Hacker News new | past | comments | ask | show | jobs | submit login
The creators of no-longer-with-us products explain what went wrong (37signals.com)
126 points by browngeek on Dec 3, 2010 | hide | past | web | favorite | 44 comments

I always find post-mortems like these a bit strange. These founders are all able to develop some pretty convincing reasons why they failed (albeit in hindsight), yet show no signs of acting on it. I understand the feeling of burnout, I have a handful of abandoned projects myself. I've never been able to look back and say that is precisely why I didn't succeed though. Lessons learned is one thing, identifying the one factor that held you back is a bit of a stretch IMO.

Interesting side note: The first failed company, Verifiable, links to the sixth failed company, swivel, as an alternative. Apparently charting is a tough space.

While I agree with you I find the same to be true of successful companies. When you ask a successful founder why (s)he succeeded you'l most often get an answer aided by hindsight.

It appears to me that nobody truly knows what makes or breaks a company. There are some prerequistes (hard work, focus on customers, etc.) that you have to have, but these aren't enough.

My guess is that it's like the weather and financial markets: Chaos (as in chaos theory) reigns, and it's just not possible to know who the winnere or losers are.

Once you understand hindsight bias, it seems so blindingly obvious. - @haruki_zaemon

Not exactly. I can't find the citation just this second, but there was an article floating about showing that successful founders tend to be much more successful on subsequent projects than the baseline.

I don't think it's chaos or luck. I've argued it before, but there are traits that make successful founders successful. The one that always stands out to me is an absolute laser focus on analytics. These founders are 'obsessively analytical' which allows them to make decisions much faster than others.

Starting a company is a race. Your trying to reach profitability (or at least viability) before your current funding dries up. For boot-strapped companies that can be a matter of weeks. For VC companies that can be years. In each of the cases in the article they lost that race.

So starting a company is largely about managing that runway. Some people get lucky and just get it right. For the rest of us tracking everything and truly being obsessed with that data allows you to see patterns, opportunities, and problems much faster. I suppose that you could say that data-obsession allows you to achieve hindsight much faster.

I have one failed startup (out of three goes). In that one I failed to keep an eye on my data. In particular, I didn't realize just how slowly our development was progressing relative to the amount of runway we had. In the end we released a product that never found a market fit and simply didn't have an opportunity to pivot. We ran out of money, and with it the business went belly-up.

My most recent venture is about 3.5 years old. In that time we've had one massive pivot (think entirely new business that utilized about 20% of the code we developed). We managed that by keeping close tabs on our development schedule and releasing as early as possible. It became clear that our initial idea would work, but the cost of acquiring customers was far too high. We felt like we couldn't optimize that cost down to something reasonable in the time-frame that we had left. We did, however, find a new business model along the way and we've aggressively pursued that all the way to profitability.

Now we're looking at different data. Conversion rates, traffic, and ad-spend rule my world now. We track everything we can think of. We've found patterns for our business that have allowed us to optimize our spend while increasing traffic. We've moved our conversion rates significantly. We still have a long ways to go, but at this point we have a nice war-chest in the bank and are at the point that we can aggressively scale. Along the way we've pivoted, just less drastically. The product has changed significantly as we've come to understand our market. We're pursuing a couple of opportunities that we never even thought about during initial development.

I feel really good about our business. I think we're going to make it. We've done that not through chaos or luck, but measurement and optimization.

You make an excellent point, and to some degree I agree with you. Particularly it's obvious that there are some traits that are needed, or at least up your chances significantly. You point out focus as a vital ingredient.

But there's a flaw to the argument. You can spend all of your time measuring success factors and acting on the data. But there are limits both to how much you data you can make sense of, how much effort you can allow yourself to spend gathering that data, and how you interpret that data.

finding a cofounder, for example, is a critical component of success, but how will you pick a good cofounder based on data? Can you even do that? Notwithstanding that the hardest part about finding a cofounder is actually locating potential candidates to pick from.

