Is fail fast a fallacy? It didn’t take long before Microsoft had to take Tay offline. The story made a big splash online, with many news sites writing about it and friends alerting friends on social media. Business Insider wrote that it was “hugely embarrassing for the company.” Wall Street Journal wrote that “the effort to imbue computers with “intelligence” has a history of creating public-relations issues for companies when things go awry.” Was this a good example of “fail fast”?
[bibshow]
The origins of fail fast
Who knows exactly where the concept of “fail fast” originated? A google ngram search shows that the concept of fail-fast first started appearing in books in the mid-sixties. It was only twenty years later that fail-fast started growing. Many of the earlier references to “fail fast” are from the domain of computer science. Fail fast was used as a concept in the synchronization of systems, e.g. a patent from the early seventies regarding synchronization of two computer systems [bibcite key=”citeulike:13988068″] or a scientific review of the transaction commit versus the byzantine synchronization problem from the eighties [bibcite key=”citeulike:13988072″].
Perhaps it was no surprise that fail-fast became the mantra of the software and computer intensive innovation scene of Silicon Valley? Early usage of fail fast was more about learn fast and improve fast. It was about daring to take the leap, even if your business plan wasn’t 95% done. Do 50%, try it in reality and improve to a new 50% level [bibcite key=”citeulike:13988057″]. A point that I am sure Edison would have agreed with:
I have not failed. I’ve just found 10,000 ways that won’t work.
A mandate to fail
Fail fast has evolved since then. A famous quote from Elon Musk, founder of Tesla, illustrates how fail fast has evolved from learn fast to a “mandate to fail”.
On risk: “Failure is an option here. If things are not failing, you are not innovating enough.”
I must admit that I have also been advocating fail-fast as a way to be a successful innovator. The logic is simple, there are more ideas available than we can ever possibly implement. The best way to test these ideas is to try them out in reality. Empiricism is one of the best innovations ever. Innovation is to say no.
Innovations are not lemmings
Source: tao lin on Flickr | CC BY 2.0Do fail-fasters treat innovation and ideas like lemmings. Just set them free and hope they won’t fall over the abyss?
Innovations are not ready-made to be picked up at a fast food counter, as Gene Hughson points out in part 15 of our ongoing blog conversation. Innovations are not lemmings. We cannot just set them loose on the world and hope beyond hope that they will survive. They must be nourished, cherished even. I don’t know what happened at Microsoft but I am guessing that they treated Tay like a lemming and let her make her own way towards the ocean.
@ruthmalan @Grady_Booch I don’t think it was Tay that made the bad choices https://t.co/ZCKd7lHbnx
— Gene Hughson (@GeneHughson) 24 mars 2016
Evolution and nature has shown us that there are two, equally valid, approaches to winning the gene game. The first approach is to get as much offspring as possible and “hope” many of them survive (r-selection). The second approach is to have few offspring but raise them and nurture them carefully (K-selection). Biologists tell us that the first strategy works best in a harsh, unpredictable environment where the effort of creating offspring is low. The second strategy works better in an environment where there is less change and offspring are more expensive to produce. Some of the factors that favour r-selection seems to be large uncompeted resources. K-selection is more favourable in resource scarce, low predator areas. You’ll find a full summary of Life History Theory at Wikipedia.
We are all talking about digital disruption and the ever faster pace of change. One of the common proof points is that the age of listed companies is decreasing. But an interesting fact here is that most companies do not disappear from being closed or bankruptcy, they continue to live on in a different constellation due to mergers and acquisitions. So are the challenges and the volatility big enough that companies should treat innovation like lemmings?
The fail fast fallacy
The Tay setback shows what happens when you treat your innovation like a lemming. Dead within a day. But did Tay had thousands of slightly different inexpensive siblings? No, the Elon Musk style fail fast approach does not really provide the kind of numerous parallel innovation streams that is called for in the r-strategy of reproduction. Would the strategy work if we just put inexperienced people on doing innovation? Perhaps the real truth is that “fail fast” is just lip service and people who become good at failing just go bankrupt?
Forget the cute mantras. No one should ever set out to fail. The key, really, shouldn’t be to embrace failure, but to embrace resilience and the ability to bounce back.
References
[/bibshow]
Part of an ongoing conversation on Innovation
This post is part 16 of an ongoing blog conversation between myself and Gene Hughson. Previous posts are listed by Gene in “Want fries with that?”. Recently, Casimir Artmann also joined the conversation from his blog Disruptive Architecture. You can also join our innovation conversation on YouTube:
Thanks!
Last but not least, thanks Mike Alatorsev for alerting me to this story:
@GregerWikstrand oh it was pricelesshttps://t.co/26zurNadxn
— Mike Alatortsev (@iTrendTV) 24 mars 2016
Image sources
- lemmings: tao lin on Flickr | CC BY 2.0
What is the consequence of a failure? Fail fast may not be an option.
Indeed, properly applied fail-fast would require that the consequences were small for the innovator/parent.
Casimir, would you save we have applied an r-strategy or a K-strategy to our Architecture Corner “project”? I’d say that we are using a K-strategy. Even though we have many episodes, they are all just feedback points on an ongoing learning journey.
I like to use the phrase “safe to fail” to help people feel comfortable with trying something new. Even then, the word ‘fail’ can be too much as it’s so loaded with the wrong connotations in many organisations. Perhaps it would be more helpful to promote “safe to learn” so we can have useful conversations about experiments to try to fill in our knowledge gaps; this article nicely addresses this approach I think: https://www.fastcodesign.com/1663488/wanna-solve-impossible-problems-find-ways-to-fail-quicker
While the term “fail fast” is indeed used in the article, I’m not sure that was really the strategy used: He never really failed – he learned and experimented.
If we use R and K metafor, when is birth of a project? For Architecture Corner concept, I would say K approach, even if we were very unsure about if it would work or not in the beginning. More like a baby a year ago and now we are becomming teenagers.
Publishing short vines about everything that comes across your mind and happens to you during the day is more R -type
Perhaps no one is really in favor of r?
Pingback: NPM, Tay, and the Need for Design | Form Follows Function
The trouble with r is that it’s a crap shoot. You can find better slogans, but not get better at the game. It’s lemmings all the way down. Like every get rich quick scheme, the house wins. The only way to win is by learning and changing the game.
Pingback: Heraclitus was wrong about innovation - Greger Wikstrand