Fail, to learn

“Ya gotta fail fast”, he said, striding back and forth with intense energy, from up in front of the podium. “If ya gonna be Agile, ya gotta learn ta fail fast“.

But… I do that all the time already, don’t I? I mean, I fail lots. I’ve been failing to write this blog-post for two days now: how could failing to do it faster actually help in this? All it’d do is get me even more depressed about it, even more quickly…

Which might kinda suggest there’s a key concept missing in that whole ‘Agile is about failing fast’ meme?

So here it is: it’s not about failing, it’s about learning.

What we really aim to do in Agile is to test the boundaries. If something breaks, it means that we’ve gone somewhat too far beyond some kind of boundary. It might be our own boundary – our lack of skill; or it may be something about the world ‘out there’. And we need to know the difference, know what it means in practice in our context, know what to do about it. Either way, we’ve just had some potentially-useful feedback from Reality Department – something from which we can learn. And yes, doing all of that , faster, quite probably is a good idea. Yet it’s not actually about the failure alone: it’s not actually about ‘fail fast’ as such, is it?

In an unknown context, a ‘failure’ is a crucial signal-event that we can use within a sense / make-sense / decide / act loop. If we’re trying to find out which way to go, which way to push our skills, then yes, we do kind-of want to fail – but only in the context of learning, and for the purpose of learning.

In terms of formal-theory, it’s a direct corollary of the crucial distinction between finite-games and infinite-games. In a finite-game, we play to ‘win’ – and failing to ‘win’ is a disaster, a cause for shame, self-retribution, and worse. And there’s always another finite-game, so all that ego-laden angst is all a bit pointless anyway. But in an infinite-game, we play to learn – and James Carse argues that, ultimately, there is only one infinite-game, called ‘Life’. The Agile meme only makes sense in context of an infinite-game, a game-to-learn – but in business especially, too many people still try to force it to fit the finite-game delusions of ‘the market’, and more. Which is why, too often, it just doesn’t work…

In essence, then, that whole meme of ‘fail fast’ is about speeding up the rate at which we push the boundaries, the rate at which we ‘fail, to learn’ – but the comma in that phrase of ‘fail, to learn’ is crucially important. If we drop the comma, what we get is ‘fail to learn’ – we learn nothing, and hence, in all probability, keep failing. Not a good idea…

So note that crucial difference: ‘fail, to learn’ is not the same as ‘fail to learn’. As so often in enterprise-architectures and the like, a subtle distinