If you think women in tech is just a pipeline problem, you haven’t been paying attention

It doesn’t matter how many girls you teach to code if you keep driving adult women out of the tech industry.

Rachel Thomas


July 27, 2015

This article has been translated into Spanish at Matajuegos and into Russian by Dmitry Si.

According to the Harvard Business Review, 41% of women working in tech eventually end up leaving the field (compared to just 17% of men), and I can understand why…

I first learned to code at age 16, and am now in my 30s. I have a math PhD from Duke. I still remember my pride in a “knight’s tour” algorithm that I wrote in C++ in high school; the awesome mind warp of an interpreter that can interpret itself (a Scheme course my first semester of college); my fascination with numerous types of matrix factorizations in C in grad school; and my excitement about relational databases and web scrapers in my first real job.

Over a decade after I first learned to program, I still loved algorithms, but felt alienated and depressed by tech culture. While at a company that was a particularly poor culture fit, I was so unhappy that I hired a career counselor to discuss alternative career paths. Leaving tech would have been devastating, but staying was tough.

Work hard, play hard

I’m not the stereotypical male programmer in his early 20s looking to “work hard, play hard”. I do work hard, but I’d rather wake up early than stay up late, and I was already thinking ahead to when my husband and I would need to coordinate our schedules with daycare drop-offs and pick-ups. Kegerators and ping pong tables don’t appeal to me. I’m not aggressive enough to thrive in a combative work environment. Talking to other female friends working in tech, I know that I’m not alone in my frustrations.

When researcher Kieran Snyder interviewed 716 women who left tech after an average tenure of 7 years, almost all of them said they liked the work itself, but most were unhappy with the work environment. In NSF-funded research, Nadya Fouad surveyed 5,300 women who had earned engineering degrees (of all types) over the last 50 years, and 38% of them were no longer working as engineers. Fouad summarized her findings on why they leave with “It’s the climate, stupid!”

This is a huge, unnecessary, and expensive loss of talent in a field facing a supposed talent shortage. Given that tech is currently one of the major drivers of the US economy, this impacts everyone. Any tech company struggling to hire and retain as many employees as they need should particularly care about addressing this problem.

Your company is NOT a meritocracy and you are NOT “gender-blind”

You don’t know if you’re color-blind without testing either

Nobody wants to think of themselves as being sexist. However, a number of studies have shown that identical job applications or resumes are evaluated differently based on whether they are labeled with a male or female name. When men and women read identical scripts containing entrepreneurial pitches or salary negotiations, they are evaluated differently. Both men and women have been shown to have these biases. These biases occur unconsciously and without intention or malice.

Here is a sampling of just a few of the studies on unconscious gender bias:

Most concerningly, a study from Yale researchers shows that perceiving yourself as objective is actually correlated with showing even more bias. The mere desire to not be biased is not enough to overcome decades of cultural conditioning and can even lend more credence to post-hoc justifications. Acknowledging that you have biases that conflict with your values does not make you a bad person. It’s a natural result of our culture. The important thing is to find ways to eliminate them. Blindly believing your company is a meritocracy not only does not make it so, but will actually make it even harder to address implicit bias.

Bias is typically justified post-hoc. Our initial subconscious impression of the female applicant is negative, and then we find logical reasons to justify it. For instance, in the above study by Yale researchers if the male applicant for police chief had more street smarts and the female applicant had more formal education, evaluators decided that street smarts were the most important trait, and if the names were reversed, evaluators decided that formal education was the most important trait.

Good News and Bad News

The Bad News…

Because of the high attrition rate for women working in tech, teaching more girls and women to code is not enough to solve this problem. Because of the above well-documented differences in how men and women are perceived, training women to negotiate better and be more assertive is also not enough to solve this problem. Female voices are perceived as less logical and less persuasive than male voices. Women are perceived negatively for being too assertive. If tech culture is going to change, everyone needs to change, especially men and most especially leaders.

The professional and emotional costs to women for speaking out about discrimination can be large (in terms of retaliation, being perceived as less employable or difficult to work with, or companies then seeking to portray them as poor performers). I know a number of female software engineers who will privately share stories of sexism with trusted friends that we are not willing to share publicly because of the risk. This is why it is important to proactively address this issue. There is more than enough published research and personal stories from those who have chosen to publicly share to confirm that this is a widespread issue in the tech industry.

…and the Good News

Change is possible. Although these are schools and not tech companies, Harvey Mudd and Harvard Business School provide inspiring case studies. Strong leaders at both schools enacted sweeping changes to address previously male-centric cultures. Harvey Mudd has raised the percentage of computer science majors that are women to 40% (the national average is 18%). The top 5% of Harvard Business School graduates rose from being approximately 20% women to closer to 40% and the GPA gap between men and women closed, all within one year of making a number of comprehensive, structural changes.

So What Can We Do About It?

These recommendations on what companies could do to improve their cultures are based on a mix of research and personal experience. My goal is to have a positive focus, and I would love it if you walked away with at least one concrete goal for making constructive change at your company.

Train managers

It is very common at tech start-ups to promote talented engineers to management without providing them with any management training or oversight, particularly at rapidly growing companies where existing leadership is stretched thin. These new managers are often not aware of any of the research on motivation, human psychology, or bias. Untrained, unsupervised managers cause more harm to women than men, although regardless, all employees would benefit from new managers receiving training, mentorship, and supervision.

