New Relic CEO on how to build a cloud company and help customers play offense with software
Lew Cirne has lived the spectrum of experiences in Silicon Valley: engineer at Apple; founder and CEO of Wily Software, which CA Technologies acquired in 2006; and, now, founder and (still-coding) CEO of SaaS vendor New Relic, which made its NYSE debut last December.
In this interview, Cirne discusses the lessons he has learned along the way and how they have helped make New Relic a success. He also shares his views how smart IT strategies will help businesses compete in the coming decades.
SCALE: How are things atNew Relic, one year into life as a public company?
LEW CIRNE: We’re very thrilled by how things are going and how the market is developing largely the way we saw it developing. We’ve had good execution on our part and our business is very healthy — we grew about 69 percent last quarter and have enjoyed four quarters now as a public company.
What’s driving all the growth is the fact that every business in every vertical is moving so rapidly to software projects to drive growth, that they suddenly realize how important it is to measure everything about their software.
Now, what’s changed is that businesses need to show three things all in one platform: the application performance, the end-user experience and the business outcome — how many units am I selling or how many people signed up today or how many articles were viewed. They all need to be measured by the same platform because every one of those three things can impact the other. A poor-performing part of your software can impact the customer experience, which can impact the number of widgets you sell today.
The other thing about it is that if you want to see a total timeline of users’ experiences, you cannot aggregate or throw away any data. You’ve got to collect it all, so this is really the ultimate big data problem. For every action that a user takes on a website, it is not just one record you need to collect. You may need to collect 10, 50 or 100 records, because behind an action like purchasing an item, the software is doing a ton of related things.
Any of these, if it fails, impacts the user. If someone purchases an item, it could trigger a transaction of flows through 10 different services. One of those services could be slow or failed — and all of a sudden the user doesn’t buy an item.
“If you write a query that does not hit an index, then it can melt the database or take forever or smoke comes out of the machine or something.”
And you’re solving for this how?
We created a database service called NRDB that previously fueled our Insights product, and now we are moving all of our products onto this database. And that means that our products are going to be lossless and capable of real-time analytics in any kind of dimension. Most databases ask you to index, so that if you tell the database what your query is going to look like, then it will give you a fact answer. But if you write a query that does not hit an index, then it can melt the database or take forever or smoke comes out of the machine or something.
That’s because prior to today’s level of scale, you couldn’t brute-force yourself through all of the data to do an ad-hoc query that doesn’t have an index in advance. But indexing is very limiting, it takes away flexibility. If you know what all the important questions are, then you’re too rigid in your thinking. And if a business comes up with a new way to segment data, but it has a two-trillion-row database, that re-indexing may never happen.
NRDB is an analytic database in the cloud that has no indexing. The way it works is, our cluster is so wide that when we execute a table scan we can literally run through a billion of events in far less than a second, inspect them all and then answer, “Of those billion, say, pageloads, how many people did X or what was the funnel of X to Y?” Pick your business question.
“If it feels like you’re pushing rope and trying to get people to try your stuff but it’s not turning into sales enough, then go back to product.”
Managing growth from idea to IPO
We last spoke around 2011, and New Relic had about 3 salespeople. When did the company hit critical mass and grow into what it is today?
For context, two years ago we hired the president of global sales from Salesforce.com. She is now our president and she manages more than three salespeople, to say the least. I guess we probably hit critical mass if we were able to convince Hilarie Koplow-McAdams that this could be the next Salesforce if things wind up well for us.
It started becoming critical mass when we were managing to get a few customers to sign up for a million dollars a year of New Relic over the phone. It was a small number of companies, but if you can close a million-dollar customer with an inside salesforce, then you’re obviously delivering something of immense value. So we decided to invest in the field and seek out more of those opportunities and accelerate that business, and that has done well for us.
If I’m a budding entrepreneur and I’m launching a cloud service, is there a natural growth curve I can expect at this point?
I think probably one of the best thought leaders on this is Jason Lemkin; his SaaStr blog is really thoughtful on how you think about milestones, revenue targets and growing your company. For me personally, qualitatively, it ought to feel like you’ve built something where you are being pulled by the market rather than you having to convince the market to give it a try and to buy it.
We saw examples of this even in the very early days of New Relic. I remember we did a multi-hundred-thousand-dollar-a-year deal in Argentina on a very short sale cycle, less than a month. Nobody spoke English on their side, nobody spoke Spanish on our side, but somehow we closed the deal. That’s where the product is meeting a real market need and that’s the sign that you’ve got something going.
