An interesting idea. The trouble here is that analysis like this tends to document the symptoms of failure as if they were the causes. The actual causes are more likely to be very complex, based on circumstances unique to the startup/people running it, require deep insider knowledge of the company, and in some cases be things people aren't willing to admit or recognize.
It's the flipside of a similar problem in analyzing why companies are successful. [1]
The little disclaimer at the bottom really says it all.
The attempt is a noble one, marred by data and insight quality issues. I think it could be useful if the site can source insightful analysis from founders/insiders and make it easy to search by market/product category. Perhaps even adding a badge to information which came from a founder.
Something I've noticed with both my own past startup failures and other startups I've known: by far the most common failure reason is "There was no reason for them to be a company in the first place."
By that I mean that either there was no customer demand for what they were building, or there were already lots of other companies that solved the problem just as well and they had no unique angle on the problem, or a key technical assumption they were relying on turned out to be false, or the market was better served by lots of little firms rather than one high-growth startup. In other words, they never found product/market fit, because there was either no market for the product or they couldn't build the product to serve the market.
The problem is that usually you can only determine this in hindsight. If everybody assumed that the only businesses that can work are those that already have a working product and customers, we'd never get any innovation. I've learned to think of "Finding a reason for the company to exist" as the primary job description for a founder, and failure means that you are doing your job but haven't completed it yet.
My hunch after starting a few side projects, and taking the leap in to a full on failed company is that while things must be tried, you should secure some level of real customer interest before diving fully in.
I'm not sure how much it takes, but there should be real evidence.
Most startups that outperform start in areas where the reason to exist isn't obvious. It only becomes obvious later. (Do we need yet another search engine?) This is why it's important to have a lot of low-capital companies searching these markets. Once they find the reason, then it's appropriate to flood them with capital.
A valid question even when Google came to the fore. We had Yahoo, HotBot, Altavista, and then Google came along. And while the others were search engines, Google was superior.
Will people switch to a better internet search engine ?
Sad fact, but true. Most companies simply have no need to have been started, and the failure stats bear this out. But as you also state, in aggregate, this process is necessary or time stands still.
What's interesting thinking about history, would Webvan and Pets.com fall under "no demand", "too soon", or just tried to do too much given the timing? Now we have services that are similar, for instance chewy.com is the modern version of of Pets.com.
"Too soon" is usually another way of saying "A critical piece of infrastructure that my business plan requires doesn't exist yet."
Instacart's founder is fond of saying that Instacart couldn't have existed before 2012. I suspect what he means is that there were a collection of technical & social changes that happened in the early 2010s that let him recast the problem Webvan was solving in an economical way. These were: 1) smartphones allow real-time coordination across thousands of workers, without hiring lots of managers 2) cloud-computing lets you run big-data algorithms to feed those instructions to thousands of workers, without building data centers 3) because of the Great Recession, thousands of workers were unemployed and desperate for some way to earn money 4) increasing urbanization has clustered people together in a city and made them disinclined to drive, which increases the demand for a grocery-shopping service and decreases the cost of servicing them and 5) everybody had Internet access and cellphones.
By contrast, WebVan spent a billion dollars building warehouses, at a time when the total number of Internet users was < 150 million. They bought their own fleet of delivery vans and hired their own drivers, at a time when employment was full and labor costs were high. They had to invest much more capital for a much smaller market, and then had much higher variable costs.
Being a "startup" doesn't repeal the laws of business - you still have to pay for labor, generate returns on capital, generate more value than you charge, and charge more than you spend. But because computers operate millions of times faster than humans and don't require wages, if you setup the business model right you can realize huge efficiencies of scale. "Setup the business model right" is the tricky part - WebVan thought that "Internet ordering" was a crucial part of the business model for online grocery delivery, but it turned out that "use existing supermarket infrastructure" and "coordinate lots of shoppers so they can work very efficiently" were the real keys, and the technology to do that hadn't been invented yet.
