Episode 1: Anaconda – Python Data Science Platform with Peter Wang
Peter Wang is the Co-founder and CTO at Anaconda, Inc. Anaconda is the world’s most popular Python data science platform and is the industry standard for data scientists. In this episode, Peter discusses Anaconda’s fast growing business, and how they use enterprise tools to monetize their open source software.
Michael Schwartz: For our introductory episode of Open Source Underdogs I’m sitting here with Peter Wang, CTO of Anaconda, a distribution of Python math libraries used by data scientist all over the world.
Peter could you start by telling us about your background and how Anaconda got started?
Peter Wang: My background is actually in physics and I end up going to software after college because I really like coding. And I sort of found my way to doing consulting using the Python scientific DAC if you will, that was kind of in the early and mid-2000s and as I did more work in that area I realize that there were significant commercial adoption opportunities for Python.
But the scientific stack was you know, it was still coming together at the time, right, there is some significant gaps and difficulties and how to use it. Then once I saw the Big Data wave starting to hit I recognized that enterprise data analytics would be something that Python be very good at, and people already starting to self service and use Python for.
But there were gaps, and so we started actually as Continuum analytics in order to promote the use of Python for data science and data analytics beyond just scientific computing that we want to do some innovative projects to fill some gaps and holes, things like that. And then also to build a enduring and sustaining open source software company.
Michael Schwartz: Who are the customers of Anaconda?
Peter Wang: Our customers are large medium sized companies that buy our enterprise software, so that is a machine learning and AI enablement platform. It’s a very different pile of bits the that most people in the open source world who download free Anaconda are used to thinking about and so that’s you know the Enterprise customers buy that in order to support their data scientists that are using Anaconda internally in their businesses.
Michael Schwartz: What’s the value proposition of Anaconda?
Peter Wang: When you look at what data scientists need to do there’s a lot of data they need access to, there’s a lot of compute they need to run their stuff on. They need to do a lot of different things that actually traditional Enterprise IT is not very well equipped to support them in.
So our Enterprise platform comes in, and gives IT an easy way to provide a governed and secure computer environment for data scientists. They can collaborate, build models, build notebooks, build interactive dashboards, gives data scientist an easy way to also deploy those in a production way.
Now in businesses lot of data science can put toys together, they can’t it for them to then throw it over the fence to get IT to make something that actually runs for other people’s business to use, that can take a very long time.
So the Enterprise platform eases the deployment, it gives IT a way to manage and to govern the kind of data science it’s done, gives data scientists access to packages and software internally inside the enterprise firewall. And then for people who have only lived in the open source world that may sound really strange but lot of people they live in corporate environments that are very logged down, they can’t get access to software.
So with Anaconda Enterprise we provide a way for IT to feel good about there being a vendor to provide the software and the data scientists are happy because they get the latest and greatest versions of all the packages they want.
Challenges Of Open Source
Michael Schwartz: What have been some of the challenges of using open source as part of the business model?
Peter Wang: So one of the things is that we both use open source in building our Enterprise product, we also, what Enterprise product does is makes open source available for people to use, right, so there’s things like machine learning libraries like scikit-learn that we don’t write and we provide. So there’s challenges associated with simply providing open source to Enterprise users.
Then there’s additional challenges of just using open source software in our proprietary enterprise software, right, there’s sort of two sets of challenges. With the former, with the distribution open source software I would say that most businesses are actually starting to understand that open source software is a key part of the software development stack.
For machine learning and AI applications in particular, it’s an integral part of it, you can’t do without using open source software, so Enterprise IT is starting a clue into the fact.
The challenges run into are that even though they are clued into this, there’s still procurement and legal and security and governance questionnaires that come down to us that sometimes you’re just like WTF, you look at this like you know like there’s no way we would make reps and warranties on a piece open source software that has global contributors, right, or yeah we have customers asking for unlimited indemnity of something or the other.
It’s like we don’t even write half of the software, it’s your users internally that want to use it. You know, we’re just making this available for you.
So there’s like some of those kinds of, I would say disconnects and impedance mismatches between enterprise expectations.
The biggest challenge I would say is that enterprises, they think of data analytics providers and vendors like SaaS or someone like that, that comes in with a big giant piece of shrink wrap, very expensive shrink wrap software.
That’s all just that vendors bits and so they’re used interacting with vendors in that way. So when we come in and we say we’re a vendor that’s an enablement platform, we have our pile of bits, we indemnify and support and provide you warranties, but then part of that is we provide fluid rapid access to a ton of additional capability in the open source ecosystem.
They really have a hard time, it’s really legal oftentimes that has the challenge of separating, you know, the bits that the vendor wrote and the bits that the vendor’s providing access to. They have a really hard time disambiguating those two sometimes.
Michael Schwartz: Can you talk a little bit more about what makes or what defines a platform and what does that mean exactly to you?
Peter Wang: So for me a platform is something that, it’s like a market right and so it facilitates an interaction between providers of value and consumers or users of value and it makes it from an M times N problem to a smaller like M plus N kind of problem.
So in our case, we have you know, there’s hundreds and hundreds of developers of open source libraries that want people to use their software, and that you know there’s a lot of capability there. And then there’s thousands, tens of thousands, millions of people inside businesses that want to use the software. But they don’t know which software they should use, they don’t know what software is secure to run, what versions are the most up-to-date, things like that.
Our platform essentially is a bridge between the open source innovation space and development around these machine learning libraries to the Enterprise, very governed environments, in which the software needs to run.
So creating a place where people can bring their goods so to speak, so in this case software packages and capabilities, notebooks, it could be datasets, whatever. And then where the consumers of those things can come and actually pick what they want, that’s what makes it a platform.
So we ourselves of course we built quite a bit of functionality into the platform but really letting this additional generated activity happen inside it.
It’s not merely, oh here’s a big pile of software and then we have some extension points so that people to plug-in things, it’s not a plug-in ecosystem.
It really is a platform that provides people with this bridge that lets people go both ways.
Michael Schwartz: On your website I noticed Anaconda.Org mentions a cloud offering. Is that a direction that you think is is promising? Or is the main focus the enterprise software suite?
Peter Wang: I’ve always been very interested in data science in the cloud.
Actually one of our first offerings as a company in 2012 was a cloud-based on demand and notebook computation system called Wakari. And then we eventually shut it down because it was a very hard way to make money at the time and it takes a lot of capital to actually get a software-as-a-service company going.
And so right now Anaconda Cloud as it exists is mostly a place for people to host packages and notebooks and to share those with people but we don’t actually execute those notebooks on people’s behalf.
In the future I cannot make any promises about our long term product roadmap, but I do feel like cloud-based execution of notebooks and of models, things like that, it’s a natural need that emerges and you know we may do something in that space.
Michael Schwartz: What are the most promising areas that you’re investing in or looking to innovate in, in the future?
Peter Wang: Well, you know, the space moves so quickly and it’s evolving quite a bit so we really are looking at I think for most Enterprise software companies would be, because of a very short time windows, or innovation sort of new, de novo innovation work.
We’re mostly right now investing in the existing projects that we have. We’re also investing, there’s a new project we just released for data access, to really solve that data reproducibility and data sharing problem; lightweight data catalog and data access library called intake that we are very excited about.
But moving forward we will be looking at doing, you know more things with regard to model management and helping people share models in a more reproducible and seamless way. And you know beyond that it really depends on how the business evolves.
Michael Schwartz: So of those different areas where you could generate revenue, which one is generating the most revenue as a percentage? Is it license?
Peter Wang: We’re primarily, the vast majority of our revenue comes from software licenses.
That’s wasn’t always the case, we actually made a pretty dramatic change over the last year. Flipped almost completely 180° around from being mostly consulting with some small amount of, relatively small amount of software revenue to the other way around. Where it’s mostly software revenue and a much smaller amount of consulting, training, services, things like that.
Michael Schwartz: Can you share any details about how you went about licensing?
Peter Wang: We start as a company to support and promote the growth of Python for data science.
So we’re something that’s very much of a community, right, it’s not like Travis and I sat down and said: Oh, aha, we have this genius idea for some crazy new cool technology, we build it and then we’ve vend it.
