It's Better to Fail Fast and Often

You don't have to be in an innovation meeting very long before someone says, "let's just test it." Fail fast, someone says, usually with a slight note of relief, because it sounds like rigour and feels like permission in the same breath.

What's actually being tested, though, is almost never the thing that decides whether the idea works. It's what it is wrapped up in. The channel, the price, the packaging, the message. The question underneath, whether anyone genuinely needs this in the first place, stays exactly where it was: assumed rather than tested, because testing it properly takes longer and sometimes returns an answer nobody wants.

That's the part of "fail fast, fail often" worth taking apart. Not the speed. The target.

The doctrine had a seatbelt

"Fail fast" has a real, traceable origin as management doctrine, not just a Silicon Valley mood. Rita McGrath's 2011 Harvard Business Review article, "Failing by Design," laid out a disciplined set of practices for spotting a bad idea before it reached full-scale development or launch. Buried inside it is a principle that rarely survives into the version everyone actually quotes: contain the downside risk, fail cheaply. The implicit promise to the people running the experiment was that a failure caught quickly wouldn't cost them their job or their standing.

That's the whole story in one image. The doctrine was born with a seatbelt built in. The seatbelt is the part that got forgotten. What travelled into company culture was the permission to fail. What quietly dropped out along the way was the requirement that the failure be small, cheap, and genuinely informative. In fact, it's often used as an excuse for lazy innovation that misses the important work entirely.

What the research actually shows

Erwin Danneels and Alex Vestal ran a longitudinal study of 106 US manufacturing firms, published in the Journal of Business Venturing in 2020, that tested two distinct ways organisations respond to failure. One they called normalising: a culture that simply tolerates and accepts failure as inevitable. The other they called analysing: a culture that deliberately dissects why something failed, through structured postmortems and genuine lessons-learned reviews rather than a quick shrug and a move to the next sprint.

The finding is the single strongest piece of evidence in this whole debate. Mere tolerance for failure has no statistically significant effect on how innovative a firm's products actually are. Tolerating failure, on its own, predicts nothing. What predicts more innovative products is the analysis discipline. Firms that fail and then interrogate why do better. Firms that fail and shrug do not.

Amy Edmondson's work gives the clearest language for why "fail fast" collapses so easily into an excuse. In Right Kind of Wrong, she distinguishes three kinds of failure. Intelligent failures are genuine experiments in new territory: real uncertainty, a real step toward a goal, informed by the best available prior knowledge rather than a guess, and kept as small as possible while still teaching something. Basic failures are preventable errors in familiar territory, usually traceable to one cause, most often a step skipped or care not taken. Complex failures sit somewhere between the two, with several causes combining in a familiar setting.

Edmondson's own verdict on the slogan is direct. "Fail fast, fail often" borrows its legitimacy from intelligent failures, but glosses over the rest of the failure landscape it's actually describing. In practice, a great deal of what gets waved through under the fail-fast banner is a basic failure wearing an intelligent failure's clothing. Nobody skipped a step out of malice. They skipped it because skipping it was fast, and fast is what got rewarded.

Worth one clean distinction here, because it's the one most commonly blurred. Edmondson's earlier and better-known research on psychological safety is about the safety to report a problem early, to say out loud that something isn't working before it gets worse. That is a genuinely well-evidenced finding. It is not the same claim as "failure is inherently valuable," and treating the two as interchangeable is exactly the kind of overclaim that let the slogan drift so far from what the research actually supports.

None of this says fast experimentation is a bad idea. It says the thing being tested has to already be informed by real knowledge of the customer for the failure to be intelligent rather than basic. Which raises the obvious next question: what happens when the experimentation is completely genuine, fast, cheap, and well run, but it's aimed at the wrong layer of the problem entirely.

Subscription shaving

I ran an analysis earlier this year across thirteen major studies of product failure, spanning 1968 to 2024. No real consumer need was cited by eleven of the thirteen as a leading cause of failure, ranked first since the very first study on the list. Detailed market studies are still skipped in more than three-quarters of new product projects, according to Cooper's data. Everybody knows it, but doing the work to uncover a real consumer insight is so much harder than running a quick experiment.

Dollar Shave Club's 2012 video is usually told as a marketing story, and on its own terms it's a good one: a $4,500 shoot, twelve thousand orders in forty-eight hours. But look at what the video was actually testing. Not a better razor. A narrow, specific hypothesis: that men were frustrated enough by Gillette's pricing and its locked display cases to switch to something cheaper and less annoying. That's a genuinely well-run intelligent failure by Edmondson's own definition, small, cheap, informed by a real frustration, and it worked. The insight was real, and the test correctly found it.

