Jeremy Au discussed common startup failure patterns, and emphasized that failure is frequent and often inevitable even for companies that make it to later funding stages. Many startups fail to deliver financial returns for investors, regardless of how innovative or pioneering they might be. He also talked about various reasons for failure, from team issues to premature scaling, providing real-world examples to illustrate these patterns:
- Good Idea, Bad Bedfellows: A strong idea can fail due to poor team dynamics, such as co-founders unable to agree on leadership or lacking the right expertise.
- False Starts: Startups that build products without understanding customer needs often fail.
- False Positives: Early customer success can mislead founders into scaling too quickly, leading to failure when they target the wrong market.
- Speed Trap: Startups that achieve product-market fit but expand too quickly into new products or markets can burn through capital unsustainably.
- Help Wanted: Sometimes external factors like market shifts or bad luck cause failure, even when product-market fit is achieved.
- Cascading Miracles: Some startups that fail despite raising large sums of money or low customer traction later spark similar successful ventures.
(00:05) Jeremy Au: So we're going to be talking about startup failure patterns. Long story short is that startups fail a lot. It is, by nature, the most default form of condition or end state for startups. So for example, this is a good analysis that's done for US startups by Crunchbase.
(00:21) Jeremy Au: So basically, at the top of the funnel, you can imagine that there's a thousand seed startups. that had already raised some capital. And out of this 1,100, 1% of them eventually reached unicorn status, including Stripe and Docker. So about 12 companies. This is their funnel from top to bottom. But what's interesting is that each stage, there was still a failure rate. So for example after seed, 48% failed. And then out of 335, only 172 made it to the next round. And out of 172, only 96 made it. So you can see that actually even for these rounds, there's still quite a big drop off rate if you think about it, right?
So 15%, 9% divided by 15% is almost like about two thirds. Closer to half actually. So if it corresponds series A, series B, series C, series D, series E, even as something like series C, only a bit over half of them actually make it to the next stage.
So what I'm trying to say here is that a lot of people will be like, okay, it is true that the early stages hydrate, but again, as I shared before, people underestimate that later stage startups can continue to fail. So something for us to be thoughtful about.
Obviously, a lot of companies will die because they fail to raise. Some of them will exit or sell themselves in a small outcome. A lot of them will end up being self sustaining, so they can't break even, but they never reach unicorn status. And then companies that break out of the exits, they have small exits from 50 million plus, 100 million plus, 200 million plus, 500 million plus, 1 billion plus.
There's another way of looking at it, is that most investments don't work out. And so what we wanna talk about is what is startup failure? So a lot of folks are kind of like, okay, you know, we use the word failure, failure, failure. You're a failure. Nobody wants to be a failure. Don't grow up like Uncle Jim. He's a failure. So there's a lot of words around failure they want to talk about, but of course one thing is for startups is are they failures or are they pioneers, right? So for example, Jibo was the world's first social robot. They ran out of cash in 2018. They had previously raised 73 million of capital from 2013. So after about five to six years, they died. And basically what they had this thesis was that they wanted to create an AI companion that would be next to a bedside table, that would be in person. And obviously, they learned a couple of things. One was that during engineering, they learned that the hardware would be more expensive. For example they found out that although camera sensors, were relatively cheap at the time the grade of cameras that they needed to have in a domestic housing environment, which can be quite dark from time to time, needed to be better than what they had thought it was going to be so the hardwares would be more expensive.
They also had to engineer middleware. Just because you have a camera sensor means that the camera can understand what a human face is. So they had to re-engineer middleware to make those cameras be able to explain what a human face is saying, and so forth. Of course, they had bad luck. The CEO was diagnosed with leukemia. The CTO stepped up for a year as interim CEO. And then out of the blue, Amazon Echo launched. Amazon Echo launched with no camera, just the microphone, so a much simpler version, kind of similar to how the USB thumb drive was invented by the Singaporean, just a simpler version of the MP3 players that he was working on. So a simpler, streamlined version of that product became normal. And if you think about it, it's kind of crazy, but like I shared before, my four-year-old and two-year-old now know how to speak to their smart assistant to ask, " Hey, Google, can you play Wheels on a Bus?"
