Using AI to Predict the Performance of Text

Insights from FirstMark’s Data Driven series, a monthly event focused on the business aspects of the data revolution: startup opportunities, new products, business models, funding trends, emerging players, and ways for existing startups to leverage their data.

Imagine knowing how well your words will be received before you actually say them. Textio believes that concept is the future of writing.

The company’s first product — Textio Talent — aims to help recruiters perfect their job postings. Using machine learning, Textio analyzes job text and outcomes data using listings from more than 10,000 companies. They use patterns to predict the performance of job postings and try to give writers the opportunity to fix lesser-performing language before it’s ever published.

Textio has tagged more than 15 million job listings to date, collecting data such as the number of people who apply for jobs, how long it takes to fill a position and the quality and demographic mix of candidates. In a talk at Data Driven NYC, Textio Founder & CEO Kieran Snyder offered some of the interesting takeaways from from analyzing millions of job posts.



The wrong words are costly.
More than 50% of recruiters’ time is spent sourcing candidates and 25% of recruiting spend goes to pay for job ads. Textio’s research suggests that candidates scan job listings for less than six seconds and a third will walk away without interest. Recruiters also write hundreds of emails to potential candidates every week and around 75% go unanswered. This data shows just how challenging it can be to capture a candidate’s attention and underscores the importance of creating documents optimized for the best results. The longer that low-performing post is out in the wild, the more money it’s costing the company.

Beware of buzzword backlash. Around 18 months ago, using the term “Big Data” in a job listing actually attracted more applicants and more qualified applicants. But the term gained popularity across various industries and it stopped being a differentiator. Then the term gots so popular that it has become cliche — actually repelling candidates. Similarly, the term “synergy” worked in job postings around five years ago. Now, the use of the term is almost taken as a joke. So, be careful about loading up a job post with industry jargon.

Find the right balance with bullets. There’s a sweet spot for bulleted formatting — you want a third of copy bulleted. Textio has also learned that the ratio of bulleted copy can impact the type of people who will apply for the job. Postings with more than 50% bulleted copy will attract fewer women. Postings with less than 25% bulleted content will attract fewer men.

Cover your culture. The language of a job description needs to match the culture of the company. If it’s a job posting for Expedia, travel vocabulary is important, as travel may be a personal passion for potential employees who would go on to be passionate about their work. Recruiters are not just trying to find the best fit for the role, they need to find someone who is the best fit for the company. Making sure that the true culture of the workplace is reflected in the language is important in attracting ideal candidates.

The benefits of better words. Snyder said Textio users get results. Qualified applicants increases by 24%, time-to-hire drops by 17%, and the number of underrepresented applicants increases by 12%. In short, choosing better language in job descriptions results in filling positions faster with more qualified candidates.

To hear more about Textio’s mission to change the way people write, see the full talk from Data Driven NYC in the video above. To hear more great talks from FirstMark’s Driven events, visit our content library and subscribe on SoundCloud and iTunes.