It’s, we’re of this community, we built more innovations to help, amplify the the efforts of the community, we’re always of the Python data science community. So, the ethos of that community is to license everything permissibly so MIT, BSD licensed, LGPL sometimes.
And so it wasn’t really feasible to do any kind of open core business model around that. Some communities, R as different right, the R Data Science community, R is all GPL based. And so classically there have been two companies, one got acquired by Microsoft, that were heavily doing work in R around open core model.
In the Python world I don’t feel like the community has an appetite for open core business models. And so we very quickly came to understand either A) we have to sell something directly to the community so our users and they’re our customers.
Or, we figure out some kind of enterprise software that we sell, that is a proxy need, fills the proxy need or as a proxy to the open source of free stuff that we give away.
So the business model we converged on was essentially that. We give away this stuff you know, you give away burgers and you sell the Coca-Cola or something like that, right.
It’s not, you get a free burger but if you want two patties, or if you want the burger with cheese then you pay extra, right. That’s not the model we have, so Anaconda, all the work I do are around Anaconda and the packages the distribution of the testing with all that stuff give away for free.
We will give it away for free forever, because we know that when businesses start using this stuff more, and more, and more, it will drive proximate need for for other kinds of things. And the first significant proxy need that showed up was the need for a management platform or some way of managing and governing the use of open source data science packages inside the business then it became managing the collaboration around Jupiter notebooks and things like that.
And now lately it’s evolved much more into deploying models and looking how models are running, you know, the actual models in production is a much larger part of what’s driving business pain and so those are prominent, all three of those things I mentioned the repository package governance, the collaboration and then the management of models of production.
All three of them are core components of our platform offering in our Enterprise product.
Michael Schwartz: Would it be fair to describe that as tools around the open source?
Peter Wang: I don’t think it’s tools as much as it’s, I mean the nature of the software itself is to provide an infrastructure for you know spinning up servers, deploying things. It’s not, they’re not really tools as such.
I mean actually the end user experience, when they’re sitting there, like if you’re a business user, as a customer that’s bought Anaconda Enterprise, you’re using the browser login to our Enterprise platform and you get the the Jupiter notebook experience there.
You can also use Anaconda as you always used it on your laptop or something, and the Conda package manager will connect your internal package repository. So it’s not like you’re getting extra tools, you have the same tools you’ve always had from an actual data science day-to-day perspective.
But using them in an Enterprise environment is much easier for you now.
Michael Schwartz: Maybe switching gears a little bit to the marketing side. Does Anaconda just sell itself? Because the community has a strong underlying, you know science Python community? Or what are some of the channels that you use to get the software out there?
Peter Wang: Well, so the enterprise software – so very few things in the world sell themselves. And the reason is two-fold, one if there’s not much demand for what you’re doing, then you have to go and tell everyone about what you’re doing.
If there is a lot of demands what you’re doing you’re going to competitors that race in and provide messaging around why their product is better than yours, right.
So in both cases if there is or there isn’t a demand and you’ve got to be doing some kind of selling.
Now for us the the machine learning platform, data science platform, that term is now a term of trade, right, in the business so we have people at the CIO level, CTO level, that are like, okay we need a data science platform, right. They know what that is, they know it’s something they need to do to support their data scientist.
Now it does help us that those data scientist will tend to already be Anaconda users, we do have access to their mail list and social media and whatever else. We have a pretty visible, pretty well-known as a brand, so the brand everything is top-notch. We’re known for our investments, interaction with the open source community.
So all of that stuff helps us in the actual sales process. But in the end of the day we do have to do marketing, we do have to inform people about what the features are, why they would buy.
It’s like any other piece of software you’re trying to sell, there’s real activity you have to do there.
Michael Schwartz: So without going into actual prices, but maybe just around pricing philosophy – were you able to quickly figure out what your price should be? Have you changed a lot over time?
Peter Wang: It’s changed a lot of over time. Depending on who you’re selling to, it really, you know, we’ve always been in the enterprise sales sort of mode, for our support training, things like that, as well for our products.
So from the beginning we had this understanding that enterprise sales is a different kind of animal than you know consumer sales, setting prices and all those things.