Formalize hiring and promotion criteria

In the Yale study mentioned above regarding applicants for police chief, getting participants to formalize their hiring criteria before they looked at applications (i.e. deciding if formal education or street smarts was more important) reduced bias. I was once on a team where the hiring criteria were amorphous and where the manager frequently overrode majority votes by the team because of “gut feeling”. It seemed like unconscious bias played a large role in decisions, but because of our haphazard approach to hiring, there was no way of truly knowing.

Leaders, speak up and act in concrete ways

Leadership sets the values and culture for a company, so the onus is on them to make it clear that they value diversity. Younger engineers and managers will follow their perceptions of what executives value. In the cases of positive change at Harvey Mudd and Harvard Business School, leadership at the top was spearheading these initiatives. Intel is going to begin tying executives’ compensation to whether they achieve diversity goals on their teams. As Kelly Shuster, director for the Denver chapter of Women Who Code has pointed out, leaders have to get rid of employees who engage in sexist or racist behavior. Otherwise, the company is at risk of losing talented employees, and is sending a message to all employees that discrimination is okay.

Don’t rely on self-nominations or self-evaluations

There is a well-documented confidence gap between men and women. Don’t rely on people nominating themselves for promotions or to get the most interesting projects, since women are less likely to put themselves forward. Google relies on employees nominating themselves for promotions and data revealed that women were much less likely to do so (and thus much less likely to receive promotions). When senior women began hosting workshops encouraging women to nominate themselves, the number of women at Google receiving promotions increased. Groups are more likely to pick male leaders because of their over-confidence, compared to more qualified women who are less confident. Don’t rely heavily on self-evaluations in performance scoring. Women perceive their abilities as being worse than they are, whereas men have an inflated sense of their abilities.

Formally audit employee data

Confirm that men and women with the same qualifications are earning the same amount and that they are receiving promotions and raises at similar rates (and if not, explore why). Make sure that gendered criticism (such as calling a woman strident or abrasive) is not used in performance reviews. The trend of tech companies releasing their diversity statistics is a good one, but given the high industry attrition rate for women, they should also start releasing their retention rates broken down by gender. I would like to see companies release statistics on the rates at which women are given promotions or raises compared to men, and how performance evaluation scores compare between men and women. By publicly sharing data, companies can hold themselves accountable and can track changes over time.

Don’t emphasize face time

A culture that rewards facetime and encourages people to regularly stay late or eat dinner at the office puts employees with families at a disadvantage (particularly mothers), and research shows that working excess hours does not actually improve productivity in the long-term since workers begin to experience burn out after just a few weeks. Furthermore, when employees burn out and quit, the cost of recruiting and hiring a new employee is typically 20% of the annual salary for that position.

Create a collaborative environment

Stanford research studies document that women are more likely to dislike competitive environments compared to men and are more likely to select out of them, regardless of their ability. Given that women are perceived negatively for being too assertive, it is tougher for women to succeed in a highly aggressive environment as well. Men who speak up more than their peers are rewarded with 10% higher ratings, whereas women who speak up more are punished with 14% lower ratings. Creating a competitive culture where people must fight for their ideas makes it much tougher for women to succeed.

Offer maternity leave

Over 10% of the 716 women who left tech in Kieran Snyder’s research left because of inadequate maternity leave. Several were pressured to return from leave early or to be on call while on leave. These women did not want to be stay-at-home-parents, they just wanted to recover after giving birth. Just as you would not pressure someone to return to work without recovery time after a major surgery, women need time to physically heal after delivering a baby. When Google increased paid maternity leave from 12 weeks to 18 weeks, the number of new moms who quit Google dropped by 50%.

Some final thoughts…

A note on racial bias

There is a huge amount of research on unconscious racial bias, and tech companies need to address this issue. As Nichole Sanchez, VP of Social Impact at GitHub, describes, calls for diversity are often solely about adding more white women, which is deeply problematic. Racial bias adds another intersectional dimension to the discrimination that women of color experience. In interviews with 60 women of color who work in STEM research, 100% of them had experienced discrimination, and the particular negative stereotypes they faced differed depending on their race. A resume with a traditionally African-American sounding name is less likely to be called for an interview than the same resume with a traditionally white sounding name. I do not have the personal experience to speak about this topic and instead encourage you to read these blog posts and articles by and about tech workers of color on the challenges they’ve faced: Erica Joy (Slack engineer, former Google engineer), Justin Edmund (designer, Pinterest’s 7th employee), Aston Motes (Engineer, Dropbox’s 1st employee), and Angelica Coleman (developer advocate at Zendesk, formerly at Dropbox).


I’m currently teaching software development at all-women Hackbright Academy, a job that I love and that suits me perfectly. I want all women to have the opportunity (and I mean truly have the opportunity, without implicit or explicit discrimination) to learn how to program — knowing software development provides so many career and financial possibilities; it’s intellectually rewarding and fun; and being a creator is deeply satisfying. Although I know many women with frustrating experiences of sexism, I also know women who have found companies where they’re happily thriving. I’m glad for the attention tech’s diversity problem has been receiving and I am hopeful about continued change.

Thanks for review, edits, and discussion to: Jeremy Howard and Angie Chang.

I do more research and further develop the ideas in this post in my later posts: on how showy, shallow diversity strategies make things worse; on bullshit diversity initiatives and some better ideas; and the research on how women are leaving tech because they can’t advance in their careers.

I look forward to reading your responses. Create a free GitHub account to comment below.