In chaos theory you can actually predict the weather if you have data that's exact enough. But you don't. There's just noway you can measure everything exact enough. I think the same applies here.

The good thing about desktop software is if the vendor goes away, the software still works.

Not too keen (as a customer) on putting my time and effort into a SAAS hobby project that doesn't scale.

Given the rate of change in OS and GUI environments, support libraries, and other APIs, most desktop apps don't work very well for long these days either.

Of course unlike SaS you can always continue to use an old version indefinitely in a VM if it really matters.

But I find the longevity of the vendor is still critical for supporting new environments and API requirements, and in practice I don't end up discounting the going-out-of-business risk much for desktop apps vs. SaS apps.

One of my biggest clients is a financial company whose core trading product is based on a commercial product they bought in the 70's (whose manufacturer subsequently went out of business in the 80's).

It's continued to run for 30 years, and they wouldn't dream of replacing it. I don't think they're an exception (although I suspect this is less true when you get out of the enterprise environment).

I should say here that I too have the 'old version' problem: The software to send samples to my Akai hardware sampler only runs on Windows 98. I have a Windows 98 PC just for this!

But you can still run it, ultimately proving zandorg's point.

>The software to send samples to my Akai hardware sampler only runs on Windows 98. I have a Windows 98 PC just for this!

Doesn't it run under emulation, like virtualbox, or even in WINE98 or such?

It also runs on a Mac II vx, with System 7 and 32mb of RAM.

Maybe we need a web framework or platform that focuses on creating portable, trivially deployable applications. So that if the original site goes down, as long as it's open source, someone else can rehost it immediately (or you can just run it yourself from the start, if you know you will be depending on it) without having to spend time reconfiguring it. Server-side installs that are as easy as on the desktop.

The key point is "as long as it's open source", unfortunately most hosted services are not open sourced. If they were, we wouldn't need any special platform.

Yes but I think it's still a significant issue for open sourced web applications -- there's still room for improvement. It's usually a lot of extra work to make deployment on new systems completely automated, so developers don't bother. For my last Django project, I have a Fabric script that will deploy an update to the server with one command -- that was worthwhile overhead to invest time into, since it will continue to save me time. But I had little incentive to spend time on automating the initial deployment, since it's something I'll probably only do once, so doing it piecemeal by hand wasn't much of a problem.

I've seen projects where this kind of environmental coupling led to the developer not even bothering to open source their work, since they knew it would take too much for others to use the code.

It's usually a lot of extra work to make deployment on new systems completely automated, so developers don't bother.

Deprec + Capistrano makes this very close to a reality for Rails. It has some niggles -- like installing virtually everything from tarballs when I would strongly prefer it apt-get everything so that I could get updates and so paths and configuration files jelled well with what Ubuntu expects -- but it has been, and continues, to be a lifesaver for me in getting new projects up to speed.

You can also do one step deploys and automate a lot of cruftiness. (I can do upwards of forty deploys to my staging server in a day when in development mode on Twilio integration -- which is hard to test without actually having an accessible HTTP server and making phones actually ring -- so I have a script which commits my changes, tags a new staging release, and pushes it to the staging server in one command.)

Isn't that why virtualization was invented?

"Vitamin not a pain-killer" is a great metaphor.

No, it's a terrible metaphor. The vitamin business is hugely successful! I did the research once and vitamins and supplements actually outsell pain relievers.

[ Edit: supporting research I did, which I've now cited twice in comments on Hacker News. I should probably write a blog post. http://news.ycombinator.com/item?id=499929 ]

I think the point wasn't so much in how successful both industries are, then he could've taken any two industries. The point was that one is very much needed, and the other is nice to have. It's hard to get people to pay for 'nice to have'.

Huh? The guy you responded to shows exactly the opposite, to the tune of vitamins being a larger industry.

Naturally--one takes a vitamin every day, but a painkiller only when in pain.

In terms of patentability and profits too?

given -R&D, seems logical that it may be the case.

That's the money quote of the whole article.

The Vitamins business is big, true, but absolutely dwarfed by the pharmaceutical industry.