What's interesting, of course, is 2023, we're now back to AI powered social robots. Digital companions have come back. We also saw Elon Musk release these humanoid robots. A lot of them were tele-operated. But again, it's the same thing, right? Try and make a robot become your friend, right? So was Jibo failure a pioneer? It's hard to tell. But, I think for the purposes of what we are talking about today, we want to define failure as the startup failed if the early investors did not or never will get back more money than they put in. Some people will be like, wow, that's a pretty chill statement. Some people will be like, okay, that's actually quite an aggressive statement. But what I'm trying to say here is that we're looking at this from a financial perspective, financial return perspective. Because again, we are looking at this class from the perspective of venture capital. We make money when there is a return on investment and the cash is paid back to us eventually. It could be a smart bet that never paid off. It could have been a pioneer for the new world that led the way. It could have generated new alumni that went on to build bigger and better things. Those are all positive externalities that society and governments benefit from, which is why governments often subsidize VC and venture capital in emerging markets and developed markets as well. But I guess from our perspective, from a VC perspective, is you don't get paid for those things. You get paid if you make money from the investment.
And so the reason why it's difficult to understand startups' failures objectively is because of three pieces. First of all is single cost fallacy. Startups often fail due to multiple factors. We, as humans, like to say, "The CEO sucked," "Government killed it." You know, like, we want to kind of simplify it to a simple reason, like one single reason. So it's not very interesting to read this multifactorial analysis of all the multiple reasons why the company failed. So we tend to oversimplify for that.
Number two is fundamental attribution error. So when we observe others, we tend to say it's their personality, their skills, their diligence. If he fails, we say, he failed because he sucked, not because the environment sucked. Now, if I fail, I tend to say I didn't suck. I say the environment sucked. So people have this kind of like, now we're saying this, "judge others for their actions, judge myself for my intentions." But basically, humans tend to have biases about how you evaluate yourself versus evaluate other people. And then when it does fail, it takes time for that failure to happen, the investors, the teammates, the media often blame the founder. Often the founder often blames external circumstances or parties. And when they're busy accusing each other, it's actually quite hard for us because they're all pointing fingers at each other. Obviously, the media is going to write different stories. And the same failure can be written five or 10 different times depending on whose perspective you look at. So what I'm trying to say here is that people try to oversimplify. People tend to, like, not weigh personal factors versus environmental factors well. And lastly, everybody's pointing at each other. It takes time for us to process and really understand the failure because it requires people to be honest, or requires a court case to bring all the facts to life.
(06:41) Jeremy Au: So there are six common types of startup failure. Good idea, bad bedfellows. The second is called false starts. The third is false positives. The fourth is speed trap. The fifth is help wanted. And the sixth is cascading miracles. So this comes from a book from Tom Eisennman. He is a Harvard Business School professor who's done this analysis. For good idea bad bedfellows, what that means is that this roughly corresponds to the timeline, a chronology of the startup over time. So for a good idea and bad bedfellows, it's quite simple. A decent idea, but the team was not a good fit. Now, the most obvious version of that would be the two co-founders are not a good fit for each other. And they compete, and they cannot collaborate, and they break up. So there's one common version of that. There are different versions of that. It could be a function of unclear who is the boss? They are co-CEOs, for example, There could also be a lack of flexibility, initiative. Maybe they were looking for investors, but the investors did not bring the expertise they wanted. They may have the wrong partnerships. But basically, the team that came together across the founders, employees, strategic partners, investors never made it.
For example, there's a company called Quincy Apparel.
(07:49) Jeremy Au: This was founded by two Harvard MBA folks, and they basically wanted to create a direct to consumer female fashion brand. So today, Love, Bonito would be a good example of a company that successfully made it in Southeast Asia, but this was very much at a time there were other startups that were also building direct to consumer. So they were basically trying to be the Bonobos, but for female work fashion.
Other companies did succeed in that space, and if you read the book, they talk about multiple reasons for failure, but one interesting part was the two co-founders never agreed on who was CEO. I think that's the problem with having two Harvard MBAs to be co-founders, is they can't decide who's the CEO. They both didn't have fashion experience, unfortunately, so they will hire the employee to a fashion experience, but that didn't really work out in terms of decisions. One of the errors they made was that they thought that they had the sleeves right. And it turns out that when they actually made the print order in terms of the manufacturing, the sleeves were the wrong size, which caused a lot of wastage of their early seed capital. And also they had an early investor that said that they were very good at, and had previously invested in fashion, but it turns out that when he brought on that investor, it turns out that investor had very generic experience, was not very hands on. They were like, they did invest in the fashion startups before, but they were very hands off, they didn't have the ability to bring the skills into it. So this company failed because they never marshaled the right expertise to make the right decisions. So that's one.
