Scaling Category-Creating Companies - An 8VC Event (Transcript)

Sep 12, 2022

Sep 12, 2022

This transcript comes from a panel that 8VC hosted for the SaaSBOOMi community during a series of events in Silicon Valley. SaaSBOOMi is a massive and fast growing pay-it-forward community of 1500+ Indian SaaS companies. We believe deeply in the ability of SaaS founders across the world to build global from day one. 

The goal of this panel was to share insights in taking software companies from 0-to-1 and scaling them, especially when playing in new and emergent budget categories. We were thrilled and honored to host Doug Merritt, former President and CEO of Splunk, and Guillermo Rauch, founder & CEO of Vercel - an 8VC portfolio company, in a conversation moderated by Bhaskar Ghosh (“BG”), Partner & CTO at 8VC. 

In this, they cover:

  • How to recognize product-market fit and what bars people set for determining whether PMF has been achieved.
  • Frameworks for discussing PMF internally.
  • Instrumenting and creating an operating system for your core business. 
  • Communications & operations challenges as your company scales. 
  • Founder patterns & anti-patterns. 
  • & much more …

If you read this and are excited, and are in the midst of creating your own category, please reach out to us at or

Bhaskar Ghosh (BG): Hey everyone, we’re delighted to have two amazing panelists today for this SaaSBOOMi and 8VC joint event. Guillermo and Doug, would you please introduce yourselves for the audience? 

Guillermo Rauch: I’m Guillermo and I started Vercel and Next.js. Our mission is to create the world’s best web infrastructure to develop, review, and ship your applications. We’re huge believers in the web as a platform that you should bet on for your startups and your projects. We’ve done a lot of open source - Next.js is an open source framework that powers things like my own personal blog ( and applications like Twitch and TikTok on the web. Vercel is a serverless frontend infrastructure platform where tens of thousands of customers are scaling their applications worldwide. One such customer is actually BYJU’S - the Indian education company. We’re deployed globally and power projects all over the world and are empowering this new generation of frontend developers that do not have to figure out everything about the cloud themselves and yet are hopefully running laps around their competitors by leveraging the power of this serverless platform.

Doug Merritt: I’ve got an unusual background where I started as a developer years and years ago custom coding for Accenture. I spent time as a coder, then in pre-sales and sales roles, founded a company where I was CEO at a young age, then managed engineering, and eventually became a CEO again. An unusual background. You’re typically just software development or sales and climb the ladder through that career path. I joined Splunk almost 9 years ago now and left officially January 31st. It was, when I joined, just $300M in perpetual license, all up-front revenue, which is about the equivalent of $100M in ARR if we had actually had a non-perpetual, subscription, cloud-based model. We’d been public about a year and half when I got there, and I got the CEO job a year later. My journey - love the product, love what Splunk does - was a huge lift and change management journey despite the growth. It was a fully stateful, vertical scale-out, on-prem only, perpetual license product in 2013. I remember first saying “What?!” - all proprietary, single C++ binary - “Why are we doing it this way?” My pitch to the board was that we had to reconstruct the entire company if we wanted to make it a lasting company. It is now serverless, in the cloud, and we transitioned all of our licensing to term and/or cloud. We ended January 31st in the fiscal 2022 year at about $3.2B in ARR. It was a tiring but super rewarding 8+ years. 

BG: That’s amazing. We’ll start with Guillermo around product-market fit. Guillermo, this is the third company that you’re building. In the domains you’ve worked in with the last two companies and now with Vercel which is in hyper-growth, how did you recognize PMF and how did you go about learning how to identify it? Would love to hear some theories and anecdotes.

Guillermo: Yeah, I have a bunch. I have a somewhat extreme take on PMF which is that you have PMF when people are literally just ripping the product off your hands and bringing down your door in desperation of how much they need your product to do their business or outcompete their competitors. I mention that this is somewhat of an extreme take because you can build really successful businesses when that characterization of PMF is not true. We’re in the technology business of putting things behind hyperlinks that are easily accessible and they can travel the entire world in 150ms. You have to err on the side of: “Things move really fast, get adopted really fast, and you know if people really want it really fast.” We have the good fortune of being open source which allows you to test PMF in a very accelerated fashion because you’re basically giving away a free version of your product. And developers also are not only 150ms away from your download, but they’re willing to try anything to make their lives better. 