You don't need salesmen, multi-level marketing or flashy advertising to sell aspirin. The only aspirin advertising is between different aspirin brands, not to convince people they need it at all. Take one, headache goes away. It's a binary test = it either works or it doesn't. That's how to build a simple business that sells. Your software should be an aspirin not a multi-vitamin. It should be self-evident on the first try how it solves a headache.

I've driven in my car late at night to out-of-the-way shops to buy aspirin. I've never bought a vitamin in my life.

you have no idea how the pharmaceutical industry works. by now there are tons of generic versions of aspirin out there, plus there always have been alternatives like thomapyrin, etc out there.

yet you know, recognize and buy aspirin at a mind-boggling mark up prize, not even recognizing that bayer has stopped producing the stuff by themselves and that it comes out of the same factories like its generic counterparts.

and why is that?

bayer still invests heavily in marketing and sales. aspirin is the most "researched" drug as there are countless phaseIV studies going on all the time. which are just a legal way of paying docs to push the product to patients. then you have key accounters and pharmacy-reps (depending on the market) out there, influencing big chains (USA) or tiny pharmacies (Europe) to stock, display and recommend the product.

never underestimate the power and smartness of big pharma when it comes to sales and marketing.

i make my living implementing crm systems for them worldwide - these guys are never sitting on their hands.

No you missed my point.

Even the generic aspirin manufacturers can sell the stuff with little or no advertising.

The vitamins business has to spend a lot of money convincing to buy people stuff, the aspirin business doesn't.

The linked interview with Swivel was interesting. I'm attempting a data visualization startup. But I've found a niche where customers have a burning need and money but lack the IT resources to implement their own solution. It's not enough to draw charts; Excel can already do that. You have to provide everything customers need, from fetching the data to analyzing the data. And it's still a tough sale because SaaS is newish and people are uncomfortable handing off critical business data. I can think of a dozen other niches where paying customers need data analysis and visualization.

I'm also doing a data visualization startup (link in profile). What's yours? Email me if you care to chat.

You have some solid competition in this space: http://www.tableausoftware.com/

If prettygraph.com is based in India his main advantage would be that there's less chance of him immediately caving to U.S. government censorship like tableau just did.

Based in the UK.

Well, I'd be flattered to be considered competition to Tableau. They are a well-funded enterprise software company. I'm just a little guy in a garage :-P

An even larger list of more in-depth post-mortems: http://foundersblock.com/featured/25-best-startup-failure-po...

Did Storytlr even bother asking people to pay for the service? Seems kind of strange that with 10,000 users you can't at least cover $500 in hosting.

At $5 a month, that's only .1% signup required.

I've seen the if only we get (tiny percentage) of (huge number) to pay (seemingly small amount), we'll be rich! arguement framed about a thousand different ways. I actually cautioned an entrepreneur I was talking to last night about this.

Sales start from 0, and they continue from there. Until you have sold it a few times, who knows if .1% is incredibly easy or impossible. Sometimes any number >0 just isn't going to happen.

I didn't say rich, I said pay the $500 hosting fees.

The fact is that you guaranteeing sales are $0 if you never ask for money.

How hard could it be to get 1%?

I mean if only ask for $1, and get that dollar from only 1% of China, I'd be a millionaire!

You would be surprised at how few people would pay $5 a month...

That's the truth. I think it's better to replace any arbitrary small monetary value with 'Will Actually Open Wallet'. Once you've got the perceived value of someone willing to open their wallet then other things can start to fall into place.

Yep, indeed. We just need to train users to pay for good software. I used to be in the "I don't pay for software, ever" camp, but then I realised people are worth my support and now buy things regularly (especially Android apps, even if the pro version does nothing more than the free version, e.g. Twidroid).

“He forgot there was a number lower than one percent.”


Well, the number I used was .1%, and that number is surely always 0 if you don't even ask.

They didn't even ask anyone to pay.

Did you read Swivel's story? They had many many more users, they did ask them to pay and convinced less than 10. That's a lot less than 0.1%.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact