HRExaminer Radio

HRExaminer Radio is a weekly show devoted to Recruiting and Recruiting Technology airing live on Friday’s at 11AM Pacific

HRExaminer Radio

Guest: Kieran Snyder, Textio Co-Founder and CEO
Episode: 101
Air Date: June 26, 2015


Audio MP3


Kieran Snyder holds a PhD in linguistics and has held product and design leadership roles at Microsoft and Amazon. She has authored several studies on language, technology, and document bias for publication in Fortune, the Washington Post, and Slate.

Most recently, Kieran built a multifunctional team in analytics, program management, and design for Amazon’s advertising organization. In her prior product leadership roles at Microsoft, Kieran created a linguistic services platform for developers, introducing new language detection, spell-checking, and other natural language processing capabilities to Windows developers for the first time. She also led a cross-company engineering effort for the native integration of Bing into Windows search. In her time at the company, Kieran was involved with search and natural language projects across several Microsoft products, including Windows, Bing, Office, and Visual Studio.

Kieran earned her doctorate in linguistics and cognitive science from the University of Pennsylvania and has published original research on gender bias in performance reviews and conversational interruptions in the workplace over the last year. She participates actively in Seattle-based STEM education initiatives and women in technology advocacy groups.

Website LinkedIn Twitter Facebook



Begin transcript

John Sumser:            Good morning, and welcome to the HR Examiner Radio show. I’m your host John Sumser. Today we are coming to you again from beautiful Occidental, California where the roses are in bloom, and people are remembering that this is where technology got its liftoff in the great state of California. Today, we are going to be talking with Kieran Snyder, who is an amazing entrepreneur from Seattle. She runs a company called Textio, and I’m going to let her tell you all about it. Good morning Kieran. How are you?

Kieran Snyder:          Good morning, it’s great to talk to you this morning.

John Sumser:            Yeah, thanks for being here. Why don’t you introduce yourself to the crowd?

Kieran Snyder:          Sure, so I’m Kieran. I’m the CEO of Textio. I am a PhD in linguistics and spent lots of years trying to hire big companies. I have always been obsessed with language. In fact, my very first most lucrative job was hustling Scrabble, trying to get people to play me in Scrabble for money. Now, we’re here doing Textio and really helping people optimize their job listings so they hire lots of great people.

John Sumser:            Is Textio and Scrabble similar? What’s the common thread there, words?

Kieran Snyder:          The common thread is words. I have always been obsessed with language and words. I’m a writer. I am a linguist, and for one time I was a Scrabble hustler. If it’s encoded with language and words then I’m probably into it.

John Sumser:            That’s amazing, so how did you figure out to make the leap from linguistics to something that’s more technical than linguistics? That’s not a move that everybody makes.

Kieran Snyder:          Yeah, you know linguistics is really interesting. It’s a pretty broad field. There are parts of it that are so tied to the humanities and literature, and the study of foreign language. There are parts of it that are deeply connected to computer science and math. When I was coming out of grad school the very first job that I got was a program manager job at Microsoft. Where I was building linguistic services for developers, so things like language detection and spell checkers. You know the kinds of things where software is encoding linguistics in a way that helps the user. That was kind of my en route to technology. It wasn’t my original plan, but I’ve never looked back.

John Sumser:            Still, the technology components of linguistics are a stretch, did you pick up technology skills or do technology classwork, or are you a sort of a self trained technologist?

Kieran Snyder:          I mean I’ve been coding since I was four years old. Right, so I have always been a hacker. My proudest science fair project ever was a supernova simulator I built in Basic when I was in seventh grade. I have always been in and out of technology. I didn’t major in computer science in college. In grad school I took a bunch, but I’ve always been kind of on the side of linguistics that is technology facing. I still didn’t plan to go into tech as a profession. It just kind of worked out that way when the exact right job opportunity came along for me. I’ve always been tech minded.

John Sumser:            Sometime, maybe later on in the conversation, we should talk a little bit about women in tech. Because there are not hoards of women who have a story like yours, and it’s hurting the tech business I think.

Kieran Snyder:          Yeah, we’re still pretty underrepresented.

John Sumser:            Yeah, to put it mildly. You’re the CEO of a startup. What does that mean? What does your day look like?