There’s a whole process to sales and sales strategy, and neither my co-founder nor I would claim to be experts in that, although we have some perspective and some background in it. So it’s important to actually work with sales people who kind of know what they were doing. And we’ve been evolving overtime, we’ve been learning overtime, how that whole thing works.
Michael Schwartz: Maybe we can talk a little bit about partnerships. Do you have a strong partner network or who are the partners that you’ve built up over time that you think have been important to you, getting out there and getting to market?
Peter Wang: So Partnerships, we’ve always worked with some other companies, but in terms of actually driving to sales success, that’s been a fairly target activity over only the last couple of years. Because using partners as a channel – their sales people have to know how to sell your product, right.
They have to know how to close a deal and include your product in the pricing so you have to have really, really clear definition of what is the product, how to talk about it as an integrated part of the partner offering, all those different things. And I would say that’s again, that’s something we’ve only been doing relatively recently.
Prior to that our partnerships really consisted of technology partnerships and sort of marketing and exposure kinds of things.
I mean the data science field itself is also relatively new, at least in terms of enterprise adoption of this stuff. So we actually partner with, there’s sort of two classes of folks: One of them are hardware vendors, actually. So Intel and Nvidia Microsoft, the cloud vendors like Microsoft, Google, Amazon all that.
But the hardware vendors AMD, Intel, Nvidia and gosh I probably could use some others there, IBM obviously, they have hardware. So with those we are partnered on technology as well as some go-to-market activities.
So we have a compiler called Numba that have distributing computing system Dask. Those are very, very exciting for these hardware partners because they want to showcase their chips running at scale, they want to showcase the kind of performance wins.
Now if they require end-users to have to write extremely low-level code in order to get those wins it’s a non-starter, right. So the idea, it’s not just performance but it’s performance as accessible.
So using our tools and technologies they can get your average data scientist with not some like C++ optimization nerd. Your average data scientist can use some of our libraries and then be running these things you know and really showcase their performance.
That’s some harder partnerships, then the cloud partnerships as well as some others. They’re really around the data science tools, around notebooks or visualization. Some cases they’ve been with our enterprise platform as well so it really depends.
So like with Microsoft for instance, we are partnered multiple ways.
So our Anaconda distribution is installed on Azure, it’s available as the Python runtime inside sequel server itself which is pretty cool. It’s something you can install as part of the default Visual Studio installer if you say I’m data scientist, and you select that profile, it will install Anaconda for you. So that’s a really cool partnership.
And then on the flip side of it, we’ve included the VS code editor inside the Anaconda distribution installer so that that’s one of the editors and ID’s that’s available for end-user.
So there’s multiple aspects of Microsoft relationship, but those are probably the most significant ones. But these partnerships are very exciting for us, they’re still in the early stages, you know, right now we would love to see driving more revenue through those.
We are investing in those relationships because we believe we will be able to drive a lot more revenue through them.
Michael Schwartz: What about non-, let’s say business partnerships, any organizations, or foundations, or other communities that you found that had been really useful for you to promote the product?
Peter Wang: Oh yeah, well so we primarily stay in the data science space, right, in the Python ecosystem.
We’ve been growing our inclusion of the R ecosystem as well because many data scientist use both Python and R. And so we’re trying to find better ways to play nicely and be a good participant in that ecosystem.
But on the Python side we’re actually I would say pretty unique in that we started the company in January of 2012 and then by March we started the effort to create a non-profit to support and sustain several open source projects in the Python ecosystem.
That nonprofit is called NumFOCUS and that’s been a really good partnership of ours.
When I started the PyData sort of global community sort of effort, we attached that essentially to NumFOCUS as well. So now NumFOCUS runs as a center point of coordination for PyData meetups, PyData conferences all around the world.
It’s really grown, just really blown up. So that’s a constant and really great partnership with that foundation. That’s, I would say probably the primary one at this point.
There’s a lot of community events and things like that, that we do as well just to be good stewards you know, and show up. But those drive a lot of awareness for us.
One of the challenges from a business perspective is, if you say too close, just mingling with your friends, you’re not going to sell much because, in our case our friends are all users of the free stuff, right and they’re not really the people who are signing the procurement checks to go and buy the enterprise software. So yeah, that’s one of the dynamics there.
