Sometimes failure is no success at all

December 16, 2013 5:49 pm

Sometimes failure is no success at all

By Andrew Hill

Mending a dud product is no substitute for getting it right first time

What struck me most forcefully in the profile of Jack Ma, Financial Times person of the year, was not the Alibaba founder’s youth, his love of martial arts, or his against-the-odds subjugation of eBay, once the dominant force in Chinese internet auctions. It was this: what a failure he was. Mr Ma was bad at maths, twice failed his university entrance exam and flopped when he tried to launch an online Yellow Pages.The traditional interpretation of his subsequent upward path would be that the entrepreneur “embraced” these failures, learning from his mistakes and riding his luck to his current success.

But failure is a fickle friend. For every Jack Ma, there are dozens who try to embrace their exam disaster and get a stab in the back, condemning them to obscurity. Even the successful worry: psychologists and coaches say many outwardly confident executives think of themselves as ill-qualified imposters, an insight that seems to divide managers into those who have failed and those who live in fear of failure.

In business, though, failure is a phase through which every great leader and hit product must pass. Facebook’s “Move fast and break things” motto is the corporate equivalent of Samuel Beckett’s dark attempt at hope: “Try again. Fail again. Fail better.”

I applaud the “permanent beta” or “perpetual pilot” approach to innovation, in which products and services are continually improved and adapted to the fast-changing market. But the number of trial “minimum viable products” that reach customers is likely to increase. That is worrying, because what may be acceptable in a virtual world does not always work in real life.

Twitter last week reversed changes to the way users can block followers, following protests from users. It probably took the touch of a button. In certain sensitive areas, patching up websites and software after release is not good enough: think only of the debacle over the early version of HealthCare.gov, the US government’s botched website for people to buy health insurance, or Microsoft’s embarrassing reworking of its Windows 8 operating system.

But the consequences of a premature release for a manufacturer of sophisticated machinery are far graver. As cars turn into “mobility solutions”, serviced by software engineers rather than garage mechanics, the minimum viable vehicle is just round the corner – and with it product recall and liability risks that would give executives atToyota nightmares.

There is a theory that if a company messes up for a customer, then resolves the problem, that customer will be more loyal than those who did not suffer in the first place. But research into the existence of this “service recovery paradox” is inconclusive. It seems to be confined to particular service failures, such as those beyond the supplier’s control. A study in 2000 of airline passengers subjected to delays drew this intuitively obvious conclusion: “Excellent recovery is not an opportunity when compared to the satisfaction resulting from error-free service delivery.” In plain English: mending a dud product is no substitute for getting it right first time.

A clear line exists. Mistakes in the months leading up to the launch of a live product are acceptable. Indeed, a manager should encourage them, because if product developers rule out failure, they will also close down opportunities for great success. But, as Sean Gourley, co-founder of Quid, which analyses patterns in data, points out, while it is “OK to fail on the journey there”, it is “not OK to fail at the end”.

Mr Gourley is describing the second of the Palchinsky principles, named by my colleague Tim Harford, in his book Adapt, after a meticulous Russian engineer who realised innovation should be done “on a scale where failure is survivable”, within “safe spaces”.

The problem comes when managers, lulled by the fact that app developers can advance their work in increments of 10ths and 100ths after launch, assume they can apply the same approach well beyond the web (Column 2.0 not working? Download Column 2.01 or 2.02 with spelling mistakes, grammatical errors and minor prose infelicities ironed out). As software and hardware converge, it is ever more important to point out that some spaces – the ones where customers expect durable, fully functioning products that do not fail fast – are not safe for experimentation.