Kieran Snyder:          Yeah, so my co-founder Jensen Harris and I began Textio about eight months ago. It’s really been quite a journey. We got into it after long careers in corporate technology. Both of us had had times at Microsoft or Amazon. We wanted to build something fundamental and useful that kind of combines linguistics and machine learning with a great user experience. There’s lots we couldn’t have prepared for. You know when you have years of working at a larger company where lots of things are taken care of for you, there’s lots you don’t know. Being the CEO of a tech startup means that sometimes you’re the CEO, and sometimes you’re doing business development, and sometimes you’re doing sales, and sometimes you’re coding. You know, sometimes you’re the benefits administrator for the company, or the facilities administrator. Right, when you’re a small shop you do basically everything. A lot of your challenge as a startup CEO is to make sure that the surface area you put out into the world doesn’t reflect how small your resourcing is. That you actually have to project out real credibility as a company.

John Sumser:            You’ve managed to land in an interesting place. You must have amazing mentors. The insight that you’re bringing to this job is a level of sophistication that doesn’t naturally flow from having gone to grad school. There’s some very interesting wisdom in your approach. What’s the story? You must have great coaches.

Kieran Snyder:          Yeah, I have tried really intentionally to pattern match off of people who I think are strong right, and so there are a couple of different kinds of people given the kind of work that we do. Obviously, I’ve learned a lot from other start up founders and CEO’s. Because there’s just some common elements to going through this journey, and so I’ve been fortunate enough to get in contact with great communities. Both of people in HR tech space and people kind of in the enterprise space more broadly, so I’ve had some great mentorship there. I’ve gotten to know lots of people who are HR and talent analysts. Who give me a different perspective since the depth of my background wasn’t in recruiting, and so there’s lots that I’ve had to learn there. I came in understanding the technology stack part of it pretty well, so our sort of core technology, the natural language processing pieces are pieces that were closer to my strength. For sure, the mentors and advisors that I’ve been fortunate to have made a huge difference, and I’m always seeking them out, so that’s been a value.

John Sumser:            Great, that’s fantastic, so let’s get to the meat of this. Textio, what does Textio do?

Kieran Snyder:          Yeah, Textio looks at job listings and forecasts how they’re going to perform before you publish them. Then helps you fix it in real time before you publish. We’ve looked at job listings from thousands of different companies, and some information about who has applied for each of the jobs. Then we do kind of what core machine learning tech does, which is we look for patterns, so if we find a set of phrases that correlates really strongly with lots of women applying for that job we kind of take note of that, or we find a set of structural patterns that tends to drive engagement with lots and lots of applicants. We’ll take note of that. Then we use all those patterns as you’re typing your new listing to give you actionable feedback, so you can get yours in the shape you want it before it goes live. That’s the essence of what we do.

John Sumser:            That’s great, so what’s market reception like?

Kieran Snyder:          Market reception’s been awesome. We launched our beta in mid March. Over the course of the twelve week beta we grew to almost six hundred companies actively using the product. We recently launched our commercial offering just in the last maybe ten days or so. We have several companies that are now signed on and using the commercial offering. It’s been great. We’ve had quite a bit of engagement with other companies around the HR tech space. Because I think we’re doing something that is fairly unique within the space. There are lots of people starting to use data science in hiring, and recruiting, and HR, but really focusing on the specifics of language, and using kind of traditional marketing language optimization techniques is a newer thing for talent. We’ve been really pleased with the reception so far.

John Sumser:            Are you able to reach outside of the sort of technology company early adopter market and see what the core HR slash recruiting pro is like as a potential customer base? Is this something that people get outside of tech?

Kieran Snyder:          Yeah, that’s a great question. When we started in the first couple of weeks of our initial beta offering about eighty percent of our companies were tech. Now it’s below fifty, so we still have a significant portion of tech because tech is sort of early adopter, as you note, and we’re tech so we know people in tech. More than half our companies now are not in technology. They range, they’re in retail, they’re in consumer package goods, finance. It’s a pretty big range. A big eye opener for me has been working with the many many staffing firms, RPO’s around the world. Because that’s a newer segment for me given my background, but we have several pretty active conversations with all kinds of companies at this point.

John Sumser:            It’s interesting, have you done a B2B company before? Have you worked in one or operated one?

Kieran Snyder:          I mean I was at Microsoft, right. I worked in Windows, so we were building the Windows platform, which is fundamentally B2B. This is my first project where I really spear headed the two part of the B2B. If you know what I mean.

John Sumser:            Yes.

Kieran Snyder:          There’s a lot in it that I’m learning.

John Sumser:            Market segments inside of HR are a surprise I think amongst people, and what might look like a single customer may turn out to be a sales channel. That’s the most interesting thing about the staffing [inaudible 00:11:50].

Kieran Snyder:          Absolutely.

John Sumser:            There must be a bigger picture. There must be a bigger picture. You don’t strike me as somebody who would start a company to do text analysis [inaudible 00:12:03].

Kieran Snyder:          Thank you.

