Andrew: Hey there, freedom fighters. My name is Andrew Warner. I’m the founder of Mixergy.com. It is home of the ambitious upstart. It’s the place where real entrepreneurs come to hear how other entrepreneurs built their businesses.
Joining me today is one of the first or one of the very first listeners of Mixergy. He’s a guy who I’m really proud to have on here because he’s a hard worker who started a business–and you’ll hear when I say hard worker, everyone says they’re a hard worker, wait until you hear what this guy did–he started a company that now when you go to AP and so many other sources, it feels like a human being wrote an article that’s interesting, that explains a topic really well, but it’s actually his system, his software that actually wrote the article based on the stats.
The AP recently, I guess, 2014, said that they were shifting to Automated Insights and almost all their quarterly earnings were being written by Automated Insights. It’s a really interesting technology. It’s a really interesting company that he launched while working a full-time job and sold not too long ago, actually just a few weeks ago. And I’m excited to have him here to talk about how he did it.
His name is Robbie Allen. He is the founder of Automated Insights. They have a natural language generation platform that takes data and turns it into stories that sound like a person, a human being wrote it. And this interview is sponsored by the company that will help you find and hire your next developer. It’s called Toptal. And by the company that will host your website right. It’s called HostGator. I’ll tell you more about them later.
First, I’ve got to welcome Robbie. Robbie, good to have you on here.
Robbie: Thanks, Andrew. Super-excited to be here.
Andrew: Me too to have you on here. Why did you listen to Mixergy all those years ago?
Robbie: So I started the precursor to Automated Insights in 2007. I really didn’t know a lot about entrepreneurship. So I started looking around for interviews and blogs and that type of thing for entrepreneurs to get advice. Yours was just starting, I think, around 2008.
Robbie: It was just a great source. I remember those early days you had some great speakers on that I was able to soak it up and learn a lot from it.
Andrew: And I keep looking at how far so many of them have come, like Brian from Airbnb. I just keep seeing him grow and grow and grow and Drew from Dropbox–so many interesting people. And you’re a guy who doesn’t have an entrepreneurship background. So, I’d understand why you’d really want to soak this stuff up. In fact, you told our producer, Jeremy Weisz that your dad was the furthest from an entrepreneur. What did he do?
Robbie: Yeah. So, he worked at AT&T for 30 years in manufacturing. In fact, he came home most days complaining about his jobs. It was so bad back in those days that the workforce would go on strikes. I remember several times where he actually had to go picket while they were on strike. It sounded like a pretty horrible existence where he worked.
So, instead of growing up thinking that, “All right, I’m going to have to go work in the man and be stuck in this horrible existence. There’s got to be a better way. Why does it have to be that way?” So, I kind of grew up with more of a curiosity of trying to make the work environment a better place to be than thinking that it had to be this type of way.
Andrew: Not that kind of place that people want to go out and picket. It’s interesting. My maternal grandfather worked at the Coca Cola factory. When my dad married my mom and became an entrepreneur, it was almost a little embarrassing, like he didn’t have a job. They didn’t understand this whole thing that he was doing, the fact that he didn’t get up at–I don’t know what time, 6:00 a.m., let’s say, when everyone else was getting up, but would get up at like 9:00 was just a shock and a horror.
She had to just say, “Please, get up and out of bed.” They were living at my grandparents’ house. “Just get up so you’re not embarrassing me with this new husband that I’ve got here in the house.” It’s completely different. You actually didn’t know what an entrepreneur was until how old?
Robbie: It was probably college before I even heard that word.
Andrew: And then you went out, you got a job, you worked for how long for Cisco?
Robbie: Yeah. I was in computer science as an undergrad and I did an internship at IBM and then I did one the following summer at Cisco. This was around ’97, ’98. Cisco told me at the time after I was finished that this thing called the dotcom boom was happening. They needed programmers. They didn’t care if I had a degree or not. So, they shipped me out to San Jose, California and I just started working there.
Andrew: So, Robbie, you were living in San Jose, California. You were living in the heart of this boom that’s happening and you still continued to work for Cisco for a total of 13 years. You started when you were 19. Why don’t you at some point in the middle of that say, “I’ve got to jump into any one of these dotcoms. They’re going to make me rich tomorrow?”
Robbie: That is a great question. Again, probably because of my upbringing, I still thought the process was you go to college. I was one of the first in my family to go to college. But after that, you go to work somebody for 30 years and you retire. I thought Cisco was that place. In fact, I did work there 13 years. I made it pretty far along. It took me a number of years. I started advising startups. Then after going through that cycle a few times, I started thinking, “I think I could try this.”
Andrew: Then before you tried it though, I said at the top of the interview that you were a hard worker. I’ve seen people say they work crazy hours. I didn’t see them do what you did. What kind of crazy hours did you have?
Robbie: So, especially during the dotcom boom with Cisco, it was pretty nuts. There were several people that I worked with where we actually lived at the office for maybe three days a week.
Andrew: We’re talking about live at the office, you’re talking literally live at the office.
Robbie: Yeah. We’d sleep at the office. The thinking–just to show you the crazy thinking back then–to get home in the bad traffic would take maybe 45 minutes to an hour. So, instead of wasting that time, why not just sleep at the office and then that could buy another two hours of potential productivity instead of wasting it in a car.
Andrew: But why, Robbie? I understand when you’re working for a startup. I understand when your employees were working for you, I’d understand why they would live at the office, you’re really changing stuff. You’re changing an industry. You’re starting something brand new, everybody’s really important. Everyone’s necessary. What is about Cisco that gets that kind of passion, that kind of hard work?
Robbie: I was in my early 20s and I was just soaking it all up. I was having the time of my life. I’m fortunate that I’m one of those people that in tenth grade, I took a business computer class and after that class, I said, “I know what I’m going to do the rest of my life.” So, I just got super lucky that I knew I wanted to work in computers and then when I got into it, I continued to love it. I’d rather do that then spend time doing other things. Again, I wasn’t married at the time. I didn’t have kids. So, I didn’t have other things keeping me preoccupied. I’d rather just learn and continue to do cool stuff.
Andrew: You’re like a dream hire. Do you feel they appreciated it?
Robbie: I left Cisco as a distinguished engineer, which was their top technical position. Ultimately I think Cisco did right by me. I know they got their money’s worth out of me as well. I feel like it was overall a good trade.