Michael Schwartz: How do you stay current on what’s a very hard domain?
Peter Wang: Yeah it’s tough. I mean I don’t… I wouldn’t go so far as to say I am current. I am current on some things and there’s other things that I’m not as current on.
We have just a very, very broad range of technical expertise under one roof here. So we have people working on compilers, people working on distributed computing, people working to visualization.
Our consultants and trainers who go out in the field and talk to customers they get exposure to a wide variety of different kinds of modeling problems, optimization problems, what’s the state-of-the-art in machine learning, in deep learning, all these things.
And I just have the great privilege of being able to talk to all these people and kind of glean from their learning as much as I can. Now I do some deep dives myself on some things, I have more time now than I’ve had in awhile. That’s part of it.
But, most recently I’ve actually spent the last I would say 3 or 4 months really nerding out on, believe it or not, the physics and the anthropology and the human ecology around open source and open source ecosystems, communities, and how we do sustainable open source; how we continue innovation while maintaining stable software.
The nature of software businesses as a whole, especially as it pertains to open source. So these are areas that I’ve been actually thinking about quite a lot and I found that my physics metaphors or my physics intuition has helped me a lot in thinking about the System Dynamics of that human ecology.
Open V. Commercial
Michael Schwartz: What are some of the challenges you think of starting an open source business or, I should say a business that uses open source as part of their model, versus straight commercial software company?
Peter Wang: I have a snarky answer – which is that the challenges that face an open source company are no, really, theoretically, no different than what face in a commercial company.
In practicality they’re different. Because a commercial company one tends to have a fairly reasonable expectation that the investors of the company and their shareholders are the ones that the company’s optimizing for.
If a company does a great job that’s measured very quantifiably in returns to shareholders. That’s the American model of capitalism, that’s what a company does. So if you screw up the company. The only people you’re going to piss off are your shareholders and your investors. Probably your employees too.
But in the open source side, open source software company there’s actually much more explicit or intentional vision that matters.
Most companies, they have a lot of like vision, mission, blah blah blah. For the most part once you get to a certain size employees are all like, yeah this is to make money. Unfortunately, you know, that’s just the reality of it. Now some businesses have highfalutin goals, most don’t actually even attempt to play to those.
But it open source companies the founders and the people who are stakeholders in it, there’s a broader set of stakeholders.
So who you piss off is a much bigger range than just the shareholders and employees. And so I would say that’s the thing – is that failure means different things.
And for open source companies usually – it’s sort of like you split the whole world of living things and prokaryotes and eukaryotes, right. Like in the world of companies, of open source companies, I would say you can cleave it in to those that do open source as a means to an end; and those for whom the open source bodies part of the ends itself.
And how open source embodies an ends – now that itself can vary. For some people there’s a religious belief that all software should be free and by golly, we’re going to build this thing to be free.
And those companies sometimes can struggle because there’s this reluctance to charge for software because you believe software is free, and so they are limited to a particular kind of growth curve because the corporate financing is available to them is of a particular stripe because they’re consulting or services company.
The ways that they can fuel innovation is also limited because they cannot go and sell a product as easily as someone else could.
So those for whom just open source because, it’s open source, those companies who have that ends, you know they’re kind of constraint to one part of the landscape.
Then there’s other companies for whom open source is an ends and a means and part of it is that it embodies some technical vision that the founders cared about. That technical vision is either only achievable through open source or gathered momentum and gathered user base, and a way of doing things, a technical perspective, that is now at this point so steeped – like the open source is part of its DNA – that it would seem like a violation and a rejection of everything that the soul of the company is, to walk away from that. That’s a different kind of ends, right.
And then there’s others like, I think for my perspective we’re little bit of that, but then also, from my perspective there is a fairly opinionated view for us that open source is, it’s a statement that were not going to lock people out by closing the software.
So it’s almost like a statement of – we may fight as a business and compete in a landscape that’s rich and filled with many kinds of competitors but one thing we’re not going to do, is we’re not going to use closing down access to software as a means of either charging rent or fighting dirty, or a way to charge rent so we don’t have to innovate anymore.
