In 2014, linguistics Ph.D. Kieran Snyder, who had spent much of her career on quantitative approaches to language, was talking to Jensen Harris, a former user-experience exec at Microsoft, when the two arrived at the same conclusion.
“We had this idea that maybe predictive technologies — natural language processing, machine learning, things I had been doing my whole career — were now at the point where we could use them to predict how a document was going to be received,” says Snyder.
The two tested their idea by building a Kickstarter predictor. It worked. Turns out the quality of a business idea had no bearing on the success of the Kickstarter campaign, Snyder reports. “But where you use your fonts, where you place your headers, how many verbs you have — super predictive,” she notes.
Snyder started to hear from human resource executives who wanted help with the language they were using in hiring and recruiting. Snyder figured that her work on predictive analysis of texts could help and she launched a business called Textio, which uses a machine-learning engine to tailor job postings so companies get more candidates who are better suited to job openings.
When a company subscribes to Textio, it provides job post language as well as outcomes — information about how the job posts performed in the past with real people. “Textio can compare your language to other posts from the past where outcomes are known.”
For example, Textio found that postings mentioning Expedia as a competitor to Airbnb delivered better results than identifying Expedia as an online travel agency.
Launched in October 2014, Textio has attracted $9.5 million in venture capital and has grown from a staff of 10 a year ago to 27 today.