John Sumser:            Is there a bigger picture, and where are you going?

Kieran Snyder:          Yeah, there is a bigger picture, right. It’s funny when we were first prototyping Textio technology we didn’t start with talent. We actually built a kick starter predictor. Because we had this intention that maybe how you write is sometimes more important in making a sale than what you write. It turned to be true, a little bit sad for me as a life long product person, but the quality of the idea you have in your kick starter project has much less to do with your ability to raise money than how you write the pitch. The fonts you use, the length of the pitch, the length of your sentences, and so that was kind of eye opening for us. We’re applying the technology right now in job listings. Which in a way are another kind of sales and marketing content where a company is trying to sell their company, but really broadly speaking we want to be anywhere you’re writing text where you’re trying to convince somebody of something. Where we can use kind of former outcomes data to help you optimize it in real time. We like to say that we’re as easy as spell check, but powered by machine learning and data, so really bringing core machine learning tech into consumer [offering 00:13:32] experiences.

John Sumser:            The idea that doesn’t quite come out there that I think you’re talking around is that you are able to teach people how to be more persuasive or more effective in their communications, doing things that wouldn’t necessarily look like improved persuasion, so like focusing on the measurable aspects of language. You’re getting an editorial insight that’s not possible otherwise.

Kieran Snyder:          That’s a great way to say it. That’s a great way to say it. You know when you’re really measuring outcomes over time you can offer a level of insight that maybe is close to intuitive for great writers, but even there is usually below the level of consciousness. Just in the jobs domain with Textio we now have close to thirty thousand unique phrases that we highlight as predictively significant. That’s a bigger list than anybody can keep in their head consciously. The more data you get the more you grow. I think that’s a great way to say it.

John Sumser:            Do you track things like syllable count and sentence length against performance, so do you end up recommending Hemingway over Faulkner as a style?

Kieran Snyder:          We explicitly don’t play in the creative writing sphere right now. Yes, we do with the kinds of content that we look at, we track a number of what we call structural elements, so you know the density of verbs that you use versus adjectives. You want a different mix and different kinds of content, or sentence length, as an example, that you used. Which is a great one, or the way you use white space on the page. Sometimes it’s more structural and visual. How many of your words are in bullet list versus paragraph prose? We look at lots of things beyond just the specifics of your word choices that turn out to make a difference.

John Sumser:            Computational linguistics is a big field, and it’s like the root of your work. The more that I get to know people who spend their time in computational linguistics it seems like the earliest conceivable days in a discipline that’s actually quite extraordinary. Tell me about computational linguistics, just a little bit.

Kieran Snyder:          Yeah, so where I went to school, which was the University of Pennsylvania, then a long history of really empirical approaches to language, you know both computational and otherwise. I was raised in a school that believed that measurement is extremely important. Computational linguistics kind of developed originally out of computer science, Alan Turing and Noam Chomsky. Where I was and increasingly in the world, there’s an attempt to tie kind of theoretical computer science with this really empirical measurement oriented approach, and so when you kind of combine those two things what you get is what modern computational linguists are doing. Where they’re looking at huge corpora, the presence of the internet has totally changed how people do computational linguistics now compared to twenty years ago, thirty years ago. You’ve got huge corpora and you start out with a hypothesis, and the internet is so democratized with respect to the availability of text, and Google made it possible for everybody to search that text, and find things that you want to find.

The tools are available for computational linguistics, broadly speaking, are just so pervasive right now. I think it’s why you’re seeing industry over industry why these techniques are coming to change how people work in a real way. We think of what we’re doing, and I think there are lots of companies in the space really making everyday business tasks smarter by means of computational linguistics and natural language processing in a way that would have been just much harder twenty years ago.

John Sumser:            That’s really interesting. The possibilities are limitless and the core idea, to be sloppy about it, the core idea seems to be how do you do the equivalent of search engine optimization in everything?

Kieran Snyder:          Totally, I mean that’s exactly it.

John Sumser:            That’s crazy, because search engine optimization shouldn’t be a thing.

Kieran Snyder:          You know it’s sort of search engine optimization for your brain. Is what it ends of meaning.

John Sumser:            Yeah, what a fantastic thing. What a fantastic thing. There’s a lot of paranoia about what happens when you can knit together pieces of data that didn’t used to exist, and how being able to add two from over here plus two from over there gives you six right in the middle, and all the sudden you can see things about me that you couldn’t see before. Do you think that computational linguistics starts to run into privacy issues at some point in time? The closer you get to predicting my behavior the more complicated that gets.

