With New Funding, Textio Trains Machines to Improve HR Writing

Textio, a Seattle startup using machine intelligence and natural language processing to help people improve documents like job listings and recruiting e-mails as they’re writing them, has raised $8 million and plans to triple its workforce in the coming year.

Emergence Capital led the $8 million Series A, with existing investors Cowboy Ventures, Bloomberg Beta, and Upside Partnership also participating. Textio raised a $1.5 million seed round earlier in 2015.

Kieran Snyder and Jensen Harris, former Microsoft employees, co-founded Textio in 2014. Their technology—which analyzes things like word choice and document structure and length, and provides suggestions for improvements in real time—has struck a chord with companies competing for talent and working to attract more diverse job applicants.

Twitter called out Textio as a key tool for doing just that when it released company-wide diversity goals last summer. Square demonstrated Textio at The 3% Conference. And although Textio didn’t attend the HR Technology Conference and Exposition, industry analysts and customers “told our story for us,” Snyder says.

“We had a lot of people that were willing to share their stories and the impact that we’re having on their hiring,” she says. “I think that’s been the center of success for us this year.”

Snyder and Harris.

Snyder and Harris.

Snyder, who earned a PhD in linguistics and cognitive science at University of Pennsylvania, and Jensen, a user experience leader credited with the Ribbon user interface that became a standard in Microsoft Office, were mulling startup ideas as they moved on from Microsoft last year. They experimented with things like predicting a crowd-funding project’s success based on an analysis of the language used to describe it. At the same time, Snyder was writing about the relationships between gender, technology, and language. HR leaders reached out to her after one of her articles went viral, wanting to understand her methodology.

Those conversations illuminated the problems of hiring and recruiting, and the potential solution for predictive text technology.

Fast forward to July of this year, when Textio released its first commercial product, Textio Talent. It has helped customers fill job openings nearly 20 percent faster and increased applicants from underrepresented groups by 12 to 15 percent, Snyder says.

Textio tracks language and other document attributes, looking for patterns that correlate with increased applications and applications that lead to screening interviews, as well as language that is more likely to attract male and female applicants. “Textio recognizes 50,000 distinct phrases that change the number, quality, and diversity of candidates who apply,” Snyder wrote in a November blog post. “The list of effective phrases is changing constantly as the market shifts.”

For example, name-checking “big data” in an engineering job listing just two years ago attracted significantly more applicants. Today, however, job listings that include that now-cliché phrase perform an average of 30 percent worse than those that leave it out. Meanwhile, the phrase “artificial intelligence” has come to the fore in the strongest-performing tech job listings during the last six months.

The company’s latest add-on, now in beta, helps recruiters improve the language in e-mails they send to candidates. Textio developed the e-mail application after noticing that 10 percent of the content customers put through Textio Talent was e-mail.

The core technology for analyzing and scoring a document can be applied to job listings and e-mails, as well as other documents. The challenge for Textio each time it adds a new document type is to gather enough examples, and data on their outcomes, to train its algorithms to recognize a winner.

“The technical work for us when we go into a new vertical or a new document type is a lot about making sure we have robust enough data to substantiate a good set of predictions,” Snyder says.

Textio trains its algorithms using a mix of customer-provided data, and recruiting e-mails and job postings it has collected on its own. That’s getting easier as Textio amasses more customers who see the value they’re getting in one area (job listings) and want to use it on adjacent documents (recruiting mails).

While most of Textio’s beta customers were technology companies, the industry mix has shifted as usage has grown. Tech now represents about 40 percent of Textio’s customers, with financial services firms making up another 25 percent. A wide range of companies, including Major League Baseball and the Monterey Bay Aquarium, make up the remainder.

Technology and financial services companies “are predisposed to quantitative approaches to problems in general and so Textio is a really natural fit,” Snyder says. “And of course technology has put a big spotlight on inclusion and diversity considerations, which is helpful for Textio.”

With the Series A funding, Textio will continue to refine and expand its offerings for HR professionals and recruiters. Snyder says there is definitely interest in other document types or industries, but for now the 10-person company—which plans to hire about 20 more employees in the next year—is staying focused on documents related to recruiting, hiring, and retention.

In the growing landscape of machine intelligence startups, it helps to have a tight industry focus and the ability to solve a problem businesses have today.

Shivon Zilis, a partner and founding member at Bloomberg Beta (one of Textio’s investors), writes of the evolution of machine intelligence companies in the last year: “startups [are] shifting away from building broad technology platforms to focusing on solving specific business problems.”

She expects that to continue in 2016.

“Machine intelligence is even more of a story than last year, in large companies as well as startups,” Zilis writes at O’Reilly Radar. “In the next year, the practical side of these technologies will flourish. Most new entrants will avoid generic technology solutions, and instead have a specific business purpose to which to put machine intelligence.”

Trending on Xconomy