kaashif's blog

Programming, with some mathematics on the side

Don't fail fast, learn fast

2022-02-04

"Fail fast" is one of the most quoted pieces of startup advice, but it's also the most misunderstood. The important part isn't the failing, but what comes after. And I guess doing it fast is pretty important too.

"Learn fast" is much better. The funny thing about learning is that you can actually learn what will fail before trying it. So there's no need to even fail at all to learn - just pick up a book and learn.

I didn't do that, instead I went for the actual failure route with my startup. This post is about how I could've avoided that by picking up a history book.

Let's go on a trip all the way from 8th century Asia to the US in the 80s and 90s, and look at some business failures I could've learnt from without re-enacting them. Maybe this will help you learn faster by just not doing what I did.

But wait, isn't "fail fast" really popular advice?

Yes! And it's not wrong, it's just really misunderstood. If you can know it's going to fail, it's better to not do the thing rather than doing it and failing. I guess easily misunderstood and misapplied advice actually is bad advice.

The goal is to learn fast, and sometimes the quickest and cheapest way to do that is by doing, failing, and iterating. But if there's a startup that did your exact idea and failed, you don't need to do it to know it's going to fail. You don't need to sell pet food online at a loss to know it's not going to work, just look at Pets.com.

There's a summary of some advice from YCombinator here, notice that "fail fast" actually isn't there, probably because a startup accelerator knows that once you tell people to "fail ...", it's more likely they'll fail and ignore whatever subtle interpretation you imposed on the "...". Going out and aiming to fail (plus some adverbs) is not the best plan.

Avoiding failure by simply learning from history is the core premise of a book I just read, In Search of Stupidity: Over Twenty Years of High Tech Marketing Disasters by Merrill R. Chapman (it's not just about marketing, I promise!). I highly recommend it and wish I had read it sooner. Even if I had, there is some chance I would have arrogantly dismissed the lessons within without living them myself, so maybe it wouldn't have helped.

My payment startup's failure has several parallels in history, all of which seem to demonstrate mindblowing stupidity (remember the title of that book). Ignore the implication that you're reading a blog post written by a mindblowingly stupid person. Knowing about these examples would've let me learn without needing to fail. Fast or slow.

Before looking at some specific cases, there's one thing I want to make clear:

Are you special? Probably not.

Coming up with a new way to fail is pretty hard, so if you fail, you'll have two options for introspection:

  1. I came up with a new way to fail, unheard of in the thousands of years of human history. I am special among untold billions of people. Perhaps even superior. I should write a book about it.

  2. I made a stupid mistake which someone else has made before and I could have predicted it if I had done enough research.

1 is unlikely. But it does make you feel good about yourself. 2 is much more likely. And yes, there's some in-between, but forget about that.

This isn't to say that you can't discover new ways to fail, but that's probably not what's going on. If you fail, try harder, read more history books (In Search of Stupidity has a list of recommended books), and you'll almost inevitably realise you were re-enacting a 9th century cloth dealer's bankruptcy proceedings.

Most of us realise we're not special very early on, but I think it takes a bit longer for aspiring startup founders. That's good in that founders are optimistic and try things without realising how hard they'll be, which spurs growth. It's bad in that founders might be more prone to the Dunning-Kruger effect and making stupid mistakes.

This is often more of a problem for young founders. See Paul Graham's essay on schlep blindness:

So the reason younger founders have an advantage is that they make two mistakes that cancel each other out. They don't know how much they can grow, but they also don't know how much they'll need to. Older founders only make the first mistake.

Being arrogant/optimistic enough to think you can make a startup work is common in young people, but often leads to failure. Lack of knowledge about history and a lack of experience are also common in young founders, but those don't have many positives and you can work on those.

Remittances in 8th Century Asia

Here's the meat of this post: my failure and how I misinterpreted "fail fast" when I could've just learnt fast instead.

My failed startup was a money transfer service. We would have taken money from you in one country, and sent money to your recipient somewhere else using only domestic transfers. Domestic transfers are much faster, cheaper, and we could offer better FX rates. We'd do one international transfer every day to rebalance the accounts.

We shut it down before serving any customers. That kind of transfer system would never have been approved by the Financial Conduct Authority in the UK, at least not as we planned to implement it. Why? And how could I have learnt that before trying it?

I could've looked at history: such a money transfer system was invented over 1300 years ago. So I guess we were beaten to market by a few years.

Most of the information I'm about to summarise comes from this report by the Financial Crimes Enforcement Network (uh oh) entitled "The Hawala Alternative Remittance System and its Role in Money Laundering" (oh shit).

Hawala is a way to transfer money over long distances without needing to actually send money using international transfers. Fast, cheap, it's exactly what my startup was supposed to do.

The system works as follows:

  • The sender sends 100,000 rupees to a broker, S. Both are in India.

  • S calls another broker, R, and tells them to send $1000 to the recipient in the US. Both broker R and the recipient are in the US.

  • R sends the money to the recipient.

Some advantages:

  • This can be done very quickly, as fast as it takes to make a phone call or send an email. This is much faster than an international bank transfer in some cases.

  • The exchange rate can be set favourably - banks have famously bad exchange rates.

  • Fees may be lower than dealing with banks.

  • No need for a bank account on either side, both can use cash.

All great. But this system has existed for thousands of years, right? And presumably it incorporated the telegraph, then the phone, and now the internet, so why isn't this system widespread? Whatever the reason is, a remittance startup like ours would likely face the same problems as hawala brokers.

The reason is simple: this system is so easy to use, so fast, so frictionless, that it makes money laundering, terrorist financing, and tax evasion a breeze. Hawala is illegal in many countries for those reasons. In particular, record keeping is crucial:

Hawala transfers leave a sparse or confusing paper trail if any. Even when invoice manipulation is used, the mixture of legal goods and illegal money, confusion about valid prices and a possibly complex international shipping network create a trail much more complicated than a simple wire transfer.

[...]

In the final stage of money laundering, integration, the launderer invests in other assets, uses the funds to enjoy his ill-gotten gains or to continue to invest in additional illegal activities. The same characteristics of hawala that make it a potential tool for the layering of money also make it ideal for the integration of money. This is when money seems to become legitimate, and, as we have seen, hawala techniques are capable of transforming money into almost any form, offering many possibilities for establishing an appearance of legitimacy.

-- The Hawala Alternative Remittance System and its Role in Money Laundering, p12-13

These dangers are not theoretical. Here's an excerpt from a report from the National Commission on Terrorist Attacks Upon the United States. If that sounds familiar, that's because it's also known as the 9/11 Commission. Yeah, this is serious.

How did al Qaeda move its money?

Before 9/11 al Qaeda appears to have relied primarily on hawala and couriers to move substantial amounts of money for its activities in Afghanistan.

[...]

Al Qaeda moved much of its money by hawala before 9/11. In some ways, al Qaeda had no choice after its move to Afghanistan in 1996; the banking system there was antiquated and undependable. Hawala became particularly important after the August 1998 East Africa bombings increased worldwide scrutiny of the formal financial system. Bin Ladin turned to an established hawala network operating in Pakistan, in Dubai, and throughout the Middle East to transfer funds efficiently. Hawalas were attractive to al Qaeda because they, unlike formal financial institutions, were not subject to potential government oversight and did not keep detailed records in standard form. Although hawaladars do keep ledgers, their records are often written in idiosyncratic shorthand and maintained only briefly. Al Qaeda used about a dozen trusted hawaladars, who almost certainly knew of the source and purpose of the money. Al Qaeda also used both unwitting hawaladars and hawaladars who probably strongly suspected that they were dealing with al Qaeda but were nevertheless willing to deal with anyone.

-- Monograph on Terrorist Financing, p24-25

This information is publicly available, thus our regulatory difficulties were easily predictable - we just hadn't done the research.

If:

  • Your business model is obvious
  • Everyone you've spoken to wants to be your first customer
  • It's not all that hard to implement
  • There's no-one out there doing the same thing

Then it's too good to be true. Maybe you should try it to discover what you're missing, but success is unlikely without a complete understanding of why there are no competitors beyond "we're geniuses and thought of it first". The easiest, fastest, and cheapest way to get that understanding is to pick up a history book.

Big Blue, Star Trek, and OS/2 Warp

I know what you're thinking: this guy is an idiot. There's no way a real company could ever do something so stupid as building a product and having to backpedal after discovering it could actually be illegal. Wrong! Enter IBM and OS/2 Warp. Most of the following is a summary of the relevant chapter in In Search of Stupidity.

OS/2 was an operating system that was supposed to replace DOS, and later act a direct competitor to Windows. OS/2 1.0 was released in December 1987, and it was mostly a flop. Expensive, bloated, and with no applications, that was hardly a surprise. But none of this resembles what happened to us. The shockingly stupid part comes later.

It's 1995. 3 years previously, IBM had released OS/2 2.0, a technically impressive, reasonably priced product (or at least, people were used to the unreasonable price), with some (but not that many) applications. OS/2 3.0 was about to be released and would presumably be a great product.

OS/2 3.0 came with a name change: it was going to be called Warp, as in warp speed, as in Star Trek. Great name! Preceding beta versions had been called Ferengi, Borg, Klingon, so the name fit. One problem: no-one at IBM had bothered to check with Paramount, the owners of all marketing rights and trademarks relating to Star Trek.

What ensued was embarrassing - IBM got all kinds of legal threats and had to walk back its entire advertising campaign. It didn't have to rename the product, but "warp" had to refer to something other than speed, and no references to Star Trek were allowed. It was a disaster. People were talking about "warping" their computers in ads, and no-one dared talk about how Warp was fast or futuristic.

How is it even possible for someone to build an entire product and market it to customers, all without checking the basic legality of what they were doing? In IBM's case, it was just complete incompetence. In my case, it was...actually also complete incompetence.

IBM not only didn't learn fast, they didn't really learn at all, and now are essentially irrelevant.

Why does this keep happening?

If IBM, with its legions of lawyers, failed with OS/2 in a vaguely similar way to an inexperienced first time founder, is anyone safe? Could anyone suddenly discover that their business model, product, or marketing campaign is actually illegal? Why does that happen?

On my end, it's simple. Young, inexperienced founders are arrogant, stupid, and largely incompetent. It takes real effort and introspection to stop being those things, but it can happen - complete and utter failure will at least get you to stop, think, and learn.

For large incumbents, it's tougher to work out. My impression is that they can't really fail in the same way a startup can. There can be failures and layoffs, but those happen over many years, and none of the decision makers are going to go hungry if their decisions are wrong. There's no feedback loop. In In Search of Stupidity, the author advocates splitting large companies up, not as an anti-trust enforcement measure, but as a voluntary program to prevent the decline into senescence and impotence we saw with IBM.

If you're running a small startup, it's easy to drop the arrogance and try to learn from history. At a large organisation, it seems impossible to drop institutional arrogance, there's little incentive for individuals to do better.

Step-by-step guide to success

  1. Read lots of history books
  2. Start a startup!

I'm a real believer in this two-step program and will try it at some point. Feel free to steal it and try it yourself!