Kieran Snyder:          Yeah, I mean I think all of technology is running into privacy considerations, right. Computational linguistics happens to work in the language domain, but analogous questions about our boundaries on privacy exist anytime anybody’s using any sort of learning science. Which is increasingly common, right. The shopping patterns that you have on an eCommerce shopping site may have nothing to do with language, and may not use computational linguistic techniques, but as they’re being tracked there are questions about privacy. I think computational linguistics isn’t special here, but it is obviously a very pervasive issue with text today.

John Sumser:            Yeah, it’s a very interesting thing. It’s as if in the privacy concerns about telephone conversations. It turns out that the meta data about the phone call discloses almost as much as the content of the phone call does. About how things work, and who works with whom, and those sorts of [questions 00:20:45]. You start to see this thing where personal data can be manufactured out of non personal components. Where things that you wouldn’t believe had enough of you in the transaction so that you could be able to see you in the aggregate. That’s a stunning advance in how humans examine the world. It must be intoxicating to be in the middle of that.

Kieran Snyder:          You know it is. Just from a purely computational standpoint the opportunities are so much bigger now for creating personal and relevant experiences for the reasons that you described. I think a large part of the challenge for anybody building a business in this space is to make sure that you’re doing that benevolently and in a way where your users, the people using your product, have some ability to understand what bread crumbs they’re leaving, and have some ability to control it too. I think that’s a really important element. There’s some studies that suggest, you know there’s some generational affect here, that younger generations have less preoccupation with privacy. I’ve seen counter studies that say, “Nope they’re just as concerned about privacy, but they don’t think they can do anything about it. They don’t think they have options.” I think it behooves tech companies working in spaces where you do collect any amount of personal usage data to make sure that your users have some options. I think that’s kind of the name of the game there.

John Sumser:            I find the evolution of the privacy conversation to be pretty interesting. If you go from say three billion people to six billion people seems like it might get a little bit more crowded, and it might not be quite as private as it was. When you get to nine billion people it seems to me that the expectation that you’ll be left alone is sort of archaic I guess. I’m always surprised that people are surprised that when it gets more crowded it gets less private. Our time has blown by. What should I have asked you?

Kieran Snyder:          You know the thing that you began talking about at the beginning about language being below the level of consciousness. The one thing that I would leave people with that I think is a great thing to talk about is how much language changes over time. When we look at patterns that we see in jobs, even just in the job and hiring space, it turns out that the language that’s effective to hire a year ago, or six months ago, isn’t necessarily effective today. I think when any discussion of language inherently includes a discussion of change. Because the patterns that work change, our brains change, language changes, and language develops. I think that that’s an interesting component here. One of my favorite HR examples, since we’re talking HR, is the phrase “workforce analytics” and how it has developed into the phrase “people analytics” over the last year or two. If you use the phrase “workforce analytics” in a job today without meaning to you may sound a little bit dated, and change the mix of people who will consider applying for your job. The element of language change is part of why learning science is so critical for any of these problem spaces.

John Sumser:            You’re looking at a particularly interesting space. I often try to help people understand recruiting by talking about fishing. In recruiting everybody’s fishing and using the same bait in the same fishing hole doesn’t produce anything interesting at all. In order to be successful as a recruiter you have to find a new fishing hole with new bait. What you see inside of recruiting is one of the fastest environments for the change of language. That’s because there’s an imbalance between the employers and potential employees. Potential employees are rarely successful statistically in their job hunts. The response rates are marketing level response rates. Employers on the other side of it are often successful. From the employer’s perspective the pace of change in bait and fishing hole is very rapid. From the employee’s perspective, from the candidate’s perspective, it changes very very slowly. I think that one of the things you’re going to find out in the coming years is that the evolution of language is just like you’d expect it to be. It’s two sided. The two sides have different velocities, different velocities.

Kieran Snyder:          Yeah, that’s a great way to say it.

John Sumser:            I am excited to watch your company evolve. It’s been great having you on the show. I wonder if you’d take a moment to reintroduce yourself and tell people how to get a hold of you.

Kieran Snyder:          Yes, thank you, it’s been really really fun chatting with you this morning. My name is Kieran Snyder. I’m the CEO of Textio. You can follow me on Twitter @KieranSnyder, or you can always kind of connect with me on, where you can write to us and tell us what you think.

John Sumser:            Thanks for being on Kieran. I really appreciate the time this morning. Thanks everybody for tuning in. This is John Sumser, and you have been listening to HR Examiner Radio. Have a great weekend.

End transcript

Tagged with:  
Read previous post:
illustration of numeral 5 on article with five links published 2015-06-29
Five Links: Recruiting is Regional (and so is HR)

Talent Acquisition and Management is most often an expression of local culture, customs, and, manners. Today's links give a good...