Most of us engineer success by trying to replicate success patterns. That’s good, but nearly 90% of startups fail, so for Founders, avoiding failure patterns may have an equally important role to play in finding success. The Partners at NFX built ten successful companies in part because, in each company, we avoided failure long enough for good things to happen to us.
There are 6 patterns that account for the vast majority of startup failures, both early-stage and late-stage, according to Harvard Business School professor Tom Eisenmann:
The following are important takeaways from my recent conversation with Tom. Consider these notes a counterintuitive advantage to understanding startup failure, in order to increase your likelihood of achieving success.
1. Failure is the most important phenomenon – and the least understood
“Here I was, supposedly an expert on entrepreneurship, a Harvard Business School educator, and I couldn’t explain the most important phenomenon in my field.”
Year after year, when I taught the first-year entrepreneurship course at HBS, my students would say, “Hey, you told us that something like 75% of startups fail, but we just did 30 cases on these spectacular success stories. These founders come talk to us and everything sounds fantastic — so what happens the other 3/4 of the time?” I took the feedback seriously and wrote a couple of cases about failures.
We’re allowed to invest in student startups after they graduate and I had done that and at least 3/4 of my angel investments failed, so I had plenty to choose from. So I wrote a couple of cases on failed startups that had been led by former students and taught them. The classes were pretty wobbly. It was very hard for the students to figure out what happened and why.
Some students were, as is often the case for MBAs, really good at analyzing by looking in the rearview mirror, “Well, isn’t it obvious that they didn’t do this or that?”
Other students, a little more thoughtful, would say, “Yeah, but some really smart investors put money behind this idea. and if it was really a terrible idea that wouldn’t happen, so there must be a lot of contributing factors.”
We couldn’t tell if failure was a result of death by a thousand cuts — or if some of the factors were more central than others.
2. What Defines Startup Failure?
One of the first things you have to do if you’re studying entrepreneurial failure is to define failure.
I taught a course on entrepreneurial failure this past fall at Harvard Business School and we did a case on Jibo, the social robot out of the MIT Media Lab. It was a robot that could actually strike up a conversation with you and move and dance and stuff. It was a really remarkable product. Jibo failed. It lost $61 million, but the question is, from whose perspective?
What is failure from the founder’s perspective? If the founder’s goal was to build an amazing product that some people would love, which they did with Jibo, then was it successful? The founder, Cynthia Breazeal, wanted to prove that people wanted a social robot in their home and she did. The next generation of social robots are out there, working with autistic kids and the elderly, and so forth. Win.
But from the investors’ perspective — $61 million gone.
You also have to ask questions about what equals a failure outcome. Does a company literally have to go out of business? Not every company fails, of course. It’s not an endearing term but a lot of investors will talk about “Zombies,” companies in their portfolio that they know will never yield a return to the investor but they’re making enough money to keep going. Is that a failure? Does something have to go bankrupt in order to fail?
I settled on a definition of failure which I think will ring true to most venture capitalists, which is the monetary instead of the societal take: “Early investors did not and never will make money.” So that is the definition I use for startup failure.
3. Why The Real Reasons Are Hidden From Us
A lot of venture capitalists are happy to explain why some startups fail and others succeed. But most of what I read, both practitioner and academic, I would describe as oversimplified.
It turns out that humans have a penchant to oversimplify and blame a single misstep for a startup’s failure.
The other psychological phenomenon I saw all over the place in failed startups was the fundamental attribution error. Basically that when something goes wrong, if it happened to somebody else, we’re inclined to blame dispositional factors like the individual wasn’t very talented or they didn’t work very hard. But if it happens to us, we blame the situation.
Previous research had been a little bit thwarted by doing interviews with people who will give fundamental attribution errors, both in the simplification and in the attribution errors.
That’s when I realized what a lot of people had done, the academic work, in particular, were just simple surveys. They would ask people like, “What are the top five reasons why startups fail?”
If you’re talking to VCs they say, “Well, weak team.”
If you’re talking to a Founder they say, “The market moved away from us in unexpected ways.”
What I learned and realized was, the case method, the approach we use at Harvard Business School was actually pretty good at triangulating what was going on in a failure situation. Because when a case is well done you’ve come at the problem from a lot of different directions and talked to a lot of different people. I’d talk to the founder, other team members and investors, look at other similar companies in the space, and so forth. When you surround the problem that way, you’ll get some attribution errors, but then you can make up your own mind about what’s really going on. The cases in the book I’m publishing soon (Why Startups Fail: A New Roadmap for Entrepreneurial Success) and the research I’m doing have ended up being super important.