Andrew: And when you decided to scale back your hours, you went from 80 to 40, you did something that I’m seeing here on Amazon, on your Amazon page. What did you do?
Robbie: I started writing books.
Robbie: Yeah. Around the time that the bust happened and all the companies were going out of business and even Cisco started doing layoffs, which they had never done before, it became a more political environment. I wasn’t learning as much. But I still had a desire to learn. So, I thought, “What I could do is write books?” It didn’t start out that way. I figured I’d write a book.
And then when I got into it, I realized that there’s no better way to learn about a topic than to write about it. I wrote the first one, which was really difficult and painful and I thought, “I learned a whole lot. Why don’t I just start writing books as a way to educate myself?” So, I wrote ten books over a five-year period, not because I was just an expert on all of those topics, but because I wanted to become an expert in those topics.
Andrew: We’re not talking about books on light topics. We’re talking about those O’Reilly books that used to have black and white animals on the cover, things like “Active Directory Cookbook,” “Windows Server 2003 Networking Recipes,” “Windows XP Cookbook,” “Windows Server Cookbook,” “DNS on Windows Server 2003,” “Managing Enterprise Active Directory Services.” Wow. You learned all those topics with the idea that you would do what with them?
Robbie: Well, part of it, it was kind of part of my job as well. So, at Cisco at the time, I had kind of grown up in a Solaris environment, which Sun Microsystems had created the Solaris operating system that’s based on Unix. But then Cisco was making a push into Windows. So, I was kind of tapped to be one of the Windows people. So, I started learning more about Windows as part of my job and thought, “Well, it’s sort of a good venue for me to start writing and delving down into it even more.”
Andrew: I want to get into how you started Automated Insights, but I’ve got to get into the mindset that gets you to do this stuff. Most people–frankly, even me included–need to really motivate ourselves to do something. I have to think about the end goal. I have to think about how exciting it’s going to be. I have to remind myself that this could have some big impact. What gets you to sit and write so many books to work so hard? What was it?
Robbie: It’s not really like work to me. Again, I think I’m one of the fortunate ones that when I’m sitting on the couch at home, I’ve got to be productive. It’s almost like a desire to be productive, almost being addicted to being productive in some way.
Some people may call it being a workaholic. I don’t call it that. I’d rather do that than some other activity, maybe some people work on cars or they have other things that they do. For me, I just enjoy accomplishing things on the computer. So, that translated into doing well at Cisco and writing books and eventually starting my own company.
Andrew: Okay. I never mean to put down anyone who works for a company of any size. Frankly, there was this old Malcolm Gladwell article about the founder of the television set. There were actually two different people who were trying to do it, one on his own, the other working within RCA.
The one that was working within an organization got the credit for having done it, largely because by working within an organization, he didn’t have to do it all. He had support and it actually came to fruition. He was able to actually see success from it. The big message from that is that we tend to put down working for big companies, but in reality, by working for big companies, we of have much more reach, much more power than if we’re doing things on our own.
So, at some point, though, you decided, “I’m going to do something on my own.” And it looks like you went through a series of ideas trying to figure out what your next one was before the first version of your business hit, right? What was your process?
Robbie: Well, so my process was I had finished writing–I had written a bunch of books. I kind of got tired from that. I was a little burnt out from all the writing. I decided I wanted to go back to school. I had always wanted to go to MIT. Again, I’m a computer science undergrad. It took me ten years to finally get my undergraduate degree.
At that point, I decided I wanted to go back to school and this time what I’m going to do is try to go learn about entrepreneurship or at least get exposed to that notion if I’m really going to start a company someday. So, Cisco was nice enough to let me work part-time and go to school at MIT part-time for a couple of years.
There I got two Master’s degrees, one in engineering management and another in civil engineering. There was just mostly getting exposure to some of the local companies, all the entrepreneurs that MIT would bring in and they would give talks. That’s when I started thinking through what’s the idea that I’m going to build my company on.
Andrew: Okay. And was there a process for it? Were you looking for specific things?
Robbie: Yeah. It was kind of funny. I’m an engineer, right? I thought, “I’ll go take the course on starting a company and then I’ll have everything I need to start my own company.” Of course, it doesn’t work out that way. In fact, it was a little disheartening. Even though they’d bring in all these great speakers, everybody from Steve Ballmer and Jack Welch to the guy that invented the Crest Spinbrush and sold it for $300 million, they all gave conflicting advice.
They all had a completely different path to how they started their company. The one thing that was fairly consistent was they said you need to be passionate about your business. That was something that really resonated with me. As you kind of have talked about here, anytime I can be passionate about what I’m working on, I’ll put in whatever time is necessary to make it happen because I just enjoy it that much.
So, when I started thinking I got this fancy MIT degree, how am I going to use that to really start a company? I think, “I really need to create a list of things I’m passionate about. The problem with that advice is by definition, you’re not passionate about very many things. Out of those, how many are great business ideas? I’m passionate about my eight-year old daughter, but she’s not really ready to start a company yet. So, I made a list–programming, web development and UNC basketball. So, I’m a big Carolina Tar Heel basketball fan.
Robbie: I thought if I could do anything related to those three things, I would just put in whatever time was necessary to be successful. So, at that point, I started looking at ESPN.com and Yahoo Sports and I thought all they showed for stats is really just the box score of a game. They don’t do any sort of historical mining. They don’t do visualizations.
I figured if I could get access to that data, I could do something significantly better than that. That was essentially the initial genesis behind a company, without really any sort of formal business plan, I had just gone through all this training at MIT and used very little of it, frankly, to help craft the approach I was going to use for starting this company.
Andrew: I wonder why you went to look at ESPN and other sports sites when it feels like everybody loves sports and everybody would love to have a sports-based business and many people can actually speak about sports and understand it maybe even as well as you versus software versus computers versus the tech world where it’s a lot harder to break through and you already had so much experience and it feels like it’s much more valued to have that knowledge. Why didn’t you pick that?
Robbie: Again, it was mostly around specifically Carolina basketball. I’m just such a big super fan that I thought if I could do something with that, it would really be something I’d devote my time and attention to. Plus, especially at that point in time, there just weren’t a lot of interesting sports-related startups. I thought the innovation in the sports media space was really low. In fact, it was dominated by these large industry dominant companies that weren’t innovating.