So, when we launched Next.js we really found PMF within weeks. We were getting adopted by some of the largest logos and companies that I’d heard of and I’d never talked to them directly! I had created Next.js to solve a very particular problem: we wanted to build our own website which was a little content form with React. The process of doing that was so tedious. We knew that React was the future (which felt like discovering electricity), but I just wanted to build a simple page and I couldn’t. It turned out that all these corporations are also feeling that pain and were interested in solving that problem. So every week we kept hearing: “Oh, is being replatformed on top of Next.js. Oh, now Twitch!” - the adoption was so fast. We even heard input through GitHub Issues like “ I just tried 100K requests per second and found this bug - can you fix it?” I thought there might be a PMF at this point. 

We’re able to cut off a lot of other things by listening to users. Now I have a nose for what PMF looks like and I have a really high bar. You can try many things and discard many things until you find a genuine PMF. 

Our company grew so much during Covid because we had oiled that machine of product-led distribution. We just kept getting signups and the product kept growing. Next.js has had two hundred million downloads in the six years that it has existed. A hundred million of those came in the last twelve months. You get a sense of how it continues to accelerate and if the system is robust, it will scale. There’s no significant differences that we’ll have to make to the software to substantiate the next two months of downloads. 

BG: Guillermo - I’ll get back to you. Doug - adding to that question, in your experience with really large product business before Splunk and of course during Splunk, if you could also layer into that answer how that works with multiple products? And we were really thrilled to learn about this framework Doug is going to talk about and would love to see you introduce it.

Doug: Just to build on what you were saying, I’ll get to the framework in one second. Part of the conversation we were having earlier is that the VC community is so focused on developer-led products. PMF is so natural when you’re dealing with a bunch of technicians that are going to give you immediate feedback, but in a cloud-first world the benefit that we have in the operating system that we create for our business is that everything can be data-driven today. Whether it is a developer instantly seeing downloads and usage, or whether it's a procurement officer or a healthcare worker - the debates that lead to politics and confusion in most organizations can be completely solved if you put an “company operating system” in place around narratives, deep thinking, and feedback loops that you can’t argue with. Questions like: “As a code push happened, was it used or not?” If only two people are actively using it, it clearly wasn’t something of value. There is capability for the flywheel, no matter who your audience is.

Guillermo: The data-driven part is huge because we had some opportunities where we could have gone down the wrong path. There were certain features that could be attractive to different audiences that aren’t high growth opportunities for a venture-backed startup. When we were developing our serverless platform, sometimes we would hear feedback from developers where we started developing a sense of whether each developer just wouldn’t pay or whether they were a qualified enterprise developer. GitHub stars are free and downloads are free for the most part, but that idea of propensity to pay is important. What is the lifecycle of the initial enjoyment that you’re able to unlock with the product? What does usage look like 12-24 months down the line? Is it an expansive usage?

One of the things we noticed with Next.js is that with our early adopters, we had made the technology something usable in a single page. For signup pages, it was PHP versus us. We made a product that was very additive to the stack. We didn’t tell them to rewrite their API diagrams. They could just bring in a new facade for the signup - Next.js enabled them to improve signups by ~200%. I don’t even know the root cause sometimes - it was probably fewer errors, a faster page load, more motivated front-end engineers - but they were able to validate the hypothesis of “Next.js is the future” with a very small sliver of product.

This is easier said than done, but can you find the developer that is really motivated to use the open-source product? Can you imagine a developer in an office that has that concrete business problem: “We want a signup page to perform better.” You have to have that dual mindset. They have to be able to take the technology and solve a real challenge. That’s what PMF is for us.

Doug: This fits into what BG was saying about the framework. What I found over and over - a twenty year exercise now - is that to find PMF there are four different macro factors that you have to consider to have real discussions and understand who you’re talking to. 

  1. What size customer you talk to (e.g. small business, vs medium or large).
  2. What buyer you talk to (e.g. manager of DevOps).
  3. What industry they are in (e.g. federal government, healthcare).
  4. What region they are in (e.g. US - East vs West Coast, India, Japan – decision making differs). 

You never have an interesting discussion until you qualify all four - otherwise it’s all theoretical. Once you get to those four, keep that rigor of simplifying and doing less within these quadrants, and be very conscious of moving off of them.

Every board slide for the past three years, I had this cube. It’s so intuitive but it’s so hard to get people to actually adhere to.

BG: As you scale through different phases, what are the common challenges that you see moving from early-to-mid-to-late phase as a combination of communication, data gathering, operations structure, and talent needs. How do you maintain this balance and make them work together? 