As evidence: I can get a 35 pound bag of dog food delivered to my house in about an hour via Amazon Prime Now. That's basically both WebVan's and Pets.com business model but on the back of Amazon's logistics ecosystem.
What makes this work is both that it's not 2001: both in that there are significantly more Internet users, they have higher speed connections, and they're much more comfortable with online purchases.
Totally this. I lived with my aunt in the bay area back when webvan.com was hot, and I remember a time when the two of us saw a commercial of theirs. I remember saying how great of an idea it was, that it could really take off, and I distinctly remember my aunt -- not a very prolific internet user but not a luddite by any means -- saying that it was weird buying groceries online. That really stuck with me, because my aunt is/was the kind of person to try something like that, but even she wasn't comfortable with it.
Plus also webvan.com blew a billion dollars building a delivery infrastructure that they couldn't support.
Yes. A big issue is confusing correlation with causality. For instance, "There was no business model, and the founding team was inexperienced and arrogant" applies to Facebook and Google as well as many failures. :-)
I feel like former employees sometimes know better why a company failed than founders. If the founder knew exactly what is/was "wrong" they probably had been able to turn it around.
Former employees are often biased by their job function. At the first startup I worked at, I had a role that straddled Eng & QA, and the lesson I took away from it was "engineering quality is critically important; we failed because there were too many bugs." In hindsight - having now seen v1 of many other startups - we failed because a lack of nerve, because many other startups charge thousands of dollars a month for product quality significantly worse than what we had. We should've shipped it, had money coming in, used that to negotiate another funding round, and then used the funding to fix the bugs. But without visibility into common sales & fundraising practices of startups, there was no way that 19-year-old-programmer-me could've known that.
If you have an all new product you need to get into the market fast. Once you have the basic features working you need to see if customers really exist. Even if it means your customers have to try everything twice because you crash the first time, if they buy your product you fix the bugs they see so that it mostly works and move on.
However if you don't have a new product you cannot do that. If you want to release something where the market already exists you can just do one part better you need to be as good in everything else. Tesla didn't release their original roadster without a heater, in the 1950s heaters were optional.
You need to figure out which market you are in and release accordingly. Getting this wrong means the death of your company. If you are in the first investors are taking a risk that people will want your product - it would be stupid to invest in perfection when customers might decide your product doesn't fill a need, better to abandon your interesting but useless product early. However when you are in the second you need to meet your user's expectations - thus I don't need to check the feature list to tell you Tesla comes with a working heater standard. Expectations is also why the early reviews of Tesla showed a tow pulling it away with a dead battery - the equivalent stupidity in a gas car would be the tow truck charging $10/gallon for gas and you are on your way in a few minutes. (Tesla has been mostly successful in managing expectations in the years since - now everyone knows it isn't the best car for cross country trips but you can do it once in a while with a little planning - there is a lesson in this too when you are a little different in an existing market make sure the downsides of your different are understood)
If you can. I thought the context of this thread was free-form postmortems, where people give a narrative explanation of why the startup failed. It's hard to aggregate that.
Also, there's the risk of aggregate counts simply reflecting the population of each job function, unless you weight them. Then it'll just say that every engineering-focused startup failed for engineering reasons and every sales-focused startup failed for sales reasons, which may be true but isn't terribly helpful for a founder trying to figure out what to do.
It's an interesting thought, almost like a "glassdoor.com" on startup failures. I'd be curious if Glassdoor actually implements some sort of a weighting scheme to counter the bias that unhappy people are more likely to post reviews, so a poorly run group would have an outsize impact.
It's because of that case I think free-form postmortems might be more insightful, despite their own obvious faults. Perhaps NLP and sentiment analysis could shed some light.
I agree--although they might not have the "full information" as jstandard points out, the employees are in the trenches all day and have to deal with the trouble spots directly. For probably all places I've worked that failed, the employees simultaneously knew the problems (and would gripe about them) but were not empowered to actually fix them. Anyone who's been in this kind of death march situation knows that feeling of helplessness--that the track up ahead is bent and if only we could just run up ahead of the train and fix it...
>If the founder knew exactly what is/was "wrong" they probably had been able to turn it around
Not always. Changing a single person's bad habits or tendencies is hard enough. 10 people? 100? Culture carries momentum even if the prime mover is found.
I agree in part. The real trick of it here is that those former employees will also be wearing their own lenses of bias. There's a good chance they don't have full information on the company.
Then the challenge becomes a game of synthesizing contradicting perspectives. That's what makes this exercise of distilling causes into "bitesized pieces" so difficult.
There is a lot that goes into running companies that ordinary employees take for granted. I don't think there's any reason the perspective of the average employee will be more accurate than the perspective of the average founder. Both will have strong biases coloring the perception of events in their own favor, and in some cases, problem employees and problem founders may blame each other instead of admitting their own faults.
In most cases, there is, of course, no simple way to pinpoint a particular place where everything ran amok. That's to be expected. The analyst needs to listen to everyone with credible knowledge of the company and use their own judgment to come to an opinion on the biggest issue. This is a subjective analysis, and not something that can be authoritatively established.
This is absolutely true, and I'll take it even a step further: great people can still form dysfunctional teams even if they all do good individual work.
Often times in an early stage startup the problem is the team doesn't gel enough to adequately explore the problem space. In those cases, the founder might conclude that the opportunity wasn't there, but maybe it's because people weren't aligned enough to push far enough in the right direction. It's impossible to know whether there was really a viable business there, or whether it was a failure of the team. But subjectively after going through several startups and other early stage projects, I know the feeling of a team that is gelling, and it can make an unquantifiably huge difference to early results.
> "Both will have strong biases coloring the perception of events in their own favor, and in some cases, problem employees and problem founders may blame each other instead of admitting their own faults."
This really resonated with me. It's truly a skill to build the self-awareness to realize when you're the problem. It's much easier to externalize or abstract the problems away, particularly in stressful times where everyone feels overworked. I've caught myself doing this.
I have to set aside part of every week to step back, breathe in, and candidly examine how I might be contributing to problems that are happening. Otherwise it's too easy to get wrapped up in the hunt for demons outside the burning house.
That's true, but ordinary employees will have a perspective that founders usually just can't see, usually due to their biases. As such, it'd be better to take in the perspectives of both, to paint a more complete picture.
My assumption has long been that ~50% of startups fail because they never actually create a useful product. But, looking at those failures does not tell you much about startups just the basic pitfalls of product development.
Similarly, many companies that fail made perfectly rational bets, but things outside their control killed them. Likewise many successful companies may simply have gotten lucky due to things outside their control, even if their initial odds where poor.
Which is why I feel like success and failure on their own tell you very little.
It is interesting that the Homejoy entry is gone. Typical analysis is that they lost business to their own pros and the class action lawsuit regarding the contractors or employees argument. My own experience in the industry suggests that bad treatment and pay for their contractors resulted in ongoing defections and that when a competitor offered just four dollars an hour more all the best cleaners left at once.
$4 an hour is I guess in the region of a 10% to 20% raise. Not sure about you but I wouldn't turn that down when the difference is working for one faceless online platform versus another.
The funny thing about salaries that many companies don't understand is that those extra few dollars can make a huge difference to workers.
There's this kind of narrative going around recently that beyond some magic figure (pick a silly number like $70k or whatever) extra money doesn't make you happier etc.
What this totally misses out on is that to have a financial future you have to make money over expenses.
If it costs you $20/hr to live and you make $24 then $28 actually doubles what you have left at the end of the month.
This is a valid point but it is worth pointing out that in this specific case most Homejoy cleaners were making $11-13/hour working usually 4-6 billable hours a day in big cities and as such could not afford $100+ a month for wireless or cable.
In addition to the points above, Timing is an another consideration. What didn't work in 1999 works today because there are users online sophisticated enough to expect a solution.