So, to me, I thought what I could do is come in as a very nimble, agile competitor to them and maybe do something that could surprise them and do things that would get me noticed in a way that they couldn’t move quick enough to do.
Andrew: So, college basketball was where you’re going to start focusing. What did that first product look like?
Robbie: Mostly visualizations, a lot of visualizations. Again, you’ve got to think back. This was 2007. Web 2.0 was kind of in its peak at that point. Because I had the connection with writing all the books with O’Reilly, Tim O’Reilly is sort of the father behind Web 2.0. So, I was involved with that movement. I thought, “What if I do Web 2.0 for sports?”
This is almost making feel like it’s dated at this point to even talk about Web 2.0. Back when you were starting Mixergy, it was a very popular topic. Now it’s a little past its prime. That was an interesting concept back then. Big data hadn’t even been named at that point. People weren’t really thinking about data, but I was kind of at the very early stages of thinking about that for sports.
Andrew: What do you mean by–can you give me an example–I’m on your site, by the way, from 2008. I’m using Archive.org to take a look at it. The topic says “Statsheet.com,” that was the URL, “The Ultimate Source for College Basketball Stats, News and Analysis.” I’m trying to figure out is this article–no, that’s written by the AP. What part did you create that was this visualization vision that you had?
Robbie: Yeah. So, if you went to like a team page, if you were able to drill down to a team page, you could see all the different charts and things that were displayed on the site. Again, that was pretty uncommon back in that time. I also created a couple of side projects. For example, Twitter was just starting to get popular around 2008-2009. What I did was created accounts for every college basketball team and I would start auto-tweeting stats and small snippets of text to each of those accounts.
My thinking was at that time ESPN had one account or Yahoo Sports would have one account. I was like, “I’m a UNC basketball fan. I don’t want to hear about Duke basketball.” So, why not make it a team-oriented approach? So, I thought since I’m automating all of this, that’s doable the way that the other companies think about it is a more manual project. They’re not going to staff somebody to tweet out about Carolina basketball. Companies still don’t do this today. It’s got to be automated.
Andrew: You wrote that yourself?
Robbie: Yeah. I wrote all that myself.
Andrew: In your spare time while you still had a full-time job over at Cisco.
Andrew: Okay. And then what was the next step? What was the next evolution of the business?
Robbie: So, it started to get more and more traction. In fact, ESPN invited me to Bristol to give them a sort of overview of what sports and Twitter even means, like again, this was 2009 and ESPN had not done much of anything with Twitter yet. So, I had already created all these accounts. I was tweeting all this stuff. So, it attracted interest, which I thought it might. So, they invited me out. This was 2009. I was still working at Cisco. They didn’t even know I was working at Cisco when they invited me to Bristol to give them this overview of what Twitter means for sports.
So, I started getting more and more attention. I thought I might be at a point now where I could go out and raise money. I figured if I could ever get to that point, I would essentially leave Cisco and devote myself full-time to that new venture. So, in 2010, I was able to raise a $1.3 million seed round and then on my 13-year anniversary at Cisco, I left the company and hired my first employee.
Andrew: When you were shopping around for money, what did you tell them the vision was for this business?
Robbie: So, one of the challenges with sports, even though people perceive it as a large industry, it’s actually a small niche when it comes to internet audiences. It’s also very seasonal. So, basketball season only lasts a few months and then what do you have the other part of the year? You have to essentially have coverage across a lot of seasons. Even college basketball, 90+ percent of the attention is during March. During November, the college basketball audience isn’t that big.
So, I could just go with hey, I’m going to build a better ESPN. There needed to be some catch to it. Well, I started the first blog dedicated to UNC basketball back in 2002. What I would do–this was during the book writing phase.
What I would do is write recaps after every basketball game, my own little summary of why they won or lost. When I was thinking about what am I going to do to propel this idea to something that can be backable, I thought, I had written all these recaps. They’re pretty formulaic. There’s got to be a way to create algorithms that essentially simulate the same type of thing.
In fact, if you look at most recaps, 70% to 80% of it is just quantitative analysis. It’s like how many points were scored. Was that better or worse than the season average of this team? Did they play better or worse? Even statements like, “Marcus Paige had a great game,” while that sounds subjective, it could ultimately be simulated through an algorithm, right? You know whether or not he scored more points than his average. Did he help the team win by looking at play by play data?
So, all these things kind of came together in 2008 and 2009 and I had some early attempts at this before I raised the money. I was convinced that I could automate, at least at that point, sports stories using data with software.
Andrew: And make revenue how?
Robbie: This kind of is the funny part. I went to this MIT graduate program and a lot of learning about entrepreneurship. I didn’t really have a solid business plan. It was much more of the sort of Twitter model of, “Hey we’ve got to get really big.” It’s the typical media model, frankly.
Andrew: How about get users? How are you going to get users for all of this?
Robbie: Well, the nice thing is sports fans were, especially back in those days, looking for new sources. There were new platforms, Twitter and even Facebook. There was nobody really creating content for them. So, it was pretty easy in those early days for me to build audience off of these new platforms.
Andrew: Okay. Let me take a quick sponsorship break and then we’ll come back and talk about how you grew it and how you went beyond sports. The sponsor is a company called HostGator, which is a hosting company. Robbie, the reason that I think it’s so perfect to talk about now is that you started a blog when you were just passionate about a topic. Not because you saw a business in it, not because you knew exactly where it was going to go, but you just started it. Then you allowed it to take you where it went or frankly just be a side hobby.
I’ll tell you, the same thing happened for me with Mixergy. Mixergy for a long time was an invitation software site. It was becoming kind of grueling because it wasn’t working out the way I thought it would. Then I went and I just installed WordPress on a subdomain. It was blog.Mixergy.com. I started playing with it without anyone seeing it. There’s a little radio button on WordPress that lets you say, “Do not show this to Google. Do not make it public.”
I kind of liked getting to play with a site that was actively public but wasn’t promoted in any way. I started playing with it and enjoying it. It was my own little creation. What would it look like if I wrote this article? I’d write a post on there. Then I thought, “This is not the kind of post I’d want to publish. I’d delete it and start over, until I ended up with what became Mixergy today, which is interviews with entrepreneurs. It started with me writing about entrepreneurs that I’d met.
