Albert Wenger is a partner at New York-based technology venture capital firm Union Square Ventures, an early investor in Silicon Valley success stories that include Twitter, set for an anticipated $1-billion IPO, and Tumblr, which Yahoo bought for $1.1 billion in June. It’s also invested north of the border, in Toronto-based online storytelling site Wattpad and Waterloo smartphone messaging service Kik. Wenger stopped in at the new startup accelerator OneEleven (which shares a downtown Toronto office with Google) on Oct. 23 for a public discussion on the state of the Canadian tech sector, how the Internet is changing the face of education and health care and why the death of privacy is a good thing.
His conversation with William Mougayar, founder of Startup Management, has been edited and condensed:
Q: Lets talk about the Union Square Venture thesis. It’s known for pinning down the large networks of engaged users very, very early on.
A: When we look at the Internet we see a technology that has some very interesting characteristics. It’s global. It’s instantaneous. It’s free. It’s connected. It’s ubiquitous. Those five characteristics of the Internet are different from any other technology that’s come before us. We spend a lot of our time thinking: What does this make possible that wasn’t previously possible? We’re very much trying to find things where you’re not using the Internet or mobile to do something 10 per cent cheaper or 20 per cent faster, but where you’re using it to do something that really could not have been done before. If you think of something like Wikipedia and then you try to imagine how Wikipedia would have done with a fax machine, you pretty quickly realize it simply couldn’t have been possible.
One of the things that this constellation then makes possible is the formation of many, many new networks. We’re particularly interested in finding businesses that have a “network effect.” If you think of historical businesses, manufacturing businesses, if you grow your volume, your unit costs decline. That’s traditional scale effect. On the other hand, network effects mean as your service grows, the value of your service to each participant increases. Where we found a lot of network effects early on was in these networks of highly engaged consumers. We’ve later begun to understand that these network effects exist in other businesses as well.
A business where it may not be obvious that there’s a network effect is a search engine. It would seem you use a search engine by yourself and there’s no network effect. But it turns out every user who uses a search engine makes the search engine slightly smarter, because the user will have given the search engine a query, the search engine will have returned a result set and the user is then making a choice which result to pick. That’s why it’s been very hard for anybody to catch the dominant search engine. So we look for the network effects, not just through the obvious network effect of one person following another person on Twitter, but also through the more subtle network effects that exist at the data layer.
When we’ve approached investing we’ve never said: “Ok we have to do investments in finance.” We discovered that there were certain types of business in finance that seemed to have strong network effects. So [peer-to-peer] lending has network effects, which is that the lenders want to be where the borrowers are and the borrowers want to be where the lenders are. That’s a type of network effect. The more lenders there are, that creates value for every borrower. Additional borrowers create value for every lender. So that has led us into those areas, as opposed to us saying hey that’s an area we should invest in.
Incidentally, we think about geography much the same way. So we don’t roll out of bed and say: Oh we should invest in Toronto. What happens is we find a really interesting company in Toronto or in Waterloo and then we go invest in that. We are usually of the opinion if the company is happy where it is we should fund the company where it is instead of telling the company that it should be somewhere else.
Q: How do you see the challenge with consumer companies where the product that the users use is not the same product that makes the money? Like Twitter and Google: It’s the advertising product that makes the money.
A: The idea that’s central to some of our thinking about revenue is the idea of a “native revenue” model. It’s best illustrated by comparing what happened in keyword advertising. Google did not invent keyword advertising. A company that came out of Idealab in L.A. called GoTo invented keyword advertising. It was later acquired by Yahoo. Their idea was simple: People are searching for things and so we’re going to let people bid on keywords. It seemed like a very good idea, except that they let anybody bid on any keyword. So the ads that showed up were the ads by the highest bidder, which were often completely irrelevant to the thing that the person was looking for. The great innovation that Google had was that they realized you needed such a thing as a “quality score” and that the quality score should determine how much somebody has to bid in order to place the ad. If the quality score is sufficiently bad you don’t show the ad at all, no matter how much money is being offered. That turned keyword advertising into a native revenue model. Native in two senses: It can only exist inside [a] search [engine]. And if it works, it ideally makes the experience better, but certainly not worse.
Almost any business on the Internet, we believe, can have a native revenue model. The sponsored Tweets I see in my timeline right now tend to be from sponsors that I actually care about. That’s adding value. We want to give companies as much time as possible to go find the native revenue model. And that may require many different experiments. The thing about experimenting is that you want to start it early and repeat it often because you don’t know what the right experiment is. But I think there are certain lessons that have been learned now that are repeatable. So the idea that if your content unit is a Tweet, your advertising unit should probably be a Tweet. That is now something that is better understood. If your content unit is a Tumblr post your advertising unit should be a Tumblr post.