Sometimes solutions have to stick around and survive through a few waves of the solution becoming relevant - either the market catching up to demand the solution, or the solution developing to meet the market.
It's a fine reason for building solutions in a lean way that can have a longer runway to let things align
> The trouble here is that analysis like this tends to document the symptoms of failure as if they were the causes.
This also holds true for successful businesses!
For example, you will forever read books on successful companies and they will say things like "we were successful because we were agile". In reality that is probably not the case, more likely success was because of a million undefinable small things all mashed together.
At the same time... I see a lot of the same errors at lots of startups. It can be hard to generalize best practices, and harder still to recognize when you are failing to follow them.
Lots of startups fail for timing reasons. There seem to be alot of startups that maybe just started at the wrong time. One example is QBotix that provided robotics for the solar industry, probably not much of a market a few years ago, but in the coming few years that should explode.
Actually for QBotix it was the other way around: the robot was designed to reduce the cost of solar trackers (instead of a motor on each tracker you had one robot that moved around to adjust each tracker). The problem was that the cost of PV fell so far that it was cheaper just to deploy more, use a fixed frame, and not worry that your panels were never really pointed directly at the sun.
I had hoped to read the QBotix page to see what they said but like others I got a 503.
Potentially a great idea. I have wanted to see data on the reasons startups failed.
One of the problems with this sort of thing is that it is not clear what the category of the failure should be. In a sense, the vast majority of failures is: no sales.
But is this because of "product-market fit" (they just didn't want it), or "customer acquisition costs" (couldn't get the word out), or "lack of runway" (can't get the word out fast enough).
That's why you almost need some hybrid of story telling and data so you can compare what Balaji Srinivasan calls "the idea maze."
I wish I could compare the series of decisions in some manner, and not just read it as a story.
But in those days the reasons usually boiled down to "this was a stupid idea with no way to actually produce revenue after the VC money ran out." Totally different from now.
Every year the idea of a startup post mortem site or study comes up and as always goes nowhere because they can't actually draw a causal relationship and tell the story of why the company in question rolled up.
Besides the fact that there is no incentive, in fact there is negative incentive, for key players to contribute to the study of failure, it also requires insane amount of depth in the very specific field which the company was operating in.
It's the same problem as failed replicated studies not getting documented, it's easier and higher incentive to just try again than to study the issue.
Really just needs to be a non profit that would run the studies and maybe turn it into a consultancy or something.
When outages happen, there are companies that do blameless postmortems, and they do get good data.
I think it's a matter of creating the right incentives. Possibly an organization like YC or a VC that's existed for a long time does. Although they tend to regard such knowledge as proprietary, so, no luck for us.
> Really just needs to be a non profit that would run the studies and maybe turn it into a consultancy or something.
Now you're on the right track.
Many non-profits like this already exist: universities, and sometimes, governments. There are lots of papers out there about why businesses fail, and governments are strongly incentivized to fund this kind of research.
I don't know for sure though, but there is probably a paucity of data about why wacky tech startups fail.
- At least until recently, it hasn't mattered very much. Actually, I would argue it still doesn't matter.
- Startups are bespoke and weird by definition, so analysis is hard. But possibly an academic could maybe test theories of exactly how much risk you should take on (I'm thinking of mcfunley's insights on "innovation tokens"; maybe that can be formalized somehow. http://mcfunley.com/choose-boring-technology)
its because web hosting companies offer tiered services, and often the lower tiers have limited bandwidth. When that limited bandwidth is reached, the web hosting company cuts off access to the site.
Who still uses traditional shared hosting for a tech related project when there are dozens of alternatives with either dedicated or semi-dedicated hardware without the overly intrusive bandwidth caps.
I front paged on HN 3 times with a blog running on a $5/mo dropplet and out of the box wordpress without any CDN. I only had to give up one cheeseburger per month to cover that cost!
Digital Ocean: Switch today, your doctor will thank you!
> its because web hosting companies offer tiered services, and often the lower tiers have limited bandwidth.