The mistake, looked at closely, sits after the launch, not before it. Once the video had validated the price and access insight, Dollar Shave Club kept doing more of the same kind of thing that had worked: fast, cheap experiments on the wrapper. Butt wipes within a year of launch, then toothpaste, cologne, deodorant, new subscription mechanics. What it didn't do, in any visible way, was keep building the deeper product and brand insight that might have given it something to stand on once the price gap it had built itself around got closed. Harry's, launched the following year with the same subscription mechanic, did the opposite. In 2014 it bought Feintechnik, a German blade manufacturer, specifically to control the quality of what it was selling, and positioned itself on premium design at an affordable price rather than simply being the cheapest option available.

Gillette closed the price gap in 2017 with a 12% cut, the first significant one in its history. For Harry's, that mattered less, because the differentiation was never only about price. For Dollar Shave Club, it mattered enormously, because price and convenience were most of what the company had ever actually offered. Unilever bought Dollar Shave Club for around $1bn in 2016. Within a month, US sales had gone flat and stayed flat for more than a year; retention had already been sliding since 2015, and Harry's kept taking share. By 2023, Unilever quietly sold 65% of the business to a private equity firm for an undisclosed sum, a steep comedown from the price it paid, now being described in the press as a fallen unicorn attempting a comeback under new ownership.

A purer version of the same mistake, with none of the early success to obscure it, is a company called ShaveMOB, which launched around the same period on a single pitch: cheaper than Dollar Shave Club. No manufacturing investment, no product story beyond the discount, nothing to test but the price point itself. It never grew into anything, and it remains a marginal, barely visible player today.

None of these companies skipped experimentation. Dollar Shave Club, in particular, ran an enormous number of fast, cheap tests over the following decade. The failure was in what those tests were pointed at. Fail fast worked exactly as designed, right up until the moment the question that mattered stopped being which wrapper performs best and became something slower and harder to test: is there a product here good enough to be worth a premium once the price advantage disappears. Harry's asked that question early and built an answer. Dollar Shave Club kept testing wrappers.

Move fast and break things

"Move fast and break things" was Mark Zuckerberg's internal motto, on the record from a 2009 interview: unless you are breaking stuff, you are not moving fast enough. It lasted as company doctrine until 2014, when it was retired in favour of the considerably less quotable "Move fast with stable infra." The stated reason was straightforward. Engineering time spent fixing self-inflicted bugs had started to outpace whatever speed the company was gaining by moving fast in the first place. It's a clean, unforced echo of the Danneels and Vestal finding from inside the one company most associated with the ethos: tolerance without analysis doesn't compound into anything. It just accumulates bugs, and eventually someone has to stop and pay for them.

The question worth asking before the test

Before calling something an intelligent failure, two questions are worth asking plainly. Was the thing that got tested actually informed by real knowledge of the customer, or did the team skip straight to testing the wrapper, the channel, the price, the packaging, the message, because that's the part that's fast and cheap to test? And if it failed, was there a written account of why, one the next team would actually read, or did everyone simply move on to the next sprint?

If the real answer to both is yes, the failure earns the label, and the culture around it is doing genuine work. If the real answer to the first is "we never actually checked," then the speed was real, but it was speed applied to the wrong layer of the problem. That's a more common mistake than recklessness, and probably a more useful one to name, because it doesn't require accusing anyone of laziness to be true. Most teams testing the wrapper are doing the part of the work that's visible, measurable, and fast, and quietly skipping the part that's slower and harder to defend in a steering meeting.

So next time someone reaches for the phrase "fail fast" in your own building: ask the question, fast at finding out what, exactly, and was that ever the thing you actually needed to know?

Sources: Rita McGrath, "Failing by Design," Harvard Business Review (2011); Erwin Danneels and Alex Vestal, "Normalizing vs. analyzing: Drawing the lessons from failure to enhance firm innovativeness," Journal of Business Venturing 35(1), 2020; Amy C. Edmondson, Right Kind of Wrong: The Science of Failing Well (2023); Financial Times, Marketing Daily, Boston Globe and CNBC coverage of the Dollar Shave Club, Harry's and Gillette pricing history (2015-2018); Cosmetics Business, Global Cosmetic Industry and Forbes coverage of Unilever's sale of Dollar Shave Club to Nexus Capital Management (2023-2025); Mark Zuckerberg, Business Insider interview with Henry Blodget (October 2009) and subsequent company statements on the retirement of "Move fast and break things" (2014).

Article 1 of the "But Is It Though?" series | redebuter.com

#Leadership #Innovation #Strategy #FMCG