That’s the way if anyone out there is looking for a business to start on the side, maybe because you just want to get your creative juices flowing again. Maybe it’s because you have a passion you haven’t explored because you’re not sure what the business ramifications are. I really urge you to do what I did, which is just go start a site, just play around with it. You don’t even have to promote it or publish it or anything.
You just have to play with it and see where it goes. Robbie, I see you’re nodding as I say this. It’s fun. It opens up your creative juices, your creative channels. And if you want to do that, I recommend going to HostGator for a simple reason–two simple reasons, number one, it’s freaking cheap. We’re talking about a really inexpensive, just a few bucks a month, not a day, to host your website with them.
Number two, they keep that site up. No matter how much bandwidth you throw at them, they’ll handle it. The reason they could do that is because they’re an enormous company. So, you won’t have the headaches of having to deal with keeping your website up or if you have any issues, having to tech support it on your own, they have tech support that will help you.
If this little hosting package is too small for you, they have ways for you to scale up and really deal with massive traffic. So, it will grow with you. If you haven’t started it, go start your site right now at HostGator.com/Mixergy. When you add that /Mixergy, I get credit, frankly, for it. So, they know that I’ve done a good job for them, but also you get 30% off of your hosting package, which is a really good deal.
Go start it, you’ll see they have a money back guarantee, so if you’re not happy, you can walk away and so many other benefits that I won’t get into here because I want to focus you on starting your site, or if you’re not happy with your hosting company, shifting it to HostGator. I’m grateful to them for sponsoring.
All right. So, you got your site up. You had your funding. It’s time to actually start making this into a business. What did you do first? Was it get more people to come to the site, more readers? Or was it form partnerships or was it improve the software?
Robbie: So, it was to scale out the strategy. We wanted to take a very team-oriented approach to sports coverage, so instead of having a single site, you could almost kind of see from my background, I kind of like to go big with things. So, we didn’t just have one site. We created a website, a mobile app, a Twitter account and a Facebook fan page for every team we covered.
So, by the time we were done, we had roughly 500 websites, 500 mobile apps. We had over 1,000 Twitter accounts because we started doing player accounts. We were trying to be one of the first and take this team-oriented approach and cover very personalized stories that for people that were fans of a specific team, they couldn’t quite get that coverage anywhere else.
Andrew: I imagine that you did that by doing it manually, right? There was not code to do it?
Robbie: So, for a lot of the things, we had to automate it. So, I hired roughly six people in the first maybe three or four months, all developers. So, for any of those things, whether it’s even creating a website, even the logos, we had even unique logos across all of those. We had to figure out ways to automate it because we just couldn’t cover that scale with sick people unless it was automated to a certain degree. So, all the content we’d create, any things that were built on stat sheets or any of the Twitter accounts, those were all completely automated.
Andrew: And were the teams okay with you doing that?
Robbie: Some weren’t. We would get cease and desist letters from some teams. We tried to be smart about the names of the websites so that we wouldn’t get in trademark violations. Although, imagine trying to create unique relevant domain names for 500 times. It’s not easy. The worst was Major League Baseball. No matter what we came up with. Unless there was no evidence of a logo, no evidence of a team’s name or resemblance of a name, they weren’t going to be happy. They were probably the hardest to pacify.
Andrew: Really? Do you have an example of a creative workaround for, say, the Yankees?
Robbie: So, like I think the initial one that we had for that YanksPride.com. So, it’s just using Yanks in it.
Andrew: Even that wasn’t good?
Robbie: Oh yeah.
Andrew: Wow. How is someone supposed to get excited about their team and talk about their team if they can’t use the team in what they’re saying?
Robbie: Exactly. They would claim that it couldn’t resemble the team in any way. Like we’d maybe put the logo–so, we had robots. All this was automated. The company mascot was a robot. So, how that played into our favor, every team, they also have mascots. So, we’d take our robot and make it look like the mascot of a team in some way, dressed and oriented robot. So, it couldn’t look too similar for MLB. Most colleges, they didn’t really care. We didn’t hear from them. But I just remember MLB was a tough one.
Andrew: That makes me a little nervous about just wearing a Yankee baseball cap in an interview.
Andrew: Did you do anything to start generating revenue at that point or you were still just starting out as a business that was passionate about a product but not yet revenue?
Robbie: Yeah. So, we had advertising on the site. Of course, you could think of us as a fledgling media empire is what we were trying to portray ourselves as. So, obviously we used advertising is the main source. We started dabbling with the notion of subscription sites. We started generating lots of advanced analytics on top of all the sports data we had. We started thinking about having subscription sites.
The challenge was, “Are we in the business to make a little bit of money or to scale this thing really big so that one day we can make a lot of money. So, typically the answer was the latter. That means we had to put away any barriers to people adopting or coming to our sites and figure that if we can get a large enough scale, we can monetize it later.
Andrew: How big was SEO for you?
Robbie: I didn’t think about it too much, mostly because there wasn’t a lot of competition. Again, there wasn’t competition on Twitter. There weren’t many team-oriented sites at this point, 2010-2011 even. There weren’t a lot of Yankee fan sites, frankly, that had the level of data and analysis that we provided. So, we didn’t have to worry a whole lot about SEO at that time just because we didn’t have a lot of competition.
Andrew: Okay. At what point did you start shifting away from sports to a more general topic or to other topics, I should say?
Robbie: Yeah. I knew we were onto something. I wrote a blog post back in 2010 called, “The Future of Sports Journalism is Automated.” I had written a bunch of blog posts in my day. Again, I had a couple of different blogs over the course of the years. But within ten minutes of me publishing that post, I got three different journalists email me saying they wanted to talk. I was like, “A, how did they even find it that quickly and why do they want to talk?”
Well, this notion that I could create software that could automate what journalists do was very interesting to them. So, that was the first hint that this notion that we’re going to automate the writing process, maybe it’s got built in PR. We’ll actually be able to track a lot of publicity just from that topic alone. So, that’s exactly what happened. I think it was Thanksgiving of 2011. We got a big write up in The New York Times. Again, we were just a really small startup. This was huge for us. Apparently the Thanksgiving edition of The New York Times is one of their largest circulated runs.
As soon as that went out, we started getting people emailing us saying, “It’s cool you can automate sports content, but have you thought about content for X? Have you thought about content for Y?” So, after a few of those, we thought maybe just trying to do this media ad-based business for sports isn’t the long-term gain for us. Maybe it’s actually doing this for more industries on more of a B2B business.