4. Avoiding Failure
So – why do startups fail? There’s always an easy answer – the cause of death is loss of blood. The company ran out of money trying to find a good opportunity and couldn’t raise and earn more. But why? Well, a “gunshot wound.” Okay, why? What’s going on? Was it self-inflicted? Was it a jealous spouse?
It’s like the Toyota Production System’s “5 Whys.” When there’s a problem in the factory, you just have to keep asking why, and that doesn’t lend itself to a single, simple, explanation as to why startups fail.
Some startups do not find the great opportunities, others don’t have a great team, or maybe they have a bad market. There is a temptation I think for everybody to look for one theory that stretches across lots of situations.
The startup failure question really doesn’t lend itself to that way of thinking. No single cause, but there are patterns and I saw some patterns repeated with early-stage failures.
The Diamond and Square framework
When we teach MBAs at Harvard Business School how to diagnose the prospects for an early-stage startup, we use a framework we call “Diamond and Square.”
There’s a diamond that represents the opportunity and the four corners of the diamond. It’s a mnemonic device to help people remember the four corners: (1) customer value proposition, (2) go to market strategy, (3) technology and operations: how are you going to build this thing and deliver it? (4) Lastly, the cash flow formula.
And then surrounding the diamond is a square which represents the different folks who have to contribute resources: (1) the founders themselves, (2) the rest of the team, (3) outside investors, (4) and strategic partners.
So we teach the students to ask, “How does the opportunity look? How do the resources look and are they in dynamic alignment? Do you have the right resources in the right quantities to actually pursue this opportunity?”
And that’s where the exploration of early-stage failure gets really interesting because what I found were teams that really had, often through Lean Startup techniques, identified a great opportunity and validated demand for a solution, but they weren’t able to mobilize the resources to capture it.
At the opposite end, there were some teams that had mobilized a great set of resources, strong founders, great team, supportive investors, but that team never managed to find an opportunity, a good opportunity.
5. 4 Tactics For Overcoming Catch 22’s
The team at NFX is all about network effects so you’re very familiar with the catch 22 (i.e. chicken-or-egg problem). It’s a logical impasse that takes the form: you can’t have A without B and can’t have B without A. For example, you can’t get a job without experience and you can’t get experience without a job.
Network effects in two-sided markets have that problem: Side A won’t come on board unless side B is there and vice versa. And so the catch 22 with an early-stage startup is you can’t get resource providers to commit, and by that I mean the rest of the team, strategic partners, and crucially, outside investors, until you’ve resolved some risk.
They’re taking a risk by lending their time, their money to you, and you can’t reduce risk until you’ve actually mobilized some resources. You can talk about it, but people will be looking for more validation and more proof.
The catch 22 is you can’t get resources without reducing risk, you can’t reduce risk until you get started, and to get started you need resources.
The four tactics to either mitigate the risk or reduce it in some way, or reduce the resource requirements are: (1) resolve risk, (2) defer it, (3) shift it to another party, or (4) get people to ignore it.
Lean testing is a way to resolve the risk cheaply and still in a rigorous way before you have to commit too much in the way of resources.
And when new ventures get funded inside big companies they don’t come with seed, Series A, Series B. It’s just next year’s money, “Here’s a big lump of capital.” But VCs will stage, which essentially defers the risk until you’ve met some milestone. “If you haven’t met the milestone we may not give you Series B.”
Partnering is another thing entrepreneurs always do. They don’t have resources so let’s go to somebody who is resource rich and borrow their resources and we’ll shift the risk to them, but we have to persuade them that it’s worth their while.
Lastly, storytelling, “fake it ’til you make it” is basically a way to get people to ignore the risk. You fake them out essentially, or in the case of reality distortion field storytelling, you’ve dazzled them.
Those are the four ways you get around the catch 22’s. Each of them has a dark side and you see some of that in startup failure stories.
The dark side of lean testing is people think they are doing it but they skip the upfront phase of customer discovery. They go right to the building. “Let’s build and put a product in customers’ hands and then iterate as fast as we can” but they’ve skipped that first few weeks or maybe even a couple of months of really interviewing customers and understanding their needs and doing ethnography and so forth.
