Open Adoption Software Interviews: Sentry
Sentry is an open source technology and managed service for automatically detecting and alerting users about errors in their code. Developed over the course of a decade, today Sentry is trusted by thousands of companies, including Airbnb, Dropbox, Stripe and more.
In the latest of our series of interviews with Open Adoption Software (OAS) executives (see prior interviews with Couchbase and Heptio), Sentry CEO David Cramer discusses how the company came to be, and why it chose to forego open source support in favor of a cloud-service-centric revenue model. He also touches on why he’s not afraid of competition from larger vendors, especially as Sentry’s open source community and product expertise continue to grow.
— -
JAKE FLOMENBERG: Tell us about Sentry and where the company is in its OAS evolution — Project, Product or Profit.
DAVID: We created the Sentry code a very long time ago now. The first lines of code were written 8 to 10 years ago, and they were written in open source to begin with. It was actually to help somebody solve a problem that we didn’t understand back in the day. They were like, “Well, I have these errors in my application and I want to know about them, and get emails about them or something.”
Then, a couple years later, I joined a company called Disqus, where they were using that same code. I said to myself, “Oh, this is clearly useful … but what you’re using is not very good.” That first month when I was there, we built Sentry to do two core things. We wanted to know when things were broken in production, and we wanted to know enough information about how to resolve them.
It’s different than logs, in that logs often have a ton of information but you don’t get a lot of actionability out of it. It’s just a stream of text, and you’re not sure what matters unless you spend a lot of time playing with it. Whereas we tell you, for example, “Hey, this is brand new. It’s the first time it’s ever happened today, and it’s affected these 1,000 users.”
It has been open source since day one. Even at Disqus, we continued to build it on top of the pre-existing open source license. That’s still true today with Sentry, the company. Open source is just part of our core value model. It breeds transparency. It breeds flexibility. One of our core mission values is everybody should be able to use Sentry. Open source guarantees that’s true.
We really kicked off the company a couple years ago, when we began fundraising. Prior to that, it had been bootstrapped. At the time, we were a little bit fuzzy on what we wanted to achieve. We knew open source was a big part of it, but we hadn’t really reached what you would consider a VC-viable business at the time. We were making some money, and we were profitable, and we had a lot of users; but a lot of them were using the open source product.
Over the course of the last year, we asked ourselves whether people really like open source because they can run it themselves, or just because open source is the right way to build certain kinds of technology. It was a hard decision, but we came down on the latter and decided to sell the product as a hosted service.
I think we’re somewhere between product and profit, in that we have forever made money on it, but we’re not what I would call an extreme success yet. We’re still burning a little bit of capital, but now we’re heavily investing on the growth and the profit side in parallel to product. I think we’ve sort of moved past the Product phase because we actually solve a big problem for people.
What role does community play in the development of, the adoption of, and ongoing use of your software?
For us, it’s twofold. Early on, and still today but less so, community was important in terms helping Sentry tackle new uses. For example, a big company wanted to use Sentry but it didn’t quite fit their needs. They were able to bring it in, change the code, change it to run on their existing infrastructure. Often, when those kinds of things happen, contributions from community members benefit everybody else.
More importantly for us, the community really helped us be experts in areas that we couldn’t be at the time. We were a handful of people, and our product works on every single language and every single platform. There’s quite a few of those, and they’re quite complex. All we had to do was build this interface and this API for people to send data. The community members went into the languages and used their expertise, and open sourced other components that worked with third parties, or would send us data, or other things like that.
That’s still true today. As a company, we build our own service and product. We also control the Sentry open source code base and every decision around it. But the community often caters a lot to things we can’t solve or are not willing to solve ourselves. Specifically, around how we acquire data and how we work with other companies. It turned out to be hugely beneficial, and of course, because it is open source, they feel the value of it, too.
How does Sentry think about making money in OAS, including which features to include in the community and enterprise versions of the product?
This is the ongoing dilemma for OAS companies. We made a commitment about six months ago that we’re not building a behind-the-firewall product. I think we are the only service out there that is a truly open source SaaS. It’s a BSD license. You can run all of it for free with no strings; there’s no crippleware.
That was a big decision as a company. The traditional enterprise open source models are you can go with a classic model, where you do open core and part of it’s closed source, or you do support and services. Our belief is that those aren’t necessary anymore. I think we have one customer, ever, that we charged professional services for. And we have a small handful running Sentry behind their firewalls, where we will support them.
