation amid accusations of sexual harassment and discrimination, it seems we are moving towards a world where tech companies are at least willing to appear decisive in slamming sexism, if even only for PR reasons.

However, heartening though it may be, there is still a distinct lack of female representation in Silicon Valley. According to Google’s most recent demographic report, 69% of its workforce and 80% of its technical staff are male. At Uber, 85% of technical employees are male. And the ratios are little better in other organizations across the industry.

This situation needs to change, not only for the businesses themselves to flourish, but if we are to benefit fully from AI. According to PwC’s 18th Annual Global Survey, 85% of CEOs have said that having a diversified and inclusive workplace population improved their bottom line, with another recent McKinsey Global Institute study confirming that gender-diverse companies perform 15% better than their one-note counterparts. It is even more important for AI. The result of male domination of tech has already led to the development of, for example, voice recognition technologies trained and tested solely by men. As a result, they struggle to understand female voices. This is a problem likely to occur in a wide range of technologies if the female experience continues to be discounted in the development of new AI technologies.

The reasons for the lack of women in tech are complex, and there is no simple solution. Diversity programs, mentorship, quotas, and so forth have all been implemented to varying degrees of success, as have schemes encouraging young women earlier on in their schooling. There is another answer that many are looking at, and it could be one of the very things at threat from the lack of diversity - data analytics.

The rate at which data analytics is being adopted by HR departments has increased dramatically over the past year, as practitioners look to harness the value of the information they hold about their employees to understand the drivers of performance, improve retention rates, and boost workplace productivity. Diversity is one area that many are already benefiting. For example, the Met Police managed to increase diversity by 15% using data analytics. Google too is using algorithms in order to avoid the unconscious bias that can impact hiring decisions.

Unconscious bias is a particular problem hampering tech companies’ diversity efforts, and it is one a number of startups are attempting to solve. Subjective hiring practices can allow bias to be a primary decision influencer. It is human nature to hire someone similar to you. Only 10 years ago, having an African American name made you 33% less likely to get a callback from a company you’ve given your CV to in the US according to a study from the National Bureau of Economic Research. As Business Insider notes, ‘…when we don't have a rigorous, replicable set of criteria from which to evaluate a potential hire's merit, we fall back on our most immediate instrument: ourselves.’ Blendoor, for example, reduces unconscious bias by hiding data that’s not relevant and highlighting data that is. Infor, similarly, builds unique ‘performance profiles’ – the profile of the ideal employee for a position, from which to make hiring decisions. They claim to have boosted employee diversity for clients by as much as 26%.

Data analytics can help in other ways, too. For one, it can identify what is holding women back. For example, one company found that very few women were going for certain promotions. When asked why, they said that they only met 8 out of 10 of the necessary attributes. However, many male applicants were going for the role that had far fewer of the skills needed. This company now looks at transferable skills when writing job postings, rather than mandating specific areas in which those skills were gained.

It is also vital in identifying when an employee may leave. It is one thing recruiting diverse talent, but if your minority talent has a high voluntary turnover rate, you haven't done much to improve the diversity of your workforce. Companies can use analytics to quickly see when a disproportionate number of a certain group of people are leaving the firm. For example, if a number of black people are leaving one department, it may be worth retraining the manager or offering some other kind of support. Predictive analytics can segment employees by demographic to pinpoint signifiers that someone may be thinking of leaving, and HR can implement initiatives to solve these issues as a result.

Finally, it helps to show that there’s a problem in the first place. Data visualization software can break down all the information to provide senior decision makers with a visual narrative to ensure the message really sinks in, and inspire them to do something about it. It is important to remember that analytics alone is not enough to change people’s attitudes. But realizing there is a problem is the first step, and analytics is then a good way of keeping on top of it to ensure that once it is solved, it doesn’t return. 


Read next:

Why Blockchain Hype Must End