So in that sense being open source company means that you are essentially committed, there’s a covenant to your customers, to your users, to the ecosystem, to your employees. There’s a covenant that you are going to innovate, because it’s so much harder to be just a rent seeking monopolist as an open source company just around the software itself.
So, it has a bit farther range than you’ve expected from the answer, but…
Michael Schwartz: Actually, you brought up one thing I wanted to ask you. Does Anaconda actually have a lot of competitors?
Peter Wang: There are some things that we do that are unique in the world. So on certain aspects of the technology and some of our product features I would say we are very unique.
Now there are certain other things that we do, in particular the software that we sell that, you know we think we’re the best but there may be other companies who think they’re our competitors, right. And so I think that from that perspective, and certainly it’s up to the customers and not us, to say.
So there are times when customers will you know bank us off on other things and so we definitely have competitors in the enterprise space.
Michael Schwartz: And you mentioned open source as requiring you to innovate. But has it also helped you to innovate?
Peter Wang: Software is a collaborative creative activity. And then some, for some problems, for some kinds of projects, the kind of open collaboration that open source represents are the most effective way to harness collective intelligence and get something really useful out.
In other cases open source is actually, the open source development methodology if you will, a way of engagement. It’s actually not necessarily the most effective to harnessing innovation.
And the dirty truth is if you look at the way the open source project actually roll. For all you know, all the Kumbaya aside, you know how they actually roll.
The most successful ones are launched by one, two, maybe three person founding team. I mean 3 is pretty rare, usually it’s one or two people, usually one actually, and they like a blaze of glory, they dropped this code in the world. And other people start glomming on and that initial nucleation site around the initial feature set, the collaborative dynamics of that. That really sets the future, like the seals the fate of the project a lot of times.
But usually open source projects they launch when it is a blaze of glory sort of innovative leap from one person’s brain.
And even now like, if you look at there’s a really sad or interesting way of looking at this, like the maintainers of some of the most critical open source projects in Python that are used daily by millions and millions of people that back billions of dollars of commercial infrastructure activity whether it’s power grid whether it’s keeping satellites up, whatever… I could fit all those maintainers in my minivan.
Now one can say, wow witness the amazing leveraging power of open source but you could also look at that say wow that’s really sad, we’re under supporting and investing these projects. So I am a big believer in open source, I don’t want this to sound like I’m like poo-pooing it, I’m a big believer in it but I’m also a realist.
And I think that open source in the early days had a lot more of the, you know, the OSI, FSF kind of days and all that.
I mean a stallman, coming out and planting a stake in the ground and saying we’re doing open source, the commercial Unix people can go you know, they can go pound sand, and we’re doing open source. That was important as a stake in the ground at the time, and over the 90’s as, you know, the Linux folks tried to educate everyone about, hey open source, okay here’s what it is, here’s free software, here’s open source.
Now, there’s a very different dynamic, as businesses are like, yeah we don’t care. Like the golden era of software is over, its services, it’s machine learning enabled services. These are where the top end of the value chain are.
There’s a real squishing compression dynamic on software that’s happening, specifically just on open source software. And if we’re not cognizant of that dynamic, if we don’t step up as fans of open source software who love the collaboration dynamics of the community, if we don’t step up and defend that and say actually, if you’re going to rely on this infrastructure you need to be paying for the maintenance of it and not rely on volunteer labor.
You know I think that’s a conversation that the free software world needs to have with the commercial world.
Michael Schwartz: We found sometimes that being open source actually is a hindrance to us. Do you ever feel that, like that open source makes your life more difficult?
Peter Wang: Yeah. For venture capital fundraising, open source is a huge liability, absolutely. Because VC’s don’t understand open source. The big ones, I would say.
Like, smaller funds that have maybe a little bit more, that can take a bit more of an opinion, be a bit more of a gambler on business models, things like that. Maybe, certainly there’s angels, you know there are angel funds that they believe in the technology or something like that.
But the but the vast majority of VC’s, they’re there to really pattern match against known business models, known growth curves, you know things like that.
So if your business model is working to charge the software they’re going to ask, well how much value is the software? What do we believe the predictable growth curve looks like for this.