Andrew: And were they potentially going to buy from you?
Robbie: Yeah. Absolutely.
Andrew: They were? They weren’t just saying, you should do this. They were saying you should do this and we will pay for it.
Robbie: Yeah. As a good example, it was one of the largest domain names registrars came to us and said, “Hey, we have hundreds of millions of parked domains.” SEO especially back then was huge. “If you could give us a small amount of relevant content about the domain,” for example, say they had KobeBryantBasketball.net. They said, “If you can even just give us about Kobe’s last performance, that would huge.” Multiply that across a hundred million domains, they could make a lot of money from it
Andrew: I see.
Robbie: So, we started getting all these crazy off the wall ideas that I never had envisioned. But that really made us start thinking about changing our plans.
Andrew: So, part of customer development, I think, involves trying to sell to confirm that you’ve really found an interest. Did you start doing that at all before you created the software?
Robbie: We did. We had already started building the software for sports. We had already done it for college basketball. We started branching out to football and NBA and baseball. So, then we had the idea that sort of a natural adjacency for us is fantasy sports. We had a lot of people in the company at that time that were big fantasy football players. We thought was missing is if we could get a personalized recap of every week’s fantasy matchup, just like we could generate a recap of a sports game, we could generate a recap of a fantasy matchup.
So, we went to all the major fantasy platforms–Yahoo, ESPN, the NFL, CBS–and we pitched this idea. The only one that bit at the time was Yahoo. They said, “Yeah, that sounds kind of interesting.” So, that was really our first foray into changing the shape of the company into more of a B2B company as opposed to a consumer company.
Andrew: And they were your first customer in the end?
Robbie: They were one of the first two or three. They were one of the big ones that we had been hunting. They were by far the biggest that we had at the time.
Andrew: I actually see that–I think I see The New York Times article. Was this the business day, “When the Software is the Sports Writer?”
Andrew: It talked about your robot army. They had this great line in there. Here it is. “We humans love story. A craving for narrative seems to be part of our nature.” That is so true. In some intellectual sense, I feel like you’re wasting my time with the extra words. All I want is the data. Why are you cluttering up the data with extra words? Still, it makes it more interesting. It still sticks in our mind. Do you know why? Why is it that someone can’t just look at a block of data and feel satisfied? Why do they need to have more information on there to feel happier?
Robbie: Yeah. So, there are a couple of things. I get this feedback to this day. I prefer visualizations over anything. Give me a chart or a graph or a table with numbers. The problem with them–again, I have some expertise in this to kind of lend to the argument–what happened with Statsheet is it was an advanced sports statistics website. What we found was that while that’s very interesting to some people, that’s a minority of people that are really that interested in digging down into visualizations.
The reason for that is a visualization or even a table of numbers requires you to do the interpretation. You have to decipher what the X and Y-axis is, the color, the size, all of that stuff, you have to essentially build your internal narrative for what that means. What most people really want is just tell me the three or four sentences that explain this. Don’t make go through the mental gymnastics of deciphering it myself. Just give it to me in plain English.
Andrew: I get what you’re talking about. I think the best example I have of that is someone offered to have a bunch of people on Mechanical Turk give feedback on my site. I thought, “Great, I’m going to get to see what all these people think of my site.” It was way too much data. I couldn’t process it.
Then he comes back and he says, “I’m thinking of analyzing this also for my customers. Can I analyze it for you for free and you tell me if it’s useful?” And I say, “Yeah, I’d love that.” And then he gave me this detailed analysis. What I really wanted was just some clarity. I wanted to know, “What do you think is broken and how should I fix it? Keep it that simple and that clear.” Is that what you’re talking about or am I oversimplifying it?
Robbie: No. I think that’s right. I think most people, especially in the area of big data, which is a little bit of what you’re hinting at. We’re in data overload. It’s almost unreasonable to expect somebody to be able to interpret the quantity of data that we have to deal with nowadays. What most people want is tell me what it means. What do people think of my site in like two sentences? I don’t want to go through and read pages and pages of visualizations. I don’t even want to read pages and pages of narrative. Just give it to me simply in plain English.
Andrew: What about this? I’m on your site right now under the example section. I clicked around and my favorite examples were the real estate examples–again, if you’re looking at data about a house you’re considering buying, it’s not nearly as interesting as seeing a photo of it and hearing or reading one of your description. But I went to the workout recap, which I also like because I get a lot of data on my runs. It’s kind of like blah. I don’t do anything with it. So, I want to see what you guys did.
Here’s what you turned this spreadsheet looking collection of numbers into. “June 13, 2015–Killing the distance! This Saturday afternoon run was a top-five ranked run for you all-time. The ten miles you ran crushed your all-time average distance of 3.3 miles per workout. The extra distance also showed up with 889 calories burned, ranking fifth all time.”
Some of this is really great. Some of it feels clunky and written by a machine. This is your example. Like the second sentence, “This Saturday afternoon run was a top five-ranked run for you all-time.” Am I misreading it or is that just part of where we are today with turning data into stories?
Robbie: So, part of it is that’s an early version of something that we did back in 2012 for a client. Part of it too, there’s some expectation that every version of a story that comes out is some unique thing that you never could have heard before. For example, what we do for Yahoo is we generate millions of fantasy football recaps every single week. I asked around if anyone interested in writing a million fantasy football recaps every week and no on raised their hand.
In fact, if you think about the process of doing that, what a person would go through, it’s very similar to what our software does. In fact, our software kind of emulates what a person does when they write. If a person wrote a million stories, there’s only a certain amount of neatness that would be there or only a certain amount you’d even want or it’s going to start to sound too out there as far as it’s trying too hard, right?
So, some of the implementations that we’ve done where maybe we’re almost trying too hard to meet a uniqueness requirement that really you couldn’t get it if you had people writing that many. So, really there’s just a variety of things kind of at play there, whether it’s really trying to emulate a certain type of writing style or just the maturity of the software and how far it’s come since then.
Andrew: So, what you’re saying is that you’re not at a place where–I think one of the articles I read about you sad, “We’re aiming for 80% of people who read it to think that this was written by a human being. We’re not trying to fool anyone. We’re not aiming for 100.” Is that what we mean here, that we’re not really at 100%? This feels a little bit clunky, but that’s where we are today.
