Textio CEO Kieran
Snyder on the
Importance of Language

• • • • • • • • • • • • • • • •

The augmented writing app co-founder discusses what language has to do with diversity and inclusion, and what companies can do to improve.

• • • • • • • • • • • • • • • •

By Michelle V. Rafter

E

xecutives have talked for years about making their organizations more diverse and inclusive. Despite all the talk, “they mostly haven’t,” said Kieran Snyder, co-founder and CEO of Textio, the augmented-writing software.

Just look at the data. Two-thirds of companies have diversity hiring strategies, and 60 percent of Fortune 500 companies have a chief diversity officer. But the rest of the stats don’t match up to the perceived commitment to diversity; only 3 percent of the Fortune 500 disclose full diversity data and only 20 percent share such basics as employee gender or ethnicity, according to Fortune.

It turns out when you have more people from diverse backgrounds in your pipeline, you hire more of them.”

— Kieran Snyder, CEO, Textio

Companies could be reluctant to share because of underwhelming results. In the technology industry alone, despite big spending on diversity initiatives, companies still employ a larger-than-average share of whites, Asian Americans and men, and a smaller share of African-Americans, Hispanics and women, according to the Equal Employment Opportunity Commission. At Apple, underrepresented minorities grew from 19 percent of the workforce in 2014 to 23 percent in 2017, according to the company’s latest diversity report. Although it’s less than a quarter of total employees, the percentage makes Apple the most diverse of 23 major Silicon Valley tech companies that publicly disclose their diversity and inclusion statistics, according to the Center for Investigative Reporting.

Snyder, who has a doctorate in linguistics and cognitive science and previously worked at Microsoft and Amazon, started Textio in 2014 to apply data to recruiting to help companies do better.

The Textio app uses machine learning-based data analytics to parse language in job ads and recruiter communications to uncover patterns that could cause some people not to apply. Snyder demonstrated this in a December 2017 blog post that shared how the phrases 10 major tech companies use in job ads could be indicative of their culture — and turn off some job candidates — including “fast-paced environment” at Amazon and “whatever it takes” at Uber. Textio clients such as Atlassian, Suncor Energy Inc. and Johnson & Johnson have used the service to widen their respective candidate pools, leading to more inclusive hiring.

In an interview with Talent Economy, Snyder talked about why diversity and inclusion efforts fail, why language matters and how something as seemingly inconsequential as a list of bullet points could stop a qualified candidate from applying for a job. Edited excerpts follow.

• • • • • • • • • • • • • • • •

TALENT ECONOMY: Why aren’t efforts to improve diversity and inclusion working?

KIERAN SNYDER: We talk to a lot of heads of diversity and inclusion, and people say, “We’re trying everything, we care, our CEO is invested, but nothing is working.” The question I ask is, “What are you trying that’s different than what you tried a year ago?” The answer is often crickets.

TE: What should they be doing?

SNYDER: Companies don’t experiment a lot, especially relative to the tech industry. Our bread and butter is experimentation. What technology have you tried? Where are you sourcing candidates that you weren’t previously sourcing from? Are you tapping existing employees to tap into their networks? I’m surprised by how often the answer is, “We hadn’t thought about that.” There are exceptions, but in general it hasn’t been the case.

TE: What are some companies that are experimenting?

SNYDER: Atlassian is an interesting example. The woman who leads diversity and inclusion has a social sciences background. She tries stuff. She implemented unconscious bias training and it didn’t work so she unimplemented it. She added Textio to the entire recruiting website. She experimented, measured and threw out what didn’t work. The company went from 10 percent women in their technical graduates engineering cohort to 57 percent in two years. They brought up the number of black and Latino technical interns to 33 percent. What I see in her is an unusual willingness to experiment and be accountable to the data and try interventions that are intentionally designed.

TE: What else can companies do?

SNYDER: In general, heads of diversity and inclusion have difficult jobs. They’re under-resourced. They don’t have money to back up what they want to do.

TE: For a long time, companies have hired for culture fit. Does that exacerbate the problem?