The second is called false starts. This is quite common where the company forgets to research the customer needs before engineering, and then they build an MVP and they launch and they're pretty excited. Normally, it is driven by a founder. eagerness and overconfidence. So in other words, this company builds the wrong product that fails customer needs. So this company is called Triangulate.
(09:34) Jeremy Au: It is not data driven dating in 2009. So it raised $1.5 million as an engineer, and basically his thesis at the time was that there wasn't Tinder, there wasn't Hinge, all these other things. So there was, Match.com. So basically he said, "What if we use your Facebook profile and your social media and absorb that data to train an algorithm that matches you with the right person that will make you happy? Now, he was very excited because he's an engineer. In the book, and that's exactly how he wants to date people, which is that he wants some algorithm to tell him who he should date. Turns out nobody wanted that. And now we see that the industry practice for dating apps today is your photo your age, your location, and your height. So those are the important things for your Tinder and maybe your occupation, right? So those are the things that are there, very visual heavy. And of course, there's a bunch of different approaches to that. I'm just saying like, he got it wrong in 2009. Nobody wants an algorithm to pull my personal data to find me the right person. And I think we see that a lot, actually. A lot of founders kind of go through the experience because They feel like people want it, and then it turns out the market doesn't want it.
The third kind of failure is called false positives. So what it means is that you did successfully find an early bunch of customers that did love the product, but then you incorrectly get the wrong lessons from them and you kind of step on the gas too early. So you accelerate. This is also driven by overconfidence and the founders are excited. Investors are excited. They say, "Wow, you know, you're doing it, you're crushing it, you should go faster." So what that means is that you build a customer correctly, and then after that, they end up building a product for the wrong people over time. For example, this is Beru.
(11:10) Jeremy Au: They raised $2.6 million for pet services. Their thesis was that, we will help property owners deploy pet walking services in your building. In other words, B2B2C, if that makes sense. So, we will partner with these buildings, and then everybody has this service called Beru, so we can walk your dogs or pets. Lindsay Hyde, she's a great lady and several failure points happened. She started out in Boston. And when she started out, it was also winter time and apartment buildings that she worked with had a lot of people that were coming in for the first time for filmmaking. So basically, the building that she was partnering with, nobody wants to walk their dog during winter time, so it's cold, so you have a lot of growth. And then two is, these people are very busy on the movie set filming. So they don't have time to walk the dog during the day. So they had a high initial growth from the early buildings. But when they expanded, they expanded to the wrong city, the wrong buildings. And so as a result, by the time they kind of figured it out, it was too late and the company had to close down. So you get initial demand, but you took those lessons and you extrapolated it wrongly.
These earlier failures that we talked about tend to be more invisible. We kind of hear that from our friends or family. But of course, the type that we see in newspaper tend to be larger, more catastrophic failures. So a common example would be speed trap. So these companies have achieved product market fit not only with their early customers, but also with the expansion customers. But they have saturated their target market. And now they start to expand to different verticals to try to make the market size larger. And then they end up losing money in all those different product categories that we see here. So the reason why is that they often have this perception that there's a winner take all. We talked about blitzscaling. So if you believe that you are a super app and that people really want to buy cars and insurance and banking and healthcare and everything because of this app, as an example, then you believe that there is network effects. Therefore, you believe that blitzscaling is good to have because it's a winner takes all market. Therefore, you're going to spend a lot of money in all these different business lines. So the concept is there's a land grab, let's grab them all, let's do it before our competitors get it. But of course, one interesting thing is that if your company is growing very rapidly because you're throwing a lot of money at it, it often attracts rivals who try to compete with you. They may themselves be VC funded, and then because of the competition, your margins will drop, right? And then of course, as you have this unsustainable pace, it tends to lead to staffing bottlenecks, disorganization, complexity, ethical lapses. And then suddenly, you know, and they're burning maybe, like, a million dollars a month or whatever it is, burning very aggressively with the assumption they're going to receive another VC round. But then the investors suddenly say, hey, I'm reluctant to invest. The CEO slams the brakes to slow growth and do layoffs because they can't grow at that rate anymore with VC money. So, for example, fab.com started out very much with effectively a daily deal for home items. Fab.com is like effectively a Groupon, flash deals of course now we now see in all of our shopping apps like Lazada or TikTok Shop, or Shopee, like all of these techniques are now used. But for them, this is the first generation of this. So they were doing flash sales. We will only sell 100 units at 80% off or so forth. And they raised $336M to a billion dollar valuation. And then the company imploded because basically, it turns out that they just find more growth. And so you're burning a lot of money trying to find new items. But then the company died.