Robbie: Well, I think it’s easy to cherry pick anything. In fact, I could go cherry pick a human written story right now and you might think it was written by a machine. There have actually been studies that have been done where a professor at a university would do a test of a version of a story we wrote and a human generated version and the results were that people couldn’t tell the difference. So, I think you could find good versions of human written stories, just like you can find good versions of machine written stories and vice versa. You can find bad versions on both sides as well.
Andrew: Okay. Most of it is fantastic. It’s the way I’d actually want to see someone write. In fact, it feels more human that humans. Like I’ll give you this one from Salesforce Report. It’s just a stack of data and here’s what you guys turn it into. “Alyssa, you started this month with $1.8 million in total pipeline. You have $900,000 in closed or won revenue against your 2015 annual quota of $1 million and this 150% of what you closed by this time last year. Damn well done!” And then you go on from there.
Like I’d like a human being to actually take a moment to actually giving me great data to acknowledge that this was actually not just nice work, but damn good work. And as I went through the examples, that’s what I saw. I’m wondering what it takes to get writing to that level when it’s done by, as you’ve called it, a robot army. Let me take a moment though and just talk about my second sponsor, second and last sponsor ad then I’ll come back and talk to you about the technology and how you get it to this level.
So, my second sponsor is a company called Toptal. I’ve told people in the audience for a long time that they are a network of developers. They consider themselves like the Ivy League of developers. They try to turn away a lot of people so they could get just the perfect group of developers or as close to perfect as they can get.
Anyway, I was reading this article on Bloomberg.com. I saw this one paragraph that stood out for me. Last spring, Aaron Rubin hired a freelance coder through recruiter Toptal for about four weeks to help get ShipHero, his company, off the ground. “To find someone that talented in New York in three days was never going to happen,” Rubin said. “Every talented engineer I have has a job.”
That is one of the most amazing stories about Toptal. They have fantastic people that they pre-vetted, that they have other developers test and make sure that they really are the best ones so that when someone like Aaron Rubin, who has a fast job that needs to get done, wants to hire people, he doesn’t have to go through all the recruiting process of placing ads or working with a headhunter or trying to negotiate and trying to win somebody over and get them in. He could just call Toptal and have access to that network and hire someone to get started within a few days and know they’re one of the best in the business. That’s the thing. That’s the power of Toptal.
So, if anyone out there is building a business and needs one developer for full-time, part-time or just a project-based work or a team of developers, you don’t have to do what this article goes on to say, which is many companies are acqui-hiring so they get a team of developers all at once. You could go to Toptal and you can hire these developers.
And I actually got to meet one of the people at Toptal here in the office and I said, “Let me see the Slack group where you guys interact with your developers.” Do you guys, by the way, use Slack as a team, Robbie?
Robbie: We do.
Andrew: What do you like about Slack? I’ll come back to Toptal in a second.
Robbie: We were HipChat users before and then we moved to Slack. Initially, I was a huge fan of Slack. I’m not sure it gives us the communication efficiency that I originally hoped. I found now that I’m looking at Gmail where I maybe just looked at Gmail before. Obviously I like it for all the reasons that lots of companies that are implementing it now like it. I think the jury is still out as to whether or not it’s really helping us communicate better as a company.
Andrew: I’m still so fascinated by slack because I don’t productivity increase from it. But I think it’s because I’m not organizing right. So, I want to see how Toptal organized all their developers. They have thousands of developers.
How do they use Slack to keep them all organized? What they do is they have channels based on location where the developers are located and channels based on the language they’re working in. So that if there are developers that are working in Boston and need to talk to other Boston developers, they can go into a room or if there are developers working on iOS, they can go into a room.
If they’re stuck with one of their clients and have a question about iOS, they can go in there, ask a question and know that somebody who’s part of Toptal can go in and answer it. It gives them the environment of a giant office with lots of people, even though they’re working remotely often on their own and just working for a client. To me that was the most organized Slack I’ve seen, but usually Slack gets really noisy and complicated after a certain number of people.
Anyway, if you’re hiring developers, you should know that you can hire great developers fast, through Toptal and they have the back or thousands of other developers will have their back at Toptal. You should go check them out not just at Toptal.com, but you should go check them out at my unique URL because when you do, they are going to set you up with–actually, let me read it exactly as they have it because I want to make sure that I’m not mislead people with what they’re offering.
Here it is. Mixergy listeners will get 80 hours of Toptal developer credit when they pay for their first 80 hours in addition to a no-risk trial period of up to two weeks. Really, go check them out at Toptal.com/Mixergy.
So, we talk a lot, Robbie, at Mixergy about how the business grew. But let’s talk about the software. How do you get data and turn it into something that reads this well? What was the original version and how did you keep improving it?
Robbie: Yeah. We built it initially and up until about a year ago it was an internal platform. So, what would happen is a customer would provide us their data, then we’d configure our platform to take that data and spit out stories. We consistently got feedback that people wanted to be more in control of that process. Maybe they saw something about a story that they wanted to tweak themselves.
So, as of this past January, we actually launched the first self-service version of an NLG platform like this. So, not, it’s called Wordsmith. If you go to AutomatedInsights.com, essentially anyone can try to access this. The way I like to think of it is we’re really trying to rethink the writing process.
Lots of things have changed in the last 20 years, frankly, but the writing process, if you think about how you write today, it’s not much different. Whether it’s in a Word doc or you’re doing it in pen and paper, you’re writing one sentence after another. Now in the era of big data, we’ve created a platform that helps you take that data and structure a story in a way so that now you can automate not just one output but potentially thousands or even millions of versions of that story.
Andrew: So, if I, for example, I guess I’m looking here at your site, an ecommerce site that sells shoes in this case, I just upload all my data to your system and you guys give me not stories, what’s the phrase for that? What is it that you’d give me back if I sent you data on my ecommerce products?
Robbie: So, now the way that we work in the self-service environment is you build your own templates. So, we have existing templates that will take product descriptions and take your data and generate them. But if you want to have your own flare, your own tone and way of speaking, you can create your own template and we help you, we walk you through the process through our platform.
Andrew: It’s not like mail merge, is it?
Robbie: It’s more than that. You can kind of generically think of it that way. There’s much more logic you can build into it so that it’s not just simple plugging in numbers into pre-written sentences, there are conditional pathways you can take through the logic based on the data that you’re providing.