Don’t try to push rope and say, “Oh, if I only only get some more leads, then we will start seeing the business.” If it feels like you’re pushing rope and trying to get people to try your stuff but it’s not turning into sales enough, then go back to product. Check on why aren’t people falling in love with your product.
“We’re a growth company and I think investors really love the growth story. . . . And every quarter we make meaningful progress on the operating margin side, so we are not a growth-at-all-costs company.”
How involved are you at this point with the product vision at New Relic?
The chief designer of New Relic sits right outside my office. I just had an hour with him earlier this afternoon, where we went very deep on new product development. We do that very regularly.
If somebody like Hilarie is leading our go-to-market efforts, she doesn’t need any help from me on how to run the business side. So while I remain engaged and interested in that, I feel like my best energies are devoted to making sure we deliver awesome product technology that makes her job easier.
It seems like public-market investors are buying into New Relic, despite its cloud business model. Was that the case when you were out promoting the IPO a year or so ago?
We’re a growth company and I think investors really love the growth story. There are very few companies growing more than 50 percent in technology right now. And every quarter we make meaningful progress on the operating margin side, so we are not a growth-at-all-costs company.
The common question I get asked is, “What’s changed for you personally as CEO since the company went public?” And the answer I give is, “A lot less than I expected.”
I was expecting to be driven by what you hear drives public companies — the 90-day cadence and Wall Street and things like that. You know what, when I meet with Wall Street investors, they often ask me, “Lew, when’s your next coding retreat?” I just had one two weeks ago.
The investors that like New Relic like the fact that it is being led by a person who still likes to innovate and drive innovation through the company. As long as I’m not the ultimate person responsible for making sure that we are executing and operating well — and we continue to execute and operate well — I think it’s going to be fine.
“I feel like it’s the right time for us because this cloud database we built couldn’t have been possible prior to today. The cost of infrastructure, the notion of deploying suites of servers with hundreds of gigabytes of RAM and nothing but SSDs, was just too much.”
Making Rupert Murdoch get up and jump
What are the biggest trends that have driven product development since New Relic began?
I think a big thing is that when companies build new digital products today, they need to be able to handle massive and unpredictable scale — that’s one of the reasons why cloud has got so much traction. And these assets have a lot of complexity to them. They recognize that they need a single pane of glass measuring everything. The end customer experience, measuring exactly what is going on in the mobile device. Measuring what’s going on in the app server and the all the various services that are running on the server side.
Our more advanced customers also want to measure what’s going on in the business at the same time. If I’m an e-commerce store, why should I wait until tomorrow to see what sales are doing right now? In fact, why can’t I slice and dice what’s going on in the e-commerce site in real time?
I remember showing Rupert Murdoch a real-time dashboard of what was happening on the Wall Street Journal website and it blew his mind. He leapt out of his chair, walked straight up to the screen and starting asking questions about his business. It was all because we were watching his software, and his software was his digital property. That’s a big part of the News Corp. business, they are digital properties.
I don’t want to suggest it was luck, but if software is eating the world, New Relic struck at the right place and the right time.
I feel like it’s the right time for us because this cloud database we built couldn’t have been possible prior to today. The cost of infrastructure, the notion of deploying suites of servers with hundreds of gigabytes of RAM and nothing but SSDs, was just too much. And it has to be an enormous multi-tenant architecture so that the entire fleet of servers could be at the disposal of any one of our customers, because basically the service bursts at query time to provide all the capacity that is needed to process that query.
That only makes sense if you’re in a multi-tenant environment, because otherwise you don’t have enough people write queries and your cluster is 99 percent idle, so you spend too much money to service a single customer.
Let me give you some numbers on how this stuff works. I am looking at a dashboard of our database right now, and it’s reporting an insert rate of 2.5 million events per second. And while it’s inserting 2.5 million events per second, our customers are doing real-time queries off of that database in real time, and those queries are reading 1.5 billion events per second. The 99th percentile query response time is 230 milliseconds right now. That’s just mind blowing, and remember everyone of those queries is a table scan.
What that means is people can just do ad-hoc queries for what in the past used to be a one-week report or 4-hour batch job. “Look at the last year of pageviews and tell me how many people looked at the blue ski jacket and then ended up buying the red mittens,” or whatever. That could be a consulting safari for a week, but now it’s a 1-second query.