(14:36) Jeremy Au: Other category is called help wanted is sometimes it's just kind of like bad luck. So you do have a product market fit, you do grow the customer base, but sometimes it's just really bad luck. So for example biotech in 1990s was very popular and then it died because everybody stopped investing in biotech. So the big drought. Clean tech in 2000s, 1990s, they're very popular, and a lot of them all basically went bust. So, Kleiner Perkins lost a lot of money as a VC in cleantech. So, the guy was correct, but he was just 20 years too early in his bet. So the entire industry was kind of like written off for a while. We saw crypto got killed for many years, but now crypto is coming back. Sometimes you just buck, you are going to make it, but suddenly because of the environment or high interest rates, or government policy decides that everybody should no longer do private tutoring, then it kind of kills your sector, and then you just have bad luck. So Dot & Bo is a furniture company similar to Wayfair in the US. And then basically, they had raised $19 million, so doing well, but they assumed that the market would be okay, and then the e-commerce bust happened around that time, 2014, 2015, and then they just ran out of money, and then the whole thing closed up shop. I think generally, this is defined the way that the professor did, but I would probably define this as more like bad luck, I would say, or it's more like macro, micro spike kill.
And then the last group that we see cluster that we see here is called cascading miracles, which a lot of us see the most spectacular failures, but also see the most spectacular success. So these are companies that have a very huge ambition but they often die with very low customer traction despite raising like over a hundred million dollars of capital. What they have in common is that they have to do several things. They have to persuade customers to do something new for the first time. Then, they have to build a totally new technology stack. Then they have to partner with the right suppliers or vendors to make it happen. And then they have to secure a regulatory partner to make it happen. And they also have to raise a lot of money. So any of these things can kill the entire company. So obviously, the legendary wins that we saw is Iridium. But of course they did it very, very early, around the 2000s and basically they died because the technology was too expensive. Customers never got on board with it, so eventually Iridium was acquired by the US government. So the US government said, you know, actually this technology is very helpful, and so Iridium, the satellites are still in play, and they're run by the US government. It's a legendary flop because they raised hundreds of millions of dollars and then they died. The opposite of that is that SpaceX made it, which is the exact same play, which is, they made a belief that can I convert everybody to use Starlink satellite dishes? Nobody thought about installing satellite dishes in your own home before, the past three years. But now, people are doing it. You have to create a new technology, which is launching satellites, the power, all these things. So SpaceX made it work. Iridium did not make it work. Segway was electrical mobility. So Segway was like, "hey, you know, can I make this two wheels? I you guys have seen it, and then can I drive it, and then blah blah blah." that didn't work out. They raised hundreds of millions of dollars. They never became a success. But, you know, Tesla made it work, which is actually it's the same technology stack, which was batteries, vehicles, gyroscope. So, electrical mobility, it just happened, you know, about 10 years later down the road in a different format of it. We saw Webvan. Webvan was in, back in the 2000s, died during the dotcom crash. But basically their concept was, what if you can buy groceries from home? It just turned out that at that time, buying stuff from home required dial up when they started it. And then also no driver had smartphones. So they were unable to track their drivers. So they had to sign routes and itineraries to the groceries. And obviously, the inventory and stock taking was way too expensive because all of it was being done manually. It was just way too early. Then FedEx made it work. And of course, the famous story for FedEx is, he almost ran out of money and then he went to gamble and then he was able to double his money and survive as a company. And now FedEx is successful today. But at that time when he set up FedEx, he wanted to create this idea of a global or pan-American network of delivery and everybody thought it was crazy because at a time, it was inconceivable that people would want to send a package to the other side of the country within one to two days. So BetterPlace is a good example. BetterPlace in 2007, they raised $850 million to build Battery Swap Network. So basically, the concept was that electric vehicles is a no brainer. Therefore, we're going to create these battery charging stations across Israel at that point of time, which is a good, from their perspective, test place because it's relatively dense, it's high income, good regulatory support. And so they created these stations where there are battery packs, and so you could drive your car to it, and then you could swap your battery instead of using normal petrol. That company died a horrible death in 2007 because there were insufficient cars that could use the battery swap network at a time. They had to make both batteries and they had to make cars that use those batteries, but if you think about it, that's what Tesla did because Tesla eventually said, we're gonna create gas stations called Tesla Gas Stations, but they use electricity and we're gonna make our own cars, right? So Tesla made it. Obviously we see that Gogoro in Taiwan made it. They have their own battery swap network. In China they made it as well. The most tremendous type of failures that we see here are often this type as well. So after failure, founders can make a decision about what they do. So for example, Lindsey Hyde. She was the founder for Beru earlier. She is now a professor at Harvard teaching about entrepreneurial failure and how people can avoid that and so forth. So you can move on with your life. You also see that founders often do rebound, revenge and rebirth startups. So for example, we see a lot of founders, I know that after they leave their startup, they feel like, you know, their identity is a founder and so they go straight into the next startup straight away. Rebound is a little bit derogatory in that sense because it feels like they haven't really thought about it, but it's more about the identity of being a founder. So you see that these founders often go through a lot of product market fit conversations after that. Revenge founders after a founder failure. So for example, we saw that Parker Conrad. He set up Zenefits which sold to HR offices and that became a billion dollar company. He was eventually forced out of the company through a board vote. And he basically left and he was very frustrated as well as his COO who is David Sacks, but David Sacks is the co-host of the All In podcast. He's currently a strong Republican supporter but he was a COO for Zenefits. He took over as effective CEO, and then he went on to build a successful VC fund called Craft Ventures, but Conrad Parker basically said, I want to take revenge. So he went back to Y Combinator. Paul Graham supported his second startup, and he decided to go after the same target customer, which is HR offices, and he eventually built a second billion dollar company called Rippling, which does all in one HR automation so there's a revenge. And now today's benefits is zero, and Rippling is a billion dollar company. And then we talk about rebirth. So rebirth would be basically like, founders who take their time off, they may be acquired, but basically like a phoenix, they kind of do some digestion of it, and then they decide to build a company that is maybe in a different field, different interests.
So for example, if you look at Palmer Luckey, he built the first company called Oculus, which is VR, which was acquired by Facebook, which is now Meta. So the Metaverse, all those goggles are done by him. He was there for a while. He was fired or let go for his support for Republican candidates, primarily Donald Trump. And then after some time out, he decided, and we shared about that, he had to make a decision, he wanted to build two companies. One company was to lower the cost of prisoner. He wanted to disrupt the prison industry by paying a success fee to prevent prisoners from going back to jail. And then his other company he decided to build, and eventually built, was Anduril, which is a defense tech startup, which does drones, anti drone weapons, sensors, fences all kinds of stuff.
But I think what we define Cascading Miracles is that these are really moonshots technologies that are hard to believe at that point of time but can be a big win for a time. So it kind of depends. I still remember back in 2014, 2015 I was hiking the Pacific Crest Trail. For one month from Los Angeles to Yosemite, San Francisco pass. And you know, part of that is you have, you come in and you come out. But I basically took a bus and I was just hanging out with this person. And I was sitting in a bus between SF and LA. And the guy I was sitting with was this old guy, he was a former engineer. And he was telling me about how he's a volunteer of the Sierra Club, which is an environmental club for protecting nature. And then he told me, "Jeremy, I'm very excited." And I was like, "why?" It's like, "yeah, you know, I'm all in on Tesla." And I was like, "Tesla? Isn't Tesla going through production hell?" And they scale up the volume. They've only done the sports car. All the papers at the time were saying that it's likely going to die. He's never going to compete with Ford and so forth. He's like, Jeremy, I got myself a Tesla because I believe in the environment, but trust me, he's going to be able to figure out this production hell and build a mass market car.
I nodded politely, I thought he was kind of like, nice for you, and I missed out an opportunity to make a lot of money investing in Tesla, because he made it work, right? And if you look at the Tesla stock, I probably should chart it out for fun and make myself regret, but if I put my little amount of savings, I didn't have a lot of savings back then, but I'm sure I could have made a lot of money with that investment, right?
So by today's standard, it'd be considered a miracle that happened, right? Now, obviously today we kind of say like, wow, it's a no brainer, blah, blah, blah. I think SpaceX is another example of a cascading miracle. We saw that last week. The latest heavy rocket that is 10x cheaper, to 20x cheaper than the conventional space industry. I mean, people didn't think it could happen. And the truth was that if you look at the biography of Elon Musk, several rockets failed and it all exploded because of the failure. If the first rocket that actually succeeded after multiple failures had failed, SpaceX would be no more. So today, we'll look at this as a cascading miracle that worked out, because these are very difficult things to do. But if it had failed, I think we couldn't have said it's bad luck or whatever. We just said, look, this was a cascading miracle kind of startup. the odds were not in their favor.
That sums up failure patterns. I think a lot of the founders would say it's bad luck. And then a lot of people outside would say, "you suck."