Andrew: I see. And the original version, was it more like mail merge?
Robbie: No. Again, we’ve kind of gone from a holy sort of developed internally services-oriented approach, where you give us your data, we’ll build the algorithms that will scan your data and find the insights and describe them automatically to now we have this self-service platform that will take your data and help you write this.
Andrew: The very first thing that you created back when you were doing it for your own sites, was it like mail merge internally with just a few–with some conditional logic in it, but mostly it was mail merge and then you kept improving it?
Robbie: It really wasn’t like mail merge at all because again, there’s a lot more analysis that went into it. Mail merge is simply, “Here’s a paragraph and I’m going to replace x with this number and y with this number.” This was, “All right, let’s scan this corpus of data that we have and figure out what exactly was interesting about this game,” and then based on that, we’re then going to pick from a set of things how we might describe it.
Andrew: I see. You’d also go back and look at how the player did in past games to compare, even back then you did that.
Robbie: Yes. Absolutely.
Andrew: Wow. Okay. I thought maybe the first version was much more MVP than that, like a real minimum viable product. No, we’re talking about sophistication and intelligence that goes beyond the game they’re covering, even back in the beginning.
Robbie: Absolutely. Ultimately, we were pitching this as an alternative to human-generated stories, right? So, the MVP, so to speak, couldn’t be much less than that. It had to kind of stand up to the sniff test if you were reading it. Maybe you realize that we’ve automated thousands of stories and you’re like, “That had to have been automated. There’s no way you could have written it.” But reading it, you wouldn’t have looked at it and said, “Absolutely, that’s automated.” You would have thought, “Yeah, that’s pretty good.”
Andrew: I have to admit to you, the reason I asked that question is because it feels like such an intimidating existing product today that I was really hoping you would say, “Andrew, we started out this small. Anyone can do it. We started out with mail merge and then we built up and we kept getting better and better.” If your first step is something that sophisticated, it feels like not anyone can do it type of story. It feels like something that only something that someone that wrote ten books and had worked at Cisco for a dozen years could put together. Am I right?
Robbie: It’s not the case that no one else could do it. But I do feel one of the reasons you haven’t heard of something like this in the past is it does require a specific set of skill sets, that is somebody that understands the writing and programming process and understands data. That’s not a common combination, right?
There are not a ton of writers that are also really good programmers and that maybe have access to data in the way that we did in order to build these things. So, I do think that we benefit from the fact that the set of skills that we have, both that I have early on and that I develop as a company are pretty unique and it’s led us to create this unique solution.
Andrew: So, then let’s go back to the story. You got Yahoo fantasy football recaps. As a result of that, people started seeing you. Other businesses started calling you and more opportunities came up. What was one of the most unique opportunities that came up at that point?
Robbie: It’s funny. Early on I would always make claims, “All right. This kind of technology works for sports because sports if formulaic.” Then we launched and we got all this feedback and I said, “Okay, really, it’s sports, finance and real estate.” Those are the three industries that NLG really world work. Then we’d get domain name registrars coming to us with ideas.
So, what I finally learned is I’m not going to put any limits on what’s possible for us to automate. We’ve heard of all things. In fact, some of the most creative things we’ve seen are internal hack-a-thons we’ve hosted where somebody recently created a “Game of Thrones” battle summary recap that I think we even have on our website that takes a data set that somebody pulled together about “Game of Thrones” and created stories out of that.
That’s why I’m so excited about our current direction, where we’re essentially going to give this platform and put it in the hands of anyone. It won’t just be us that’s doing it. Essentially anyone that has data can now develop stories. We’re going to go from generating over a billion stories a year that I think in the near future we’ll have certain customers generating a billion stories a year.
Andrew: I can see that. I’m trying to think of how we would be able to use it.
Robbie: One of the ways is personalized emails, especially if you’re in a business where you’re trying to reach out and contact people on a regular basis. Imagine instead of just sending one email, you’re essentially using mail merge to replace names and whatever. If you have some information about that user–their last logon, their access history, the last time they paid, subscription information–then you could actually create a tailored stories specifically for them. If you know who they’ve seen in the past and you have some data about your guests, you could potentially customize a story directly to that individual and drive readership.
Andrew: I could see that. Let me see what happened next. I could see that. I’m trying to think of how we would be able to integrate it into our email system. But that’s something that Michael our developer would have to deal with, not me.
Robbie: Our API is super-fast. It’s intended to work in a real time kind of environment where you pass data and we generate the content, conversational bots are a really interesting space that we’re exploring now. We’re getting a lot of interest. Just to enable–imagine playing a video game and the video game and maybe the sort of narration that’s driven from some of the games that use narrated scenes could be personalized based on what you have. So, imagine playing Madden and we can actually give you a personalized narration of the game based on how you’re performing and your history.
Andrew: Oh wow.
Robbie: So, there are a lot of possibilities that we’re just at the beginning of exploring.
Andrew: How tough was it to raise money?
Robbie: It’s varied. The seed around took me, I think, a month to raise that. It was the easiest money I ever raised. The second round was a little bit harder. The third round was easy. It just kind of depends on timing, where the industry is at, who you talk to at that time.
Andrew: You told me before we started that Jason over at Bull City Ventures was considering investing but he had a requirement and that requirement was that you leave Cisco. If you did, then he’d invest in the company. Why didn’t you just leave Cisco before he even said that?
Robbie: Cisco was essentially funding the company. That was where the initial investment was coming from was my salary. The offer that Jason made me was not enough to really move the needle. I thought if I could continue more on my own, then I could raise a significant seed round later, which I did.
Andrew: Okay. There was a period there where you ran out of money or you got close to running out of money, right?
Robbie: Well, not really.
Andrew: No? Here’s what I’m seeing from the notes. “When we were raising series A, we were running out of money and got an acquisition offer. We had two months of cash left and we would have needed four months of cash flow.”
Robbie: Yeah. Our runway was shrinking just like it does with every company. So, we had essentially our series A round sewn up, but we just so happened to have gotten an acquisition offer around the same time. So, we were in that situation where you can’t tell investors, “Just give me a couple months before I’m going to take your money.”
It’s a timing game. So, we either had to make a bet, “All right, are we going to get acquired out of the gate, but we’re going to have to get a bridge to make it to that. There’s not guarantee that would even happen. Or we continue moving forward and raise this next round of funding?” Obviously we decided to keep raising the money.