Keeping up with developers
Have application containers affected how New Relic designs products?
We provide a first-class experience for Docker container monitoring and seeing the connection between the application, the containers they’re running in, and the underlying operating system and server that the container is running on — all of those Russian dolls of containerization.
If the same instance of a physical or virtual machine is running three containers and one of those containers gets very very sick, what does it do to the other two containers running in the same host? That’s the kind of problems New Relic solves now.
What is the approach to keeping up with the different technologies that customers might be using? Does it make a difference when users switch between one cloud and another or one program language to another, or to microservices?
One of the things that is a challenge and that creates competitive advantage for us is the pace of change on how people are building modern software. What technology stacks they are using, whether they are running in cloud. That pace of change is pretty torrid. It seems like every six months there is a new Javascript framework for the front end. Ember is so 2010, Angular was 2014 and the new one is the one out of Facebook.
You get my point. We have a fast deployment model and we try to service the latest technology just as rapidly as we can develop support for it, then all our customers get that benefit.
“The sins [of Wily] were that the product was too hard to use and our Band-Aid for hard-to-use products was more people — more support people, more services people, more pre-sales people.”
When you started the company did you envision growing from a single application into this whole platform?
It definitely evolves over time as you learn more from your customers and you learn more about your business and the market. I started off just thinking, “Here are the sins I don’t want to commit again if I start another company.” The sins were that the product was too hard to use and our Band-Aid for hard-to-use products was more people — more support people, more services people, more pre-sales people. We don’t want to commit the sin of a hard-to-use product because you end up spending a lot more money on people to help the customer use it, and the customer prefers an easier to use product.
You also don’t want to commit the sin of getting into that vicious cycle of being addicted to the large upfront license and maintenance fee, so that every quarter you need a mega-deal to pay for the next quarter’s expenses. I want to be in a subscription business with a nice healthy financial model.
I wanted to deal with products that had natural adjacencies where we could build more products. We started doing this for just Ruby on Rails application performance management. That’s a very, very small market, but early on I said, “Well, I can always just expand into other languages like Java, .NET and PHP.” We did that over time, and along the way we just started adding more capabilities, addressing natural new problems as we went.
I always like to have a pretty clear one- to two-year view of where we’re going, and then a much fuzzier three- to five-year view of things. But if we lock ourselves into that too much, we miss the opportunity to learn and adapt. And as a company we’ve always done well by learning and adapting, and that’s why I think we’ve made good bets for the long term.
If you had asked me five years ago, “Would New Relic look like this?” there’s no way I could have predicted it. I had no idea.
“The better you are at playing offense with software . . . the brighter your prospects are. If not, then you’re going to get Ubered.”
If you were to look out three to five years though, where do you see things going?
I think our strategy for New Relic is we’re going to do more with the data we collect. We’re going to be smarter about the data we collect and we’re going to collect more types of data. For example, we just acquired this company called Opsmatic that has a real-time server-monitoring technology that reports all the configuration changes on servers. Imagine if all the data we are collecting from software is married with all the data by about what’s going on in the server, such as who logged into a server just prior to that server having a problem? What was the thing they changed on that server?
You know what brought down the New York Stock Exchange last summer was somebody logging into a server and making a config change. That’s how that data can make all our other products better.
We don’t have specific products ideas on it today, but in particular we want to service the needs of the operations people who have these enormous environments. They’re busy, they’re overwhelmed, and they don’t have time to do anything but barely keep everything running. We want to make it easier for them to manage enormous environments by taking on the burden of doing more with the data that we collect for them.
Do you have a sense of where your customers are heading, or where the application app-development space is heading generally?
The ultimate winners in every vertical space are going be the companies that see software as an offensive weapon. They play offense with software. The losers are the people who only play defense with software. An example of playing defense with software is using software to automate the existing back-office processes just to reduce cost. That’s how people thought of IT in the 1990s and 2000s; IT is there to make sure things don’t break.
Whereas the winners . . . look GE and rearchitecting themselves as a developer company and a software company. Look at how banks are thinking about their new customer-facing entity as no longer the branch, but the mobile app. That is how people do their banking.
The better you are at playing offense with software, at creating a great customer experience to drive the growth of your core business, the brighter your prospects are. If not, then you’re going to get Ubered. There’s going to be another company that will do your vertical what Uber is doing to taxis.