Staging can go bad when you pick the wrong investors.
Partnering goes bad a lot basically because big players are hard to negotiate with. It’s hard to align your interest with theirs if you can get their attention at all.
And then the risk with storytelling, reality distortion fields, and so forth, is that the reality distortion field folds back on itself and the founder is persuaded of the truth of what they’re doing and doesn’t hear the universe saying, “This idea really isn’t working.”
6. 3 Early-Stage Failure Modes
Here are three of the failure patterns that were made clear from the “Diamond and Square” exercise:
“A false start” – good founders/team/investors, but they start building too quickly and waste a cycle and their capital on a bad first product.
“Bad bedfellows” – a good idea that never gets traction due to poor founder fit, weak team, and poor investor fit.
“False positive” – you get off to a good start with early adopters and then it turns out that mainstream demand doesn’t share the same needs and you’ve actually mobilized the wrong resources to pursue the mainstream market.
The “False start” is a difficult and tricky one. For example, an HBS alumnus, Sunil Nagaraj, launched an online dating site called Triangulate straight out of business school.
Sunil was an engineer. Like a lot of engineers, he is great at building things. Sunil loved to build things so he dove headfirst into launching Triangulate without really studying the market, without getting a good sense for customer needs, without running what we would think of today as good minimum viable product tests.
The first version of the product was off target and he spent several months building it and launching it before he figured that out. He had a fantastic team, they were really agile and could iterate and build new things fast, so he went through a couple of pivots. But he had only raised $750,000 and he only had time for a few pivots.
Eric Ries in The Lean Startup book defines runway as not the number of months of cash you’ve got left before you exhaust your balance, but rather, how many pivots can you complete before you exhaust your cash balance?
The “False start” is essentially you get going too fast and you waste a pivot so you have less capital available and you can try fewer things. That’s what happened to Sunil and Triangulate.
It’s very understandable. Entrepreneurs have a bias for action. They love to build. They’re told to launch early and often. A lot of the rhetoric, I would say, pushes entrepreneurs in that direction.
7. Why “Crossing The Chasm” Is Harder Than You Think
Most of us who’ve been in this business for a long time have read Geoffrey Moore’s Crossing the Chasm book. That book puts a spotlight on the difference between early adopters and mainstream customers.
I think I knew about it conceptually, but to see how hard it was for early-stage founders to deal with this question of: Do you build a product for the early adopters? Do you build a product for the mainstream? And how do you bridge from one to the other? And the mistakes you could make if you get that wrong and how it can tank your company — that was eye-opening.
If you create a product that’s perfectly suited to the early adopters, it often won’t be the right product for the mainstream.
Fab.com, an e-commerce company is a good example. They started off with a really highly curated set of funky, distinctive products that were sold to flash sales. They had rhinestone-encrusted motorcycle helmets, things like that, and the early adopters were crazy about this stuff. This very distinctive vision of design that the Fab founders had was bought by a lot of people and referred to their friends so it took off on social media. And the first cohorts that Fab recruited were just fantastic; some of the best that e-commerce has ever seen.
The repeat purchase rate was high and the average sales price was high. They would take this deck to the venture capitalists and say, “I’ve never seen anything this great.” It was there in Philadelphia and they raised what, $350 million or something?
VCs pumped a lot of money in and the expectation was that they could go, go, go. Well, the next generation of customers, the next cohorts, weren’t nearly as excited. They didn’t repurchase. The first cohorts came through social media referrals, so the cost of customer acquisition was zero for them.
But they had to buy the next cohorts. Now, you have this LTV, lifetime value of a customer, CAC, customer acquisition cost, squeeze, where the customers became less valuable because they weren’t repurchasing and they became much more costly to acquire. And they burned through a ton of capital, mostly because the mainstream market didn’t share the taste of the early adopters.
8. Early Stage vs. Late Stage Startup Failure
The big difference is when a late-stage startup fails, it leaves a giant steaming crater in the landscape, hundreds of millions of dollars and hundreds of employees are gone — and we all read about it. Early stage, things fail and you know about it if you’ve invested in it, but they don’t quite have the impact. The late-stage failures can put your hair on end.
The failure rate of late stage is almost equal to the failure rate of early stage — if you use the definition we did before which is “investors didn’t make money.”