Thanks for explaining the obvious. I get the logic behind it. Still, I think bandwidth is cheap nowadays and a website should never have to go offline in case of an unexpected traffic spike. Other hosters reach out to the owner before shutting it down.
In case this helps, you might want to A/B test your homepage with and without coffins, watch the bounce rate. I bounced, the feelings created by my amygdala were too strong and overruled my curiosity or any desire to learn!
I used to do something like this on Downside.com, but only for public companies where you could get the financials from the SEC. Then you have some hard data. Startup Graveyard seems to be listing companies that failed before they even launched. There needs to be some minimum qualification, such as "actually had at least one paying customer".
I'm actually interested in all commercial ideas that got funded. That means there was enough things in place to generate confidence from someone else other than the (possibly deluded) founder(s). Once customer cashflow appears, the business recipe gets a partial validation and the entire context changes. In my view the part before that bears a higher risk and that makes it more important.
It's sad that all the content, source code and other intellectual property is just lost. Does anyone know of a resource that has or lets companies open source their code instead of locking it away forever? Like a farm sanctuary but for failed startups?
Well Patent trolls prey on all those dieing startups, researching the patents they hold to find ones they can buy up and use to sue as many others as they can.
It would be interesting to include (but really hard to do) startups that 'exited' through acquisition that failed to clear the liquidation preferences. This sort of exit could also signal a 'mistake not worth repeating'.
I think a very useful feature for this site is a "I worked here and want to talk about it" listing.
That way, one thing I could do is research my idea / etc for similar startups that failed, call up the people, and ask what went wrong.
The goal here, it seems, is to help you not fail at something similar. The best information will come from those that worked there. So, why not index just enough information to find those people, and then help connect you to them?
Great idea. Site is hugged to death at the moment I think
A quick issue that arose in my head is what happens if people turn away from ideas because they see it in the graveyard? Some ideas may have seen the light of day too soon...timing for startups is important, as shown in this TED talk - https://www.youtube.com/watch?v=bNpx7gpSqbY
With this resource at hand, people might mistaken bad timing for bad idea.
This is cool and it's always smart to check past implementations like this. As long as you learn why the failed not that the idea is impossible. VR is a good example; I didn't see any companies on that list but many gave up on VR because it was too early to be possible until a few years ago. Another good example is digital currency.
Also; you can learn from a comp like Clinkle which should def be added to the list
I like this idea; however, this is very subjective. Laying out 6 solid reasons for _why_ a startup failed is a bold statement. In reality, there usually isn't a concrete reason why, but I guess it's a good exercise to at least think about it collectively.
But there sort of is. They're all just unique. But formulas are suppose to not be unique, so that's the paradox. I think one way to solve it though is with degrees of detail. At the end of the day most companies that succeed just keep making money, and those that fail just run out of cash. So from there, increase the resolution and see what patterns emerge. There will be formulas.
I knew it was too good to be true when I first started using it. Same with VidAngel (basically streaming version of Redbox). Any product that bypasses the bureaucracy/legal requirements of the MPAA/RIAA to give consumers exactly what they want is doomed to failure.
I know this is very tangential, but I hate the design decision to have the hamburger menu mixed in to the graveyard stuff at the top of the site. It reminds me of those hidden object puzzles. It very much prioritizes aesthetics over usability. With more thought, the aesthetics could be included without having to sacrifice usability.
Nearly every image on the site being a ~600x450px jpeg seems wasteful when almost all of these have easily accessible (or at least easy to recreate) SVGs.
It's the flipside of a similar problem in analyzing why companies are successful. [1]
The little disclaimer at the bottom really says it all.
The attempt is a noble one, marred by data and insight quality issues. I think it could be useful if the site can source insightful analysis from founders/insiders and make it easy to search by market/product category. Perhaps even adding a badge to information which came from a founder.
[1] http://www.tomorrowtodayglobal.com/2011/12/09/good-to-great-...