Yesterday Knewton, the Union Square-based online education startup, announced a $33 million Series D round led by Founder’s Fund, the VC firm co-founded by Peter Thiel. That might explain why Betabeat heard Mr. Thiel’s fellow co-founders Ken Howery and Luke Nosek were throwing a pre-game party Friday night in New York.
Existing investors Accel, Bessemer and FirstMark also participated in the round, along with Pearson, an education publisher, putting Knewton’s valuation higher than $150 million, according to TechCrunch. Another New York City-based education startup, 2Tor (get it??), raised $32.5 million earlier this year. But what sets Knewton apart is the adaptive learning algorithm the company developed, which figures out student’s weakness and can be applied to any type of curriculum.
Indeed, after trying its platform out in test prep, Knewton is now being used by all 10,000 incoming freshman at Arizona State for an online math readiness course.
Betabeat talked to COO David Liu about why Knewton isn’t making teachers obsolete, how its adaptive learning algorithm works, why Mr. Thiel would invest in an education startup and why Mr. Liu thinks Knewton is, basically, going to take over the world of personalized education.
It seems like there’s a lot of buzz around online education startups recently. There’s 2Tor, Veri, Code Academy, and General Assembly may even get into the game.
We definitely see ourselves in the center of it. We have this huge industry now transforming to digital and it’s not enough. E-books are not going to disrupt an industry. But it is a very fundamental step that needs to happen for the traditional materials industry—getting it online is a huge step. Once you do that, Knewton becomes incredibly viable for a lot of these content providers because what we do is we take content and tag it and break it down to a very fundamental level.
That’s where adaptive learning comes in?
To give you an example, we take, let’s say a text book in geometry, and we break that down. If we know you’re weak in a concept, we’ll look within geometry. We’ll find maybe you have issues with triangles. But maybe it’s not just triangles, it’s right triangles. Then we’ll go all the way down to the Pythagorean Theorem. We will know because the content is tagged at such an elemental level that we can reformulate that content back to you based upon what you do on our system. One half of it is breaking the content down to that level, no one else is doing that.
How do you tag a text book down to that level?
We have a template. We used to do it manually, but certainly auto-tagging is something that’s being done today. Even the largest publishers in the world are beginning to automate a lot of that work. Just doing that is the first step. Then you have to put it into a hierarchy, you have to put it into what we call a “knowledge graph,” which kind of formulates what concepts goes before another, what comes after. So that’s a little bite of our proprietary stuff.
How do you know where a student’s weakness is?
What we also do on the front end of that is we’re measuring and we’re processing thousands of data points an hour on every student on the system. They are answering questions, and what we call assessments on questions. Or how much time they’re spending on a piece of video content. Or how much time it takes for them to read through content. Whether they’re moving faster in the morning or in the afternoon or in the evening. Imagine what a user does on Facebook or Netflix, we’re doing the same thing in terms of measuring what data they’re producing on the system when they’re on a digital course that we’re powering.
This is where the algorithm comes in?
We’re taking all that information, we put it into our algorithm. Then based upon where you are in the course or where you’re supposed to be–or based on your individual ability as a unique student–we then produce that next bite-sized piece of content that you need to master. What we’re finding is that a lot of people in the industry are beginning to throw around the term “adaptive learning,” but we’re really one of the few companies that actually does adaptive learning this way. Most companies when they talk about adaptive learning, they’re talking about adaptive testing, that’s been around for years. If you get a question wrong, it gives you an easier question.
Well, you mentioned Netflix, which has faced a lot of pushback from the old guard. Are you going to make teachers obsolete?
We’re absolutely not making teachers obsolete, we’re taking all the dirty work out of teaching so that professors can spend a lot more time teaching instead of trying to pour inefficiently through data—or even not even having access to data about individuals in their class. We’re so incredibly precise and surgical about understanding how individual students are progressing, so we can highlight those individual problems so quickly so professors can come in and do with the class what they do best and that’s instruct and teach.
That’s how they’re using Knewton at Arizona State?
That’s what Arizona State saw first with our college readiness course and now we’re powering a blended learning first year math course across the campus. What they found was that the data was just as valuable as the technology. We’re able to now create a user profile of each student, confidential to the professor, obviously, and they’re now able to make sure during a semester that students can progress as quickly as they need to. Of course the learning outcome is what we’re all interested in. We want to make sure that the kids are learning and that they’re learning in the most efficient and effective way. When they jump into another course, after they take another one of our courses, you don’t have a cold start. So when the student comes in that professor already knows their strengths and weaknesses.
Is Knewton going to focus on the college market from now on?
We started out by proving the efficacy of the technology in test prep and we sold those courses ourselves and now we’re in higher ed, licensing the technology to universities and building some of these courses for universities so they can see the power of platform and the technology. The next step is to continue to broaden this out so that we can now power any content provider’s content and education so that we’ll be powering the materials part of the industry.
Why focus on materials?
That does a couple of things. One, you don’t have to change what people are doing today because text books still drive a ton of syllabuses in any course of study. It’s the big flow of how education content is being distributed, but now electronically and through the cloud. The second thing this does is drive big adoption of the platform much faster. So while we can still and will go school to school, the materials space, the materials industry, big publishers and all the rest hit millions of students all at once at different schools all around the world. That’s where I think the potential of this becomes very powerful because we can revolutionize that industry very very quickly.
That’s where the API comes in?
We’re on our way in the next year or so to opening up our platform where any company or individual will be able to come in and begin to build just unbelievable courses using our technology. So we’ll have a toolkit just like any another API provider and you’ll be able to build a business or build a textbook, if you will, with the world’s best content in it, including publishers, nontraditional publishers, and even open education resource content from an MIT or Code Academy, we want to power all that.
What are Knewton’s revenue streams?
Well we’re still working on all this stuff, so our business model overall is to license our technology. We’re a tech platform and we license that technology. When we open up our platform it’ll be a freemium model platform, so that if you’re a non-profit and you don’t charge for content, we’re not gonna charge you access to our APIs, but if you are a business then we’ll charge you some variable rate to have our technology.
If your platform is offering content, wouldn’t materials companies like Pearson want a cut?
For a publisher, I suppose that we would license it to the publisher and then share in the revenue. I mean they’re interested in making their content and their course solutions much more effective, so you know ultimately hopefully it will change the economics of their business as well.
Peter Thiel has been very local about his lack of faith in the power of a formal college education. Why go with Founder’s Fund?
We’re incredibly thrilled. Founder’s, outside of Peter Thiel’s comments about education, I’m just commenting on Founder’s on a higher level, they believe in investing in companies that can change the world in their respective industries. Look at Tesla, look at Palantir, look at SpaceX. So we’re incredibly thrilled to be working for them because they believe we’re that type of company for education. Ultimately we want to be able to provide personalized learning and better outcomes for every person on the planet. It’s a big, big meaty kind of vision and we’re gonna build many paths to that goal.
Do you find it surprising that Founder’s would be looking an education startup at all?
With respect to Peter Thiel’s comments, again I’m not gonna comment on his earlier comments. I think what they’re interested in as a firm is to really fundamentally improve, big big markets that have been under-performing. I think that’s really all he wants to do. So I think it’s a great statement about us that they found something that they believe can actually make that change.