Startups that survived did this…
New research reveals when startups start scaling, how it relates to their likelihood of failure, and what startups that survived did differently. Get ready for a wild ride, it’s not what you think.
Do startups that scale fast fail fast? Let’s break down new Wharton research that answers that exact question—with mind-blowing results—and also something much, much more interesting: how do some startups scale fast and not fail? How do they survive?
I’ll break down the findings and why everything I thought about scaling fast was wrong. 🤯
Heads up, this new research is still a draft, so the final version may change, but it’s already been reviewed by many academics, so the main takeaways are unlikely to change.
Two conflicting hypotheses
The researchers analyzed two opposing perspectives on when startups should scale.
The first perspective is a widespread belief in Silicon Valley and Ricky Bobby’s life motto: If you ain’t first, you’re last.
Reid Hoffman, Linkedin’s co-founder, calls this idea blitzscaling, which prioritizes speed over efficiency because you'll likely lose out if you’re not first to scale.
Rocketship success stories of companies like Airbnb, Uber, and Beyond Meat are often cited as examples of why this theory works. On the surface, these companies embody the scale-fast-or-crash mentality and reap the benefits of amassing resources, contracts, network effects, and economies of scale.
That’s the popular Silicon Valley scale-fast hypothesis: the earlier a startup begins to scale its business, the less likely it will fail. In the study, it’s stated like this:
Hypothesis 1a: The earlier a startup begins to scale its business, the less likely it will fail and/or the more likely to successfully exit (i.e., an IPO).
Or is the opposite true, does scaling fast increase the failure rate of startups? That’s more of a business school counterargument and the second hypothesis tested in this study.
Hypothesis 1b: The earlier a startup begins to scale its business, the more likely it will fail and/or the less likely to successfully exit (i.e., an IPO).
This second hypothesis suggests that startups that scale too fast have less time for experimentation—talking to customers and testing their assumptions—which makes them more likely to launch an unproven idea. Researchers call this commitment risk, the mistake of coming up with a bad idea, building that bad idea, and then being so committed with time and resources that it’s too difficult to pivot to better solutions.
So, which hypothesis is correct? Do you experiment or scale fast? Or better yet, can you do both? We’ll dig into the findings, which are still a draft, so these numbers aren’t “final-final” even though they probably will be.
This research is so exciting because it’s the first time anyone has analyzed the timing of startups scaling on a large dataset (more than 38K startups!) with math.
How the researchers measured scaling is really clever, I won’t get into the details here, but if you’re a data nerd 🤓 like me, check out the entire paper. I stayed up late reading all 60+ pages because that’s how I spend Friday nights, but you’ll see, it’s more fun than you think.
Ok, on to the findings.
Wild Fact #1: Four long years
Out of 38,217 startups, the researchers found that, on average, startups post a job for their first sales and manager positions at 49 months—the same time for both positions—roughly four years in. The researchers used job postings for the first sales and manager positions to indicate that a startup had begun scaling its business.
This finding is wild. I was shocked by it. Four years is much longer than most people would expect, including your mom and all your friends. If you’re a founder, they’re all wondering what you are doing with your life, chasing your startup dreams for that long.
The research shows more than 77% of startups scale after two years, and only 4% scale in the first six months. Wild. What happened to overnight success? But did these companies succeed? We’ll get there.
Wild Fact #2: Smart money scales slow
There’s a common belief that VCs want to see their portfolio companies scale or die as fast as possible. One speed, go. Surely, of the companies that scale fast, they must have VC money. No. Only 0.1% of the startups that scaled in the first six months were VC-backed, barely a fraction of one percent.
Compare that to a whopping 19% of those that scaled after 24 months being funded by VCs. This seems to imply that the smart money wants you to take time to experiment and validate customer assumptions before you blow all their smart money on a dumb idea. Smart.
Wild Fact #3: Scale fast, fail fast
Wild fact number three is the answer you thought you were here for, but I warn you, wild fact three may be misleading if you don’t stick around for wild fact number four…
Scaling within the first 12 months represented a 20% to 40% increased failure rate. Bad news Ricky Bobby, if you’re first, you’re last.
“We find that startups that begin scaling within the first 12 months of their founding are 20 to 40% more likely to fail.”
The study used a few different complex models, and the results consistently showed scaling within the first 12 months represented a 20% to 40% increased failure rate, where scaling was measured by the first job post for a manager or sales position.
And overall, fewer companies failed when they scaled after 32 months, about three years, compared to those that scaled early. (Keep this in mind when we look at Airbnb later.)