Everybody else, which is a lot of customers, is using our SaaS service — they’re small and big companies now. I strongly believe that is a viable future for open source: If you have something that can be a service, it should be a service.
Why do you think hosting OAS as a service is such a good business model?
There’s no way our customers can do it cheaper than us, or as reliably. Then there are all the challenges of running behind-the-firewall software to begin with, such as upgrades. Yes, there are possible trade-offs on security and compliance and other stuff, but ultimately we think our expertise in hosting Sentry tips the scale in our favor. Why not just pay us to host it, instead of paying us so you can host it yourself?
How do you view the open source community as a potential source of paying customers?
Originally, we took the position that Sentry is not a free service. If you want it free, you run it yourself. We’ve since changed, and now you can start for free. We want people to understand from the get-go that they want to use the SaaS service rather than hosting Sentry themselves.
For example, we’ve got big, wealthy companies not paying us a dime, because they use open source. But does that mean that two years from now they’re still going to not pay us? What we’ve seen in the past is often the bigger the company gets, the more willing they are to pay us, and the more willing they are to consider SaaS.
Twitter, Red Hat and Spotify have explicitly chosen to convert part of their infrastructure to our SaaS offering. While they actually have expertise in running Sentry themselves, they just don’t think it’s the best solution anymore.
Why do you believe OAS is the right model for your company?
You can’t make progress if you’re constantly rebuilding every idea. Just look at all of the infrastructure that powers cloud services, for example. While Sentry has developed some of its own infrastructure, we use a SQL database, and a few other databases, that are open source. We use a web server that is open source. If we had to build all of those, we’d never be able to do what we want to do.
As we look forward, I think people will start thinking of services a lot more as tools that empower them and let them move faster. Sentry is a supporting infrastructure tool. You know if you use Sentry, you no longer have to go worry about a reactive mindset where you’re saying, “Oh! I need to be doing heavy QA on my product. I need to be digging into logs all the time.” We’ll just tell you about it, so you can continue your mission of building your business.
From an addressable market perspective, the question is whether companies would build something themselves if it didn’t already exist. We think that’s the case with Sentry, for sure. The fact that Sentry, or any technology, is open source just gives people the piece of mind that should something happen, they don’t need to stop everything and rebuild those same capabilities from scratch.
You know what you’re getting in terms of the code. Plus, open source companies often have open governance models, which means you can contribute and you can participate. And today open source software is almost always better than the proprietary version. That’s usually because there’s a larger mindshare pushing it in the right direction.
Are there any points you want to help educate the world on, on why OAS is a viable software development methodology, or any misconceptions you want to clear up?
For us, open source has been a highly defensible position. For example, Square uses Sentry open source. For somebody else to come in and compete with us there, they’d have to first unseat something that Square can run for free and that is a very good product. That’s very hard to do.
I think it’s the easiest way you can grow market share, but you have to have a good product. There’s a huge misconception that all open source technologies fall victim to bad product design because you have a lot of people with different ideas pushing their agendas. If you ask anybody using our products, they’ll tell you absolutely the opposite.
How do you perceive the threat of another company, say Amazon Web Services, deciding to host its own managed version of Sentry?
We are a little bit less worried about that than some other companies might be, because we’re not building pure infrastructure. We’re building a product that’s open source; it’s not something where we sell support and services, it’s not something where we sell consulting. A company like Amazon would have to convince customers to pay for their service instead of ours, but ours will probably always be better because it’s the only thing we’re focused on.
I think that’s more a hypothetical idea, and not something that I can foresee happening.
A more real fear might be an attitude of, “Well, if Splunk does something, or even New Relic, is that good enough?” For a lot of companies with lots of vendors already in place, it is. But the great thing is there’s constantly new companies, and they don’t care about that. That’s where we’ve seen a lot of our growth. We got into Uber and Airbnb and all these big companies when they were tiny, and we’re still there. That’s where we see the future.
The biggest thing is there is no singular solution for where your tools exist, or where your infrastructure exists. It’s a multi-faceted market and it always will be. For example, we’re not going after the market that Amazon’s competing for. Amazon’s competing with Google and Microsoft; we’ll let them compete with each other. We’ll build a version that works on all three platforms.
It’s the same idea you have with Apple versus Android. There’s tools that work on only one, but often they quickly realize they need to be adaptable. Your company can’t build your app just for iPhone, because you want the entire market and not just half.