Doug: Just at a high level, if we use an example of those four quadrants, a lot of those problems become more simple if you can become clear as a company and disciplined about those factors. If we’re going after mid-sized financial services companies and their CIO in the US. Now whether it is a sales person, an engineer, a PM, or a product marketer, you atleast are getting rid of one of the elements that can create misinformation or people talking above each other without realizing they are doing that.

The other aspect that is so beneficial beyond PMF is the entire scaling exercise of the business becomes significantly easier if you have those four quadrants. You now know what types of sales reps to hire, what locations to target, how to optimize SEO, what product marketers to hire and what language to use, and what trade shows to go to. The entire world becomes so clear. The difficulty always that you read about in every business book is - “I’m beginning to optimize and reach critical mass and either need to leap to another industry, buying center, or country.” All the risk is that there is a continuous PMF flywheel that we’re never done with - with Splunk at $3.2B ARR we’re still going through this. Creating the operating system within the company to keep focus on those vectors and be explicit about when to make a leap to something new. The difficulty with the leap is communication, coordination, and time. There’s now new people, with new skills they need to add to the picture (e.g. because the sales representatives with experience in financial services don’t know healthcare) - it’s not about learning but sometimes about relationships. It’s a very “go slow to go fast” that wasn’t in vogue in the past three years but we now have the opportunity to drive this discipline so we have this high quality today.

Guillermo: I’ll add one that is very related to this. One of the biggest mistakes of young, growing companies is that they under-instrument their business. They don’t have enough data flowing in and informing decisions. You realize too late that retrofitting that instrumentation - that ability to X-Ray - is a multi-domain problem. 

Looking back, one of the things I found the most valuable is that we said “let’s pause and instrument to get insight in real-time about how the business & technology is operating.” With fast-growing PMF came a lot of noise and a lot of potential for noise. A lot of this also involves saying no to things. There’s a lot of attractive side quests - especially in prolific engineering-savvy teams. If you have a really good engineering team, they might be compelled to reinvent things every single week. They might have the right reasons and technical talent to do it and you still have to say no. 

Early days you can get really compelled to do a lot more than you should be doing. You learn the hard way that you’re going to be cutting stuff and that comes with some pain.

BG: I’ll come back to GTM at a later phase. We’ll move very briefly. [To Guillermo] - anything you want to call out in open-source - getting developer interest and network effects as building an open-core business. Any tidbits or takeaways?

Guillermo: I’ve now had the fortune to develop 5-7 really successful open-source projects. My biggest open-source project is one called ms which takes in a timestamp and converts it into a human-readable format. A lot of internet applications that have relative time - e.g. “3m+2s ago” - that’s probably my library because it has 150 million downloads a week! But I can’t build a business on that.

Even if I charge $0.00001 per installation, they still are not going to pay. I think you have to think about open source in the context of well-understood, real business models. Technology gets invented and replaced very often, but successful business models and things that people are willing to pay for - and create business centers with important personas - don’t emerge every day. 

The Web3 space is a really interesting case study - is the way you build a successful business there with token economics and charging for usage or as a traditional company like Chainalysis?

I think you have to stick with open-source that has a well-understood business hypothesis to go along with it.

BG: (To Doug) My first manager was your old head of security. How did you guys think of and start the security business which became huge. What went well and what was hard? 

Doug: Nobody at Splunk initially envisioned that, and it was our customers that showed us. The benefit of Splunk became much more synonymous with these monitoring and predictive analytics applications that deal with structured data. To make those work you had to structure the log data first. Splunk’s core initial value was to take “garbage in,” not automatically transform, enrich, or cleanse it, but have the index and the language - SPL - allow you to interrogate that data and find structure. This whole orientation was: “when something goes wrong in a data center that no one thought was going to go wrong - a one-off unknown-unknown event - you have to deal with chaotic data.” As our customers ingested raw logs initially for the IT-ops team, because it was non-structured, anyone who got interested in what could possibly be in that data could make sense of the data using SPL. 

Back in 2011-onwards, security teams heard the success that IT-ops teams were having in finding needles in the haystack which is the same activity they have. They’ve got a SIEM tool and other tools that are trying to refine structured data and they cannot show us the answer. They’d say: “you’ve clearly got the answer using SPL to show us how to stop something from happening - can you go up the stack?” That consistent demand led to the SIEM application. It now is an almost-$2B revenue product. The market found us. 

BG: Switching to leadership. What goals did you set as a leader, not just as an entrepreneur, for your previous companies and forward-facing?