SNYDER: A lot of heads of diversity and inclusion would agree with that assessment. It doesn’t mean their partners in talent acquisition or hiring managers are on the same page. I’ve worked in tech for a long time, and culture fit as a component of hiring has come up on every team I’ve been on. It takes time to change that conversation. It takes someone in the room when you’re on an interview debrief saying, “We’re hiring for values alignment,” not culture fit.

TE: Tech companies’ efforts to improve diversity and inclusion get a lot of attention. How do other industries compare?

SNYDER: Oil and energy is a big, growing segment of the economy with similar challenges. Suncor, when they got their job descriptions to be gender neutral, doubled the number of jobs that women were applying for. A U.K.-based grocery chain called Co-op that is a client was in a similar situation. They have reasonable gender balance but not in tech roles. After a quarter of writing job posts in a way that was more mindful of bias tone, they saw 27 percent more women applying. We talk about Johnson & Johnson a lot. In 2017, after they started using Textio, they had 90,000 more women apply for roles, which represented a 9 percent increase, mainly focused on science, research and tech roles. They saw a comparable bump in hires, which is what you would hope.

TE: If language is so important, how can companies change the language they use? Are there certain terms to avoid?

SNYDER: It’s important to understand when bias is unconscious it’s not something you can have an easy theory about fixing. It’s not just changing three things and being good to go. Here’s an example. Say you’re looking to fill a technical manager role. It turns out when a job description includes a description for someone to manage a team you attract statistically more people who identify as men v. women. When you ask for people to develop a team, you attract more people who identify as women v. men. When you ask for someone to lead a team, you attract a mix. And every manager would say they do all three: manage, develop and lead.

TE: What else can affect whether or not someone applies for a job?

SNYDER: It’s not just about the words. Sometimes it’s about how you format them on the page. It turns out that the job posts that perform best, where you get the highest percentage of qualified people from all genders who apply, have one-third of the qualifications formatted as a bulleted list. If you go above half bulleted qualifications, you see a significant drop-off in women applying, and if you go below 25 percent you see a drop off in the number of men applying. There’s nothing discriminatory about writing a job description in bullets, but statistically it changes who applies for a job.

TE: What are other steps that companies can take?

SNYDER: One of the biggest is being mindful of where you source candidates. At Textio, around August 2017, we started to ask our own hiring teams to try to have 50 percent of our post-résumé review phone-screening candidates be with people from some underrepresented group in the industry for any given role. We were pretty loose about it. We asked them to make their best effort in order to see what would happen. This quarter, about three-quarters of our hires have been women or people from some other underrepresented group. We didn’t change anything else about our interview process. We didn’t not interview white men who were qualified; we just asked people to do their best to find a more representative group of people. It turns out when you have more people from diverse backgrounds in your pipeline, you hire more of them. When we talk to customers, we ask where they are sourcing. We love people who go back to school, change careers, active volunteers who pick up new skills. It shows intent to learn, and they tend to be good teammates and employees. But if you’re recruiting from the same six schools or candidate profiles you’re not likely to change your candidate pool.

TE: What’s Textio’s employee makeup?

SNYDER: In the last nine months, we’ve gone from 45 or 50 people to 75. Today, we’re about 50/50 men and women and a small population of people who identify as nonbinary. Our exec team is 60 percent women, as is our board. We’re about 35 percent non-white. About three-quarters of their hires started out in the Seattle area. Twenty percent to 25 percent of employees who are people of color relocated to be part of Textio, most frequently from California.

TE: How do you address skeptics who say artificial intelligence could also introduce unfair biases into results?

SNYDER: AI is a monolith that means many things to many people. We talk more about augmented writing because it’s what we do. In our case, it’s important that when the platform gives people feedback there’s no opinion. It’s based on real data about who has responded to which jobs with which language patterns in the past. For anyone using machine learning, it begins with having the right dataset.

TE: What’s next for Textio?

SNYDER: Our road map over the next several months is attaching Textio to additional kinds of writing where we think it can have an impact. We have customers using Textio to look for patterns in language that lead to more people of color applying, and they’ve seen a 10 to 15 percent gain in applicants of color over one-and-a-half years. Other customers are using it to screen out biases against age and veteran status.

Michelle V. Rafter is a business journalist in Portland, Oregon, reporting on workforce and technology for Talent Economy and other publications. To comment, email editor@talenteconomy.io.