What happens with a lot of late-stage startups is they’ve got momentum, the Series C investor buys in at a high share price and the thing goes sideways. And it doesn’t necessarily fail, but there’s a down round or even an up round, but a lot more capital comes in, in a different place in the liquidation stack and if the thing is sold, Series C might make some money but it isn’t going to be the five or ten times they were hoping for and often they may lose money. So if that’s the metric, not getting all your money back, then 75% is indeed the number.
Late-State Failure Pattern #1: Cascading Miracles
There are three failure patterns that I’m looking at for late-stage failure. The first is “Cascading Miracles.”
“Cascading miracles” actually came from John Malone, the entrepreneur who built TCI, Tele-Communications Inc, the biggest US cable company. He got it from a mentor of his, the phrase, and it’s basically that many things have to go right and if any one of them doesn’t, the venture fails.
It’s a math equation where you multiply a bunch of outcomes and if any of them goes to zero, the whole expression goes to zero. You used just 20 coin flips. If you just take five coin flips, so if there are five things that have to go right and there are 50/50 odds, the chances you’re going to come up heads five times in a row is three percent. It’s like spinning a roulette wheel and hoping you land on 31 or whatever.
Where you see this is in audaciously bold startups where there’s this really ambitious innovation plan and often a founder who can sell that.
Because of the scope of the innovation, it’s going to take a long, long time to develop the product. The nature of these businesses also suggests you will have network effects in the background or high switching costs, attributes that are going to cause you to want to go fast once you actually do launch, so there’s a scaling imperative. There’s a long development cycle, there are often partnerships, there are often government approvals, because again, ambiguous legal standing and a whole bunch of things conspire. You’ve got a moving target because you’re not actually going to launch this thing for five years, seven years. Of course, the markets keep moving, technologies keep moving. You have to figure out whether you want to incorporate the new technology.
Jibo was a good example of “cascading miracles.” Iridium, if you remember, satellite phones. $6 billion in losses. Satellite phone service anywhere on the planet. Segway was that kind of business. Silicon Valley veterans will remember GO Corp, which was pen and tablet computing back in the ’80s, early ’90s. And the cases we use today, SpaceX and Tesla, are probably examples of this working. They flipped heads many times in a row, at least so far. But the case I use in the book is Better Place. Project Better Place, which was $900 million spent and lost on an effort to create a network of charging stations for electric vehicles, launched in 2007 and out of business in 2013.
There’s a tendency to overestimate demand. If you saw the original projections for Iridium or Segway, what they thought they were going to sell, and because of the delays you eventually cut some corners with the product. The costs, because of the delays and partners not carrying their end, the costs escalate, and so ultimately when you launch, it’s disappointing, as it was for Jibo, as it was for Segway, as it was for Iridium, over and over again. And these things, when they go down, it’s hundreds of millions of dollars lost. They would have had to have a series of cascading miracles for them to work.
Shai Agassi was the entrepreneur behind Better Place. He was positively Jobsian in his ability to dazzle an audience. And with Segway, Dean Kamen, same thing. He was a really riveting and inspiring presenter. Segway was backed by John Doerr of Kleiner Perkins. Steve Jobs himself wanted to put $50 million in and Dean Kamen wasn’t comfortable with the amount of control that Jobs might want. Bezos wanted to invest. Some of these ventures get going in boom times at the height of a bubble, but not all of them.
The positive side of someone with a reality distortion field is that every company has financing risk, and if this person through their personality is really good at fundraising, then you as an investor think they’ll attract more capital and will get more and more advanced with this.
The financing risk is particularly acute because the development cycle is so long. Better Place was a cleantech startup, started in 2007 and launched in Israel in 2012. You got to live with five years of financing risk and we went from boom time for cleantech investing to basically people deserting the sector.
Late-State Failure Pattern #2: Help Wanted
One of the late-stage failure patterns is “Help wanted” and it’s basically a late-stage venture that still has product-market fit. The customers love the product, the basic formula in terms of LTV/CAC is on track, but something on the resource front goes awry. It might be a mistake made, or it might just be misfortune. We’re in the middle of a pandemic and a lot of startups will fail not due to any fault or bad decisions by the entrepreneurs.
Dot & Bo was the example in the book, it’s an online retailer of home furnishings and the formula was working. The demand side was very strong but two things went wrong.
One, they got hit by a big downdraft in spending on e-commerce circa 2013-14. It was probably a 50% decline across the board and even healthy companies can’t raise more money in a downdraft like that. E-commerce fell out of favor and stayed out of favor for long enough that if you had just stepped on the accelerator as you were heading into that period of financing risks, you were in big trouble.