If your business model is we’re going to get users, they don’t care if give away the software, right.
So it’s all about, when you talk to the VC’s, how do you present them with articulation of your long-term value or not long-term value, but the returns you generate on their investment. That’s often the only conversation they really care to have.
So if they’re investing in software and you just giving it all away it literally looks like you’re just giving away their money.
In fact – if you were to say we’re going to take the money, buy iPhones and give it away to everyone, they would at least understand that a little better, because it’s something tangible in an iPhone right, they can say well we’re going to get on the other side of it.
You know, or you’re going to give away anything like MoviePass or whatever, I wanna give away free movie tickets to people and lose money on movie tickets. VC’s are not afraid to lose money, they want understand what that money is buying them. Right.
If it’s buying something else, some other number – eyeballs, or users, or something else, which you can then show how that converts in the long-term revenue, they are happy for you to give away the money.
But if it’s simply we’re funding software development and you’re giving it all the way, and you can’t show conversion that they believe in to revenue then it’s not going to fly.
VC Alignment With FOSS
Michael Schwartz: Were you ever concerned that perhaps the investment would come with strings that would make you to give up some of the mission or culture of open source?
Peter Wang: As we were talking to investors that was certainly one of the concerns.
I mean that with the reason we picked the investors we did was because we felt that they were mission-aligned and that was that’s a real luxury, we were very lucky in that.
And we worked pretty hard of the fundraising thing, especially as first time founders, not knowing what we’re doing. We talked to a lot of people. Probably screwed up a lot of meetings but, we ended up with some folks that actually really understood, not only the commercial potential of investing this, in the space in this company but also the mission, so we’ve been pretty lucky with that.
Again I would say that it’s rare to find that.
Michael Schwartz: What advice would you give someone who wants to start an open source software business today?
Peter Wang: Ultimately the act of going into business, it’s not separable question from why are you going to business at all, right?
Like if you’re going to start a business it’s because you either A) want to fund some activity; B) you believe you have some unique thing that you can sell to the world or some unique service you can provide to the world.
There’s actually very few, a small set of valid reasons for starting a business, in my view. And attached to each of those reasons, if you do open source as core activity the question you have to ask yourself is, am I doing this as a means to an end or am I doing this, is this one of the main things I’m starting business to do. If it’s the latter which I think is really where the meat of your question is, if someone says I want to do open source software but I also want to make money doing it somehow.
There’s many people who can, by building open source software, use it essentially as a marketing or brand awareness tool and then they can freelance, they can build a very healthy consultancy from getting their name out there and being known. That shouldn’t be downplayed as a valuable thing, and you can build a reasonably good small business around that.
But that’s not going to get you into unicorn billion dollar valuation territory, it just won’t, because to get to that level you have to somehow get to a certain size of revenue.
How do you take down certain amount of revenue, either you produce something like a piece of software that’s extremely valuable that you sell a lot of. Or you have a ton of people doing a lot of work.
Scaling up a consultancy is highly non-trivial, most software geeks who care a lot about technology are not wired to scale-up, a lot of meatware if you will. Once you get to a certain number of headcount you have got to figure out as an entrepreneur and as a technical entrepreneur how to let go of parts of it and let someone else actually help you grow that and that will change the culture and that will probably change part of the mission too.
So if you want to scale the certain revenue sizes you’ve got to do that either the basis of providing a really valuable, scalable service, or software.
I would just say understand the dynamics. Don’t blindly rush in with a whole bunch of optimism and then just curse your fate like, there’s a real dynamic here.
Investors coming in to help put more gas in your gas tank. They want to understand what roads you’re on, how far you’re going to go. If you can’t communicate to investors on the basis that they understand, about what your business model revenue models are, then you have no business asking for them for their money. You know, don’t get mad at them, right.
So, I mean it sucks in a sense that we live in this world where people do not think more about investing capital in socially-aware activities or in a generative sort of effective labor. That’s a broader conversation outside the scope of this interview probably, but that’s a systemic thing which I hope will resolve in 20-30 years time. But at this point of time it’s just the reality of the investment you know field, what it looks like.
