Clear eyes, full hearts, can't lose

For startups, there is no step two.

In August of 2013, 45 companies used Slack. The product was still in private beta, and its test accounts were almost entirely small businesses and startups, ranging between five and 75 people. The nascent product had promise, though; as CNET said at the time, Slack put "group messaging, file uploads, and notifications into one tool that syncs across all platforms."

Over the next ten years, nearly everything about Slack has changed. Its four dozen trial accounts grew into more than two hundred thousand paying customers, $1.5 billion in annual revenue, and active deployments in 77 of the largest 100 companies in the world. It launched scores of integrations, a developer platform, and features that make it possible to use Slack for external as well as internal conversations. A global pandemic dramatically changed how people work and communicate with their colleagues, compressing a rise in remote work that would’ve taken forty years into a month.¹ Slack’s CEO quit; Slack got acquired by Salesforce; Salesforce’s CEO quit. Compared to what it was in 2013, nearly everything about Slack today is incomprehensibly different. 

Everything, except, what Slack is. When I first used Slack, I would’ve called it a chat app; today, I’d still call it a chat app. In its first Product Hunt listing, Slack christened itself as “real-time messaging;” Slack now defines its product as a “messaging app for business.” ChatGPT—which isn’t an oracle or an encyclopedia, but a kind of poll of the internet, and a rough approximation of what the average person on Reddit might say—describes Slack as an “app that helps people work together by letting them chat, share files, and stay organized in one place,” which is almost identical to how it was defined by CNET a decade ago. The world has changed and the business is unrecognizable, but Slack’s center of gravity—best revealed by how people would describe it in one sentence of simple words—has been written in stone.

This is true for nearly every major technology business.² In 2010, UberCab was a “service that offers an on-demand car service via an iPhone app.” In 2023, Uber has a new CEO, a new name, and a dozen new corporate appendages³—and it’s still primarily an app for ordering a car from your phone. Stripe launched as an online payments processor for developers; despite ballooning into every business from banking to a publishing house, Stripe is now, first and foremost, an API for accepting payments on the internet. Fifteen years ago, Twilio released an API for sending texts and making voice calls. Today, the title of Twilio’s homepage is “Communication APIs for SMS, Voice, Video & Authentication.” Asana is still a task management app. Salesforce, despite acquiring a number of companies like Slack, is a CRM. Twitter has spent nearly two decades trying to figure out its second feature. 

Of course, companies mature. Markets blow up and fade away. New features get bolted on; brands change; CEOs give keynotes about reinventing the future, about relaunching the vision, about next chapters, about versions 2.0. But in all but the most dramatic cases—the true pivots, when major lines of business are shut down—companies’ foundations rarely budge. The banners under which they launch is the banner under which they grow, go public, and in some cases, collapse⁴ and die.⁵ 

In other words, the two-step startup is a myth. It’s not uncommon, in early stage fundraising decks, for founders to pitch their first product as a stepping stone to a bigger one. We’ll start with building a simple alerting service for monitoring when your brand gets mentioned on social media, they might say, but we’ll use the data we collect to build an AI-powered ad platform. We’ll launch as a data catalog, and use that as a wedge to expand into our true vision as an end-to-end data lifecycle management application. We’ll begin an office leasing agency, but eventually harness the energy of we to elevate the world’s consciousness. 

These stories are tempting to tell. They feel simultaneously measured and ambitious. They appear to explain how a niche product with barely any users can ladder its way into a big market and $100 million in annual revenues. They feel clever, like Elon Musk’s master plan. They “preserve optionality.”⁶ 

They also don’t work. When the first step doesn’t go as expected, companies spend their entire lives trying to finish part one. And when the first step does work, companies get captured by their customers, and can rarely escape the shadow of their initial successes. One way or another, prerequisites become permanent.

Put differently, for a business to grow into something big, its initial product has to grow into something big. The market might change—it could grow underneath a product as it did for Uber, or new markets could open up, as enterprise buyers did for Slack—but the soul of the product almost certainly won’t.⁷ 

Clear eyes

The gut check for founders and early-stage investors, then, is this: Do you believe that the product description in the VentureBeat writeup about the seed round can also be the first sentence on an S-1?

In some cases, the answer is a clear yes or no. But most of the time, it’s a harder call. Plenty of companies struggle and stumble, start slow, or stall out at a couple million dollars of revenue and twice that in expenses. They see occasional sparks of interest for their initial idea, which are enough to keep the pull the business along but not enough to light a bigger fuse. 

The mythology of Silicon Valley tells these companies to keep going. Airbnb famously teetered through a near-death experience in its early days; lots of other companies fought through long slogs before becoming “overnight” successes. It’s easy, in these situations, to keep tacking on features, in hopes that some tipping point is just one or two releases away. It’s easy to experiment with partnerships, or to launch a free offering, or to rebrand under a more current buzzword. 

But for companies living in this trough of sorrow, they need to be clear-eyed and honest—uncomfortably, brutally, searingly so—about the goal of these efforts. Are you exploring these things because you’re convinced that your original idea—that one sentence description in the VentureBeat article—is still white hot with potential, and can carry you all the way to an IPO? Or are you exploring these things because you know that it’s not, and you’re looking for a partial pivot?

Because if it’s the latter, the battle is lost. Yes, full, hard pivots can work. Slack, for example, spun out of a gaming company that was building a wholly different product. Instagram began as a social media app for bourbon drinkers. Segment went through two pivots before becoming what it’s known for today. In these cases, the first product wasn’t renovated into the second; it was torn down and picked apart for scrap metal. 

Partial pivots, by contrast, are a prolonged death march. If a product only attracts a curious interest from people who never seem quite ready to buy, a couple new features won’t ignite demand. Partners won’t turn a mediocre idea into a great one. And if people are reluctant to spend money on something, they won’t be excited to use a free version of it either. For a startup to succeed, its founding idea has to work, or it has to change. And slow arcing pivots are neither. They’re the aimless distractions of a company that lost some deals and is on its way to losing itself.⁸ 

Full hearts

At this point, I’m probably supposed to talk about how important data is for developing conviction in an idea. If you carefully track how people are using your service, I should say, you can figure out if your idea has product market fit. Buy a BI tool; use the modern data stack; trust the science.

For early stage companies, however, this sort of analysis is somewhere between useless and actively harmful. Because if an idea doesn’t quite work, you won’t compute that painful truth; you will feel it. You’ll feel it in your wavering belief that you’re making the right bet. You’ll sense it in how you push down your own doubts in investor pitches. You’ll hear it in prospects’ voices when they politely tell you that they think what you’ve built is cool, that they’ll follow up after the call, that your email got lost in their inbox but they’re definitely still interested in checking it out. 

Startups don’t die because leaders miss these feelings; they die because teams ignore them. They die because we’re supposed to be empirical and data-driven, and these anxieties don’t show up in our dashboards. They die because that same anxiety is uncomfortable to admit, and even more uncomfortable to address. They die because we’ve made our emotions inadmissible evidence.

But there’s urgent meaning in that emotion. A head full of doubt can still be a road full of promise, but a heart full of doubt can’t win. 

1. Though really, it all happened in one day: On March 12, 2020, the day that Trump suspended travel from Europe, that the NBA season got canceled, and that Tom Hanks got COVID. 

2. “Survivorship bias!,” you scream. “The airplane, you fool, the airplane!” 

Yes, but that’s the point? If very few of the companies that kept trying to contort themselves into some new idea didn’t make it home, then…don’t do that? 

3. These include: A food delivery service, a self-driving car division, a self-driving semi-trailer truck division, a helicopter service, a vertical-takeoff-and-landing aircraft division, a bike rental service, a scooter rental service, several AI labs, an alcohol delivery service, and Postmates. 

4. This is poetically true for Secret, which apparently imploded in part because its CEO sold a bunch of shares during one of their fundraising rounds and used that money to buy a Ferrari. He tried to hide it from Secret’s employees, but someone found out and posted it on Secret. “I never thought leopards would eat MY face,” sobs man who created the Leopards Eating People’s Faces app.

5. The same is for data companies. Segment is a switchboard for web events, despite dabbling in ETL. Fivetran is ETL, despite dabbling in data transformation. dbt Labs is data transformation, despite its recent dabbling in semantic layers. Transform, which dbt Labs bought, is a semantic layer, despite dabbling in reporting. And Heap is reporting, despite dabbling in being a switchboard for web events.

6. This is a bug and not a feature.

7. Startups, you might say, are what you thought they were.

8. At which point the company might not die, but will stumble along, stuck in time, leaving everyone wondering how on earth it’s still around.