And it turned out Dot & Bo’s demand model relied heavily on virality and social media. It can be harder to turn that off than the kind of customer acquisition that relies on paid marketing, so that was another misfortune.
I suppose you can always ask questions about whether management should plan for something like that, whether they could see it coming, but the other problem they had was shipping couches across the country; that’s logistically and operationally intensive like nothing you’ve ever seen, except maybe apparel manufacturing.
It took them the longest time to figure out how to do that and how to do it efficiently and effectively. They went through three vice presidents of operations before they found somebody that could get the logistics and operations and shipping under control.
They had strong demand but they had poor margins because they were expediting things. People were canceling orders and there were all sorts of queries to customer service. Anthony Soohoo, the entrepreneur, I think would say he made a mistake in these hiring moves.
The first person he hired because he knew he had these operationally intensive demands. He wanted a generalist, a chief operating officer type, and hired somebody who was good at that but had never really had e-commerce experience of shipping heavy things. The second person had more of that but had the kind of big company background we were talking about a little while ago and didn’t work out. Finally, he nailed it the third time, but by that point, he’d burned through a lot of capital, and then he hit the financing risk.
It was an incomplete team. One missing manager in a really decisive senior role. Having the wrong person in there made it “Help wanted.”
It takes three to six months to fill a position like that, and then you bring somebody on and it’s three months or six months before you figure out whether it’s working. So it’s 12 months of burn because you’re missing one person. And if you’ve gotten everybody else in place, then that just increases your burn every day.
Late-State Failure Pattern #3: Speed Trap
I talked a bit about Fab.com. “Speed trap” is the case there. We can blame this one on the venture capitalists, although entrepreneurs can be complicit.
Early momentum, VCs buy into a company that’s growing fast at a high share price and expect more of the same, entrepreneurs love that. Who doesn’t want to grow the thing? You step on the gas and the customers that arrive later aren’t nearly as attractive as the ones who came in the beginning.
If it’s just a pure software business, it’s a little more forgiving. But if you have to operate warehouses and call centers and so forth, you’re now hiring legions of employees that you have to train. You have to layer in middle managers. You have to create processes and so forth, and you’re going as fast as you can.
You’re often experiencing chaos operationally. You have problems culturally because you have old guard/new guard conflict.
The new guard is jealous of the option gains that the old guard is sitting on. The old guard looks at these new people who don’t understand the mission, who just see this is a job. The new guard are specialists and they think their skills aren’t appreciated. The old guard are generalists and their jobs increasingly are, “What do we do with Fred? He’s pretty good at a lot of things but he can’t run performance marketing, he can’t run the warehouse, he can’t run the call center.” You have all sorts of problems. Of course, the right answer is to slow down and fix things, but you have a lot of pressure to keep going and that’s where the speed trap comes.
9. What People Get Wrong About Network Effects
Network effects are all over our curriculum at Harvard Business School. To get an HBS education, students will be studying a lot about network effects, so certainly having the syllabus is half the battle. The NFX website is the syllabus now, so thank you for that. Despite this effort, here’s what startups tend to get wrong about network effects.
There are a lot of people who will point to virality and assume there’s a network effect. (See: Viral Effects Are Not Network Effects). Sometimes there is, but if it’s just a word of mouth referral, that’s not a network effect, that’s word-of-mouth referral.
And the other place where it gets confusing and I think people, even experts on network effects, can have a reasonable debate about this, is a situation where density of the network, sheer traffic, scale of some sort, makes the product more valuable.
I tend to reserve “network effect” for when the users of the network are actually interacting with each other, usually through a platform, and if there’s no interaction I don’t consider it a network effect. Then it’s a scale economy.
You get into businesses like dockless bike sharing where the company buys a whole bunch of bikes and the more bikes there are, the more comfortable you are as a user that there’ll be a bike in the right place at the right time. But it’s not really a network effect.
Google in some ways is the same. With a huge amount of search traffic, they can draw better inferences about what the best listings are to serve up to you. The product gets more valuable the more people use it, but they’re not really interacting with each other through the product. So those definitions are an issue.
I still see a lot of mistakes assuming that network effects are stronger than they are. But at the same time, I think it’s easy for founders to wave their hands and say they’re going to harness network effects without really understanding what they are or how strong they are.