Michael Schwartz: What were some of the businesses that you looked to as you were looking to scale Anaconda as businesses that you could model after?
Peter Wang: We recognized that actually because we were not doing open core that took a lot of our peer crowd of open source companies off the table, right. So like, MongoDB is an example.
Michael Schwartz: Could you just define open core real quick?
Peter Wang: It’s where the core of the software is open source and then to use, to put more data through it, or to run it on certain kinds of machines with more cores, or to do blah blah blah you know additional features and additional whatever, charge money for it.
Michael Schwartz: So like an enterprise version and a community version.
Peter Wang: Yeah, community version only supports 100 users, enterprise is unlimited. Community version can only run in the cloud, enterprise version can run on prem, like things like that, open core yeah, it’s best way to define it I guess.
Michael Schwartz: I see.
Peter Wang: The core of it is still open source, you know it’s still legit open source.
Dual licensing I think also falls under that. Right, so like you basically, the core is open source GPL and then commercial license. That’s that dual licensing is also something, something that people do.
So you can use it for free for to a non-commercial setting but as soon as you take it in the house in a commercial environment your lawyers and their total like allergy to GPL will cause you to go and give IT a call to go buy the commercial license of the software so you don’t have to risk, you know, virally contaminating your internal software with GPL. Those are the kinds of models, and so we felt like many things in the data analytic space were in that kind of model.
And certainly databases have this a bunch. We didn’t really, we didn’t see those as comparables. And when you look at platform software there’s RedHat. Really not that much else to compare to.
Now we look at some of the Java framework companies that manage to sell, some got acquired by RedHat, some got acquired by Oracle. For the most part I mean once we figured out that we’re just going to sell enterprise software, you know by seats, sometimes by notes to customers.
It actually became pretty straight-forward how to think about the business, so in that case it was really drive a lot of usage of the open source stuff and that’s going to drive a smaller but correlated usage and demand of the enterprise stuff, that was it.
Michael Schwartz: What did I miss? Or is there anything else you want to add?
Peter Wang: I would say that right now I sort of just glanced by this comment earlier but I really do think that right now the world of software is in transition, so people who want to start open source based software companies now should think long and hard about what is the value chain for software actually look like.
It may be better just to do a software-as-a-service kind of thing and then you get to own the customer relationship, you own much more the value chain, etc.
Now there’s downsides to that as well but that’s something to think about because I think that the, and Jason talks about software eating the world, and that may be true, but I think less and less of what it eats has really high caloric value right; more more the high calorie stuff is going to other kinds of things.
So I would encourage you think strong about what it means to be a software business in the modern-day, especially as things like Amazon, Google, Microsoft eat the world of cloud services.
The second thing is understand what it is about open source that you love, that’s intrinsic to your mission and figure out, be very, brutally honest with yourself about what of that you want to preserve and what are the right mechanisms for preserving that.
Don’t assume it’s just a money problem don’t assume it’s just, oh if I got out from this like horrible soul-crushing corporate job I could do open source all the time.
There’s always going to be a whole bunch of yak shaving from a business and management perspective no matter what you do. Understand why you’re taking that particular road if you’re gonna take that road.
And the third thing is open source doesn’t exist in vacuum. It exists in a human ecology of users, contributors, competitors, and evolving technology landscape.
You need to understand to think strongly about whether or not the thing you’re building has long-term sustaining ecology value and if it does then it’s worth investing. Otherwise you might want to think about how to pull various other pieces of your ecosystem together to something that’s more valuable as an agglomerate.
Michael Schwartz: Peter Wang, Founder of Anaconda, thank you so much for sharing your wisdom with us. Best of luck.
Peter Wang: Thank you very much, thank you.
Michael Schwartz: Well that’s it for this first edition of Open Source Underdogs. Special thanks to the Linux Journal for co-sponsoring this podcast.
To the All Things Open conference we’re launching on October 21st.
Music from Broke for Free, Chris Zabriskie and Lee Rosevere.
Production assistance from Natalie Lowe. Operational support from William Lowe. And from the staff at Anaconda.
Next week we’ll talk to Netgate, a rare bootstrapped open source company who’s also based in my hometown of Austin, Texas.
Until then, thanks for listening.