Where things sometimes go wrong is that entrepreneurs can underestimate how long it can take, not only to figure out the revenue model, but then tweak it and then build the sales infrastructure and the back-end infrastructure. It’s the hidden part of the product to make it all actually work.
Q: This brings me to the topic of Tumblr [and its sale to Yahoo.] Could it be said that they took it as far as they could inside and had to sell because they were not able to monetize on their own? It is a reality that not all entrepreneurs can take their company as high as they can without help from the outside, or without needing to be acquired?
A: When you look at a company such as Tumblr and its trajectory, David [founder David Karp] built something absolutely amazing from day one. As far back as when we invested it was already at a miniscule scale, but it was growing very, very rapidly. When you have something like that you’re still faced with a challenge of building an actual company that can deliver all that. In Tumblr’s case that required a couple of different things that needed to be built. One is a massive infrastructure; Tumblr’s traffic is very image and media-rich and its dashboard is actually quite complicated. Then there’s the ad side of the business that needed to be built. I would say Tumblr is an example where the building of the ad side started late and was ramping very, very nicely, but it was still a ways to go before you had a profitable company. In that situation, I think it’s a legitimate choice for an entrepreneur that if somebody comes along and makes a very compelling offer, to say: I sort of know we’re on the right trajectory, but there’s also still a lot of risk involved. And I think David and Marissa [Yahoo CEO Marissa Mayer] developed a very strong rapport and it seemed like a perfect fit from both sides. Otherwise the company would have had to raise more money, absolutely.
Q: Let’s talk about Canada. We’re talking about $500 million invested in soft tech. It’s a lot more than last year. But in New York last month there was $200 million. But about half of the [Canadian] deals had some U.S. venture capital in them. Is that good or bad for Canada?
A: Our observation is there are a lot of interesting companies being built here. What we believe is that the money follows the interesting companies. People say Union Square Ventures, you helped build the New York ecosystem. But that’s misreading cause and effect. We’ve been successful investing in the New York ecosystem because it’s there and eventually that ecosystem begets itself. When you have an exit like Tumblr there’s a whole bunch of people who make money, which they will then invest in other startups at the early stage, long before we invest usually. It feels to me like instead of looking at necessarily the current gap, it’s more important to look at the trajectory and the trajectory [in Canada] feels like it’s very good.
Q: Are you getting a lot of requests from Canadian companies right now?
A: We don’t really track the origin of requests.
Q: Are you going to invest in a Canadian company in the next six months?
A: Not that I know of. But I don’t even know anything we’re investing in, in the next six months until I see it.
Q: Let’s talk about education. You’ve been talking about homeschooling on your blog. You’re a big believer in homeschooling.
A: Four or five years ago we were all sitting around at Union Square Venture and thinking about what are some of these areas where these network effects are going to play themselves out? One of the things that is also central to this thesis is “unbundling.” Think about a newspaper. The reason why it historically gave you sports and politics and business news and culture news and stock quotes is because it was very expensive to print and distribute. But with the Internet that all went away. With one click I can go from ESPN to Politico. So we started to think about what other areas had that same characteristic and it became clear that one of the biggest bundlers of information was education.
If you think of universities, you go for four years and you get all your courses from the same institution. How did that model evolve? Well that model evolved with the Sorbonne, which was the first modern university in the 1700s when, if you wanted to hear a person speak, you had to be in the same room at the same time and preferably close to the speaker. If you wanted to read a book, you had to be in the library, where they had the one copy of the book, and hope that 17 other people weren’t trying to read it. None of that is true anymore.
I got interested in [homeschooling] because I had been thinking that not only is university such a bundle, but K-12 is such a bundle as well. Why do we have this bundle and why is it we think every 13-year-old needs to learn the same thing in math? Where did that come from? [What’s been found] is the rise of modern schooling is coincident with the rise of industrial society. So we’ve created an industrial process for creating people who fit into an industrial society. The more you start to think about it, the more preposterous that idea appears.
At the beginning of 2012 we moved back to the city and I basically convinced my wife we should experiment with six months of homeschooling [their three kids aged 11-13]. Then they went back to a school for a year. Toward the end of that school year they said to us: Can we go back to homeschooling? So now we’re back [homeschooling] as of this past September.
Q: How does it work, do you and your wife alternate?
A: No. My wife is an entrepreneur. She’s got a startup called Ziggeo and she’s busy with that. We hired tutors and they come to us. It’s pretty easy to find tutors in New York because New York is full of very talented people whose jobs give them a lot of flexibility. So one of our tutors is a published writer who spends a lot of her time writing, but she’s completely in control of her own schedule. Another is a professional after-school tutor, who has a lot of work starting at 3 p.m. but has nothing between 8 a.m. and 3 p.m. And of course we also have the opportunity to avail ourselves of all the latest technology. Khan Academy [which offers free online courses] has come an incredibly long way. They have thousands of videos and they have exercises and they’ve tied the exercises to the videos and they’ve created this knowledge map. It’s really amazing. And schools generally don’t use it. Just like they don’t use Wikipedia.
Q: Is this the future of education?
A: I’m not suggesting that our model of hiring tutors is scalable. But some of the components of what we’re doing is the future of education. The inverted-classroom model: A lot of our tutors will tell the kids go watch this video of this TED talk and then when you come back we’ll discuss the TED talk. As opposed to the tutor presenting what’s in the TED talk. This inverted classroom model is something that is scalable. It will just take a long time for institutions, which have for a very long time been used to doing things one way, to adopt these quite racial, almost 180-degree flip models.
Q: What about health care?
A: Right now there’s a lot of legislation that deals with the privacy of health care data. The legislation is all aimed at trying to say this is a patient’s data, it needs to stay private, it needs to stay protected. In the pre-Internet days that made a lot of sense. When you had a paper record of the patient, you didn’t want that paper record to be randomly flying around because there was really no upside for the patient in that.
With the Internet that actually should be turned on its head. What we should really be looking for is legislation that lets each and every one of us publish our own medical records and then protects us from the downside. In that kind of system all our medical knowledge would explode along with our ability to understand where certain diseases come from, where they’re prevalent, or what new diseases are developing. By doubling down on the idea that patient privacy is the problem and that we need to figure out ways of locking up patient records and making them only accessible to the patient and the specific authorized doctors, we’re denying ourselves this huge opportunity to grow and understand why we’re sick in the first place. We’re just at the beginning of a very radical change, because with the Internet often things are the very opposite of what they were before. So for us individually, and as a society, to wrap our heads around the idea that it would actually be desirable for our medical information to be out there is a lot of steps from where we sit today.
Q: Talk about another one of your interests: machine learning. And make it interesting.
A: We now actually have machines that are better than humans at many interesting tasks. Such as image recognition. Such as driving a car safely. The rate of improvement here is really, really dramatic. As recently as 2005 the best [self-driving] car got 50 miles into a desert circuit that had no other cars, no traffic lights, no intersections. And now it’s 2013 and we have self-driving cars that have driven hundreds of thousands of miles on public roads accident-free.
The challenge to all of us is multiple-fold: What are the areas where we’re going to use this? And how are we going to build this so it’s useful for most people? If we can build something that is better at diagnosing human disease, how do we make that widely available as opposed to building something and then selling it for exorbitant amounts of money to the health care system and making the health care system even more expensive rather than cheaper?
The other big challenge is what do we do about work? It’s no secret that we’ve had this very strange economy for 10-plus years now where the economy keeps growing, but wages are stagnant and the share of labour as part of gross domestic product is declining. I firmly believe that’s the beginning of the post-industrial age. It’s the age that’s been talked about since the late-70s, but has finally arrived, where machines are better than humans at a lot of things. So the demand for human labour is actually decreasing. That will be a huge challenge for how we organize society in that world. I think we’re just at the beginning, but the trend lines are already visible. And that will be a huge challenge for the system as we know it.
Q: Surveillance, we all know it’s going on right now. I think you believe it’s being done the wrong way. So what is the right way?
A: Your privacy is, in fact, dead. There’s a start-up in New York that’s letting people turn their old Android phones into cameras that people can put out in their windows to film the streets to measure where people are walking. You’re going to be taped. You’re going to have an electronic footprint. It’s just a fact. The big question is what do we do about it and how does government operate?
The fundamental problem that we have is not that surveillance exists, but that governments are trying to be secretive about it and that government is trying to use secret evidence and secret trials and that is completely undemocratic. We need to figure out how to avail ourselves of these technologies, just like with health care. I don’t think we can fight these technologies. I don’t think we can put the genie back in the bottle. But I think we can apply these same sorts of technologies to government. We need to double-down and go in that direction. Again, it’s very far from what our thinking is today. But I think that’s the way forward. The way forward is not for us to try and have better encryption and to figure out how we can stash our data away and prevent government from snooping. We have to have a government where we know what snooping they’re doing and where we can trust that the snooping they’re doing is actually in our interests, as opposed to used to oppress dissent.