Guillermo: I’m originally a technical person. I’m a front-end developer that happened to create this really popular framework and a really successful business on top of this framework. My number one mission for the company is that I’m instilling a company culture that is truly “Vercel.” I’m defining our identity, mission, and vision and trying to create a scalable model such that the thousandth employee we hire can be effective without top-down direction.

I mentioned data-driven decision making frameworks as a way to unlock this kind of thing. What’s working when you’re 10-20 people is very different from the scalable, organic, semi-biological system that a company becomes when it’s at 300, 500, and 1000 people. So much of my job is defining that culture and what success looks like. Defining what the metrics that come from that instrumentation that we go after. Defining our objectives and strategy and letting it flow, as well as being strategic and opportunistic about embracing the very best ideas bottoms-up.

I very much love the DRI framework of Apple - where you sit in an organization doesn’t determine the value or weight of your opinions or ideas. I make it a point to collaborate with folks all across the organization, not just in my glass house with an executive team. At the end of the day it comes down to: to create a generational company you have to go much beyond yourself. As a Founder-CEO I have the opportunity to do this. 

BG: Looking back, what goals had you set the first three years and how did they evolve? What were the main takeaways of what went well?

Doug: I had the advantage of being at Splunk for about 13 months before I interviewed with the Board for the CEO job. That gave me enough context to be very clear about where we were then and what we needed to do to be successful. There was a five-year plan that I gave an overview to the Board in August 2013 that we more-or-less followed all the way to the end of my time at Splunk. It started with what needed to transform within the business - product, culture, process. Outside of that, we tried framing - based on what we were doing - the market growth and what we should look like 5-7 years down the road. I set a really ambitious $5B revenue target for the company and started to break down for the Board what we’d have to do to move from a search tool to a multi-product company that went after the IT-Ops, SecOps, and DevOps buying centers.

There’s an interesting McKinsey study at the time that had all of the companies that made it to $100M, $500M, etc. In enterprise software it turned out that 18 companies had made it to >$5B. If you wanted to create a seminal, generational company, there were known external quantifiable metrics you had to achieve as well as internal, qualitative metrics. 

BG: Any thoughts around CX and customer lifecycle management?

Guillermo: A lot of companies end up developing major strengths and weaknesses here. We’re strong with developer experience. One of the things we’re fortunate to have developed organically by hiring great people is that we’re known for our customer experience prowess and customer obsession. 

This was defined by our culture and who we are when we hire determines that. It has become a very unique weapon for us because the technologies we’re dealing with are not entirely straightforward and the world is hard to predict. We have customers that see extraordinary amounts of traffic or move so fast that they’ve tripled the size of engineering for front-end products in just a matter of months. It has been really rewarding working with them.

Doug: We were really purposeful about that as well - certainly for the past 7-8 years. At the end of the day, pre-sales, sales, and customer success should rally around a document that shows who is purchasing, what business process they want to enhance, ROI expectations, and timeline to achieve that ROI. That document became the bible for handoffs.  

If you want to be successful you need to maintain details on what customers need to achieve. We complemented that with growth metrics on the commissioning side to show ROI. 

Guillermo: There’s another side to that. You can get too caught up in how much time the customer is using the product and don’t talk to them about why they’re using it or what they’re trying to get done. Especially for large customers, being very aligned on the business objective that is driving the choice of vendor, and making that relatable to every individual in the organization is important.

You have to get aligned on success metrics. Getting the truth of where you’re missing - between your customer’s and your own - and democratizing that truth helps you scale. 

BG: What is the one mistake which you don’t want the founders here to make?

Doug: The one that I would be ten times more diligent about - politics and empire-building ultimately kills a culture and that goes back to who you hire, clear job specs, and clear metrics. There is an operating system for a company, just as there is an operating system for the technology that we use, and to drive that operating system, which I tried very hard to convert Splunk to, means that data rules these decisions. That needs a lot of discipline. Do I know how to engage this person, quantifiable metrics, etc.? 

At Splunk we hired 3000 people in 18 months. We had new people managing new people. Internal discussions about budgets and headcount which radically impacted the culture. 

Guillermo: For early-stage, most of it all is just coming to terms with the truth that “Product might not work, people might hate it, technology is hard, and your first customers might just be your friends.” It’s hard to go from five to fifty customers. Your first customers don’t have that staying power or distribution power. The vast majority of products go nowhere. You need to be really honest with where you stand with PMF and growth. 

Continue Reading