How Airbnb’s Diversity Strategy Increased Women in Data Science by 100% 

Across the technology industry it is a shared value that workforce diversity is a good thing – morally good and good for business. Yet most companies remain tenaciously male.

Airbnb was one of those companies, despite its recognition that gender uniformity could limit its ability to relate to its two-sided market (rentees and renters) and lower its conversion rates of convincing property owners to rent houses and apartments – and travelers to rent those properties.

Two years ago, Airbnb decided to do more than just lament its lack of diversity and actually do something about it, starting with the data science group.

Airbnb, of course, is widely known for its adroit use of data in support of the 30 million guests, as of 2015, who use its services per year (annual revenues for the privately held company are estimated at nearly $1 billion). But in spite of its explosive growth, according to Elena Grewal, a data science manager at Airbnb, the company was convinced it needed to add a stronger women’s perspective to the work it does.

Elena Grewal of Airbnb

“Airbnb is a global community, our mission is to belong everywhere,” Grewal said at the Structure Data conference held last week in San Francisco. “It’s really important to have a team that reflects our community, to be able to gain the right kind of insights from our data, to have people with a variety of opinions and perspectives in the room drawing insights from the data. We knew this was something we needed to focus on.”

Grewal said the recognition that something had to be done took place in 2014, when the company realized that in the previous year only 10 percent of new hires had been women.

“This was really a wake-up call that we couldn’t continue the way we were,” she said. “The women on our team were quickly becoming outnumbered and we knew that having a diverse team was important to us. We weren’t being intentional in our hiring to create as diverse a team as we needed.”

The first step taken to implement Airbnb's diversity strategy by the data science group was to – analyze data. They examined the job application conversion funnel and were surprised to find that fully 30 percent of job applicants were women, yet women comprised only 15 percent of the group. So names were removed during the initial stage of filtering applications and resumes.

But it’s what comes next that probably has had the biggest impact on improving diversity at Airbnb.
Grewal explained that one of the first things candidates are tasked with is a take-home data test, an analytical challenge using Airbnb data. The resulting grades are used as the basis of selecting candidates for the next round in the hiring process.

“We noticed the grading rubric we used was a little bit fuzzy,” Grewal said. “We would look at their analysis, look at their data visualizations and see if it’s good, and then move the candidate along, but it wasn’t very clear or objective.”
So the hiring group made the expectations and grading criteria clearer. “Did this person remove the duplicates in the data? Did they find this answer correctly? It was a pretty powerful approach for making this process more uniform and clear for graders.”

The objective was to eliminate the likelihood of unconscious bias by streamlining the grading process and leaving it less to the discretion of the graders which candidates would make it to the next step.

Those who earn top scores on the data challenge are then invited to come on site at Airbnb to work with data scientists for a day and then present their findings to a team of interviewers. Here again, Grewal’s group looked at the composition of this group and the experience it offers the candidates.

“We realized, because our data science team was already quite skewed toward many more men, that the audience was often all men,” Grewal said, “so a candidate might look at the audience of all men and be a little bit intimated.”
After initially deciding that at least one woman would be present during the candidate presentations, they then took the step of having at least 50 percent women there.

Another piece of the strategy: provide more support to candidates throughout their day at Airbnb, including assigning a “buddy” to have breakfast with each candidate, stay with them and answer questions and be there during the presentations at the end of the day.

The results are impressive. By the end of 2015, the data science group had doubled the percentage of women, from 15 to 30 percent. And Grewal said the group’s methodology is beginning to be adopted across Airbnb in its effort to increase underrepresented minorities.

“It’s actually something we’re starting to scale now,” she said, ”so we hired a data scientist to work with the recruiting team to build out the data pipelines for the entire company, which is really exciting. This kind of analysis is available to anyone at the company, not just the data scientists. Many of the processes that we changed to improve the number of women we believe are transferable.”

Grewal said the group’s efforts appear to extend beyond female diversity to also include greater female job satisfaction at the company, as reflected in a quarterly employee survey, in which workers are asked questions like: Would you recommend Airbnb to a friend as a place to work? Do you see yourself working here in a year? Do you feel you belong at Airbnb?

“It’s hard necessarily to claim cause and effect, but one thing we have found is our team actually has some of the highest satisfactions scores in the entire company,” she said. “One hundred percent of the women said they see themselves working at Airbnb a year from now. So those are promising things to us. In creating a diverse team we have created a culture where people feel they belong on the team and are happy and proud to be part of the team.”

As of today's date of publication, a visit to the data science section of Airbnb's career page on its web site shows that eight positions are open. It also shows two articles written by members of the group - one of them, "Designing Machine Learning Models: A Tale of Precision and Recall," by Ariana Radianto.