Diversity in Data Science
Sourcing, hiring, and growing female talent
Here at Coursera, we are proud to have a data science team that is nearly half female. The problems we’re tackling demand creative approaches and, as the literature consistently shows, diversity unlocks innovation.
Yet only 16% of technical roles at major tech companies are held by women. What lessons have we learned in building our team?
Sourcing diverse talent
Diversity — including gender diversity — demands conscious and consistent efforts to source the right mix of talent. Consider the very top of the recruiting funnel and its three main pipelines: organic applicants, referrals from existing employees, and sourced candidates.
- Organic applicants come in through our job postings. A number of publicly available tools help us check that our job descriptions are inclusive — and pinpoint what to change when we’re missing the mark.
- Referrals come in through existing employees. These can be an incredible source of talent: strong performers tend to refer other strong performers. But people also tend to refer people who look like them. So if there’s a lack of diversity on the team today, leaning too heavily on referrals can exacerbate the disparity. The solution? Start diversity efforts early — the investment will pay off in spades as the team scales.
- Sourced candidates are those we go out and actively recruit. Here, we work to build a diverse network — for example, by hosting inclusive events with groups like Women Who Code, and by reaching out to select passive candidates who might be a strong fit encouraging them to apply.
Removing bias in screening and interviewing
Once the right mix of candidates is in the pipeline, we can bring intentionality to balanced screening and interviewing. For many of our technical roles, we’ve implemented automated screens on HackerRank; relative to having a recruiter or hiring manager as the first point of contact with all applicants, these screens allow us both to cast a wider and more inclusive net, and to do a more objective initial evaluation. At the interview stage, setting up panels of interviewers who are themselves diverse for phone screens and on-site conversations can help ensure fair evaluation — and experiencing the diversity we’ve already cultivated can help candidates assess fit.
For gender diversity in particular, connecting the dots between the technical work and social impact during the interview process can also help attract female technical talent to our teams. While many technical roles do provide a unique opportunity to make an impact in education, health, and other societal challenges, the link is not always obvious in the classroom or even while interviewing. Talking to potential candidates early and often about the why behind the work can make these roles more appealing to top female talent.
Retaining and growing diverse talent
Recruiting is the start of the story, but only truly valuable when all employees are treated as equals in the workplace, and grown and empowered in their new roles. This includes being deeply honest about whether otherwise similar employees have access to the same high-visibility projects and the same transparent feedback (and where they do not, fixing it). More formally, we can validate that compensation and promotion decisions are unbiased. For example, we conduct an analysis of pay parity at every compensation review to ensure fairness across the company, including between men and women in comparable roles.
Many of these actions take conscious thought and daily effort across the company — but the benefits of cultivating strong and diverse talent in data science and other technical roles far outweigh the costs. Every day, I have the privilege of seeing colleagues challenging each other to approach problems from different perspectives, and to try solutions that they might never have dreamed up alone. And that’s exactly what diversity is about — making all of us more thoughtful, creative, and successful as a team.
Originally published at blog.coursera.org on January 31, 2017.