Andrew: You told our producer that you went through the acquisition process a few times. Why?
Robbie: It’s just sort of the natural course of running a business, right? We were approached by a variety of different companies over the course of our existence.
Andrew: Why did you entertain them? Why were you open to an acquisition when the business seems to keep growing?
Robbie: It just entails having a conversation. It wasn’t like we explored anything in depth. In general, I think it’s good to at least have the conversation, make sure you know the people you’re talking to because if they don’t acquire you now, maybe they’ll acquire you down the road. So, it’s more of a corporate development process that I think most companies should entertain, not to say that you should go through significant diligence process or anything like that.
Andrew: Why did you sell the business?
Robbie: So, it turned out to just be the perfect timing. The private equity firm that acquired us was extremely eager and very helpful. They believed in our vision, wanted us to continue forward. But instead of having to worry about, “Where’s the next round of funding going to come from?”
They’re a $15 billion private equity firm, Vista Equity Partners. They said, “Hey, we love what you’re doing but we want you to focus even more on product development, not the service business that you created. How about don’t worry about the funding piece, focus on building a product and we’ll be there to support you. It’s just been an amazing past 13 months.
Andrew: Okay. You said to our producer that you wouldn’t talk about revenues. Why not? Why not be open about what the revenues are at this point?
Robbie: We’re just a private company. I think it’s pretty common that most private companies don’t really share that because they don’t have to. There’s not a lot to be gained from it.
Andrew: On Mixergy, they do tend to say it. I don’t know why, but they do.
Robbie: Yeah. I would be surprised that somebody is at series B stage and they’re sharing their revenues, I don’t know.
Andrew: I’ll have to look. I think they do, but I’ll have to look. So, you’re saying once you get to series B…
Robbie: I would say once you raise money. I think it’s pretty uncommon once you’re even at a series A that you’re sharing revenue publicly.
Andrew: Do you remember the day when you sold the business when you signed the contract or when the money hit your account?
Robbie: I do.
Andrew: Which of those days do you remember once and what did it feel like?
Robbie: So, I hadn’t told my wife about it. We had gone through this process a few times. There were a few times where we had gotten really serious and then it didn’t happen. So, you kind of learn as an entrepreneur that you have the conversations, but you don’t take them that seriously, frankly, until the ink is dry. Even for this last process, my wife had told me, “I don’t want to know about it. Let me know when it’s a done deal.”
The last time I went through it, I said I’m not going to tell her until it’s a done deal. I got questions from her, “Why are you talking to your attorney so much?” I think she was a little suspicious. But I announced it to our employees and I took my wife out to dinner that night and that was the first time that she knew that we got acquired.
Andrew: And how did life change after that?
Robbie: It hasn’t changed that much. I have an 18-month old son, so it’s not like we’re able to just travel the world. Again, I’m still running Automated Insights. What I tell people is even with all the publicity we’ve gotten with the money that we raise, with the acquisition, I’m more excited about 2016 than I’ve ever been in the company’s history because of Wordsmith and all the great traction that we’re seeing with the product. So, the day to day for me and I think even if you asked our employees has not changed all that much.
Andrew: I don’t know if the word is gauche–is gauche to ask if you became a millionaire from the sale?
Robbie: Probably like the how much revenue we’re making, that’s another detail I probably wouldn’t share. I did very well. I’m very happy with where I’m at.
Andrew: And one of the things that you told–I keep coming back to the notes from our producer. You and Jeremy seem to have had a great conversation. I keep looking at these notes. You said you’re determined to have your kids be exposed to entrepreneurship, to actually know the word entrepreneurship before they hit college age. So, you did something about it. What did you do?
Robbie: I don’t really believe in regrets. If I would change one thing, back to you earlier question, I probably would have started a company much earlier. But I wanted to make sure is that my kids know that it’s an option for them. To do that, I’m not creating a college fund for them. I’m creating an after school fund. After they graduate high school, they can do one of two things with the money. They can either go to the college of their choice they get into or they can use that money to start a company and I’ll be their first investor.
Again, I don’t know if they’re going to be entrepreneurs or not, but I at least want them to know they have that option. I think college will always be there. In fact, who knows what it looks like in another 10 to 15 years. But I get asked to do lots of talks about the future of jobs and all of that. To me, if you want a future-proof job, there’s no better job than to be an entrepreneur. It requires so much creativity.
That’s the one area, even software developers. There have always been attempts to try to write the software writing process. We’re not that far a long and it’s still going to be a while before we get there. But I do it’s going to be quite a long time before we can automate what an entrepreneur does. My one piece of advice to them is if they want to have a future-proof job, they really should think about running their own company.
Andrew: I think it makes sense. I wish my parents had done that. I wish I hadn’t gone to college. I know I could have used that money in way better ways. Frankly, I didn’t get any financial aid from school. So, it was tons of money, tens of thousands of dollars that I could have used $20,000 of it to have more creative educational experiences over the four years and then started a company and maybe failed and then continued to have the same life that I have now if not better.
Andrew: All right. Finally. After having listened to Mixergy, what’s it like to go through it? What do you think?
Robbie: It’s been awesome. It’s a bit of a flashback. I remember all those early interviews that I’d watch. So, it’s been a pleasure to talk to you. I’m glad you had me on.
Andrew: All right. I’m glad you were on. Thank you so much for coming here. It was a little awkward when I was asking you about that one article that I saw, the one about running.
Robbie: The running?
Andrew: Did it feel a little awkward for you?
Robbie: A little bit. I’m used to it.
Andrew: It was for me too. That’s one of those where to this day I remember hold your pose, don’t move, don’t start to backpedal, don’t apologize for asking the question. Let it go and see what happens.
Robbie: I think for your benefit it shows that it’s not completely a puff piece where you’re just going to ask me softball questions. You’re going to ask me hard questions too and I have to be ready to answer them.
Andrew: And I thought you answered it really well. Thank you so much for being on here. Anyone else who wants to go check out the site, it’s AutomatedInsights.com. Of course, my two sponsors are HostGator.com/Mixergy and Toptal.com/Mixergy. Finally, if you have not yet subscribed to the podcast, please do. You’ll get every single interview automatically delivered to whatever device you love. Thanks so much. Bye, everyone. Bye, Robbie.