The graphs and charts in the study are a bit confusing because they use multiple data points, which need to be compared to the already astronomically high failure rate of startups in general, no matter when they scale, but the overall results are clear. Scaling early was bad.
Interestingly, this was also true for companies that were first to market, when you would think the saying, “if you’re not first, you’re last,” would matter. It didn’t. Scaling slower was better for the average startup.
And what about two-sided platforms, where you have buyers and sellers, like Airbnb? A top benefit to scaling fast is network effects. You need a large network of buyers to buy and sellers to sell, and you need both fast because one doesn’t exist without the other. Surprisingly, the relationship between scaling early and failing was three times greater for two-sided platforms than for non-platform companies.
On all accounts, scaling early was bad, except for one.
Wild Fact #4: A/B testing changes everything
The theory behind this research is that scaling too fast results in less time for talking to customers, validating ideas (which I cover in detail in my video on the problem space), and experimenting before wasting money on a large scale.
So, the Wharton researchers said, if this is really about experimenting, let's identify the companies that use advanced A/B testing software for experimentation. In most cases, these companies are paying a lot of money to run experiments, gain customer insight, and experiment with multiple options before launching anything.
The results are staggering. There was no meaningful risk for companies that A/B tested and scaled in the first six months, while companies that A/B tested and scaled in 7 to 12 months actually had a reduction of 9.7 to 12 percentage points in the likelihood of failure. This flips the research upside down and says scaling fast is bad, except for the companies that use A/B testing software, then it doesn’t matter, and they might even perform better when they scale fast.
The researchers conclude that the risk of scaling early does not apply to startups that engage in A/B testing, supporting their argument that scaling early is a risk because it increases the likelihood of commitment risk. That is, scaling early is only a risk because it increases the likelihood that the startup didn’t take time to talk to customers, experiment, and ultimately committed to building products that nobody wants.
And yes, both of the unicorn startups I worked at did a lot of experimentation before scaling.
The blitzscaling misconception
Timing is not the issue; it’s about talking to customers and experimenting. This is the misconception of blitzscaling, where Reid Hoffman recommends prioritizing speed over efficiency. Nowhere does he suggest that you should blitzscale a bad idea that hasn’t been validated with customers or experimentation—quite the opposite.
Reid Hoffman was a series A investor in Airbnb, and Airbnb is a popular example of blitzscaling, scaling fast, even mentioned in the Wharton research we just discussed. Hoffman invested in late 2010. I don’t know when they hired their first manager and salespeople (which is how the Wharton paper measured starting to scale). But, they hired their first product manager in early 2011, which may have been around the same time because their team was really small before that. This chart below makes it look like they scaled like a rocket ship without time for user research or experimentation.
But let’s ask Airbnb’s co-founder, Nathan.
“... and Paul Graham told us something else he said, ‘it's okay to do things that don't scale.’ It's counterintuitive a little bit, you're making an internet company the whole idea is that it's self-serve and it can scale to millions but he said when you're trying to find that product market fit it's okay to do things that don't scale, and he asked us where are your users, and we said well our users are everywhere, he said well no where do you have the most users, we said New York, he said go to New York and meet all your users…”
– Nathan Blecharczyk, Co-founder Airbnb (Source: YouTube @17:45 - 18:14)
You can’t just look at a chart of when their user base scaled fast, after 2010; you need to go back, years back, to 2007.
Airbnb started in late 2007 by renting out air mattresses solely for events—Reid Hoffman famously passed on that initial idea—and it was through the founders living with their customers, literally, over years, that they found that users wanted to rent out entire homes, that users wanted an easy payment platform, and that users needed help taking pictures. (This was a time before Instagram when camera phones were terrible.) I found a TechCrunch article from June 2010; Brian Chesky, another co-founder, was living full-time with customers in Airbnbs while their apartment was still the tiny company headquarters, years after their first booking, and this was just months before Reid Hoffman's series A investment.
Blitzscaling, much like overnight success, can happen after years of hard work.
As with Airbnb, sometimes we look at when a startup scaled rapidly without considering all the years of experimentation before that. And that’s what this research shows, for the first time, on a large dataset, with math! More importantly, it’s not about scaling early or not; this is about focusing on the customer and experimenting before you scale the right product.
And, get this, the person who originally told Ricky Bobby, “If you ain’t first, you’re last,” was also taken out of context. You’ll have to watch the end of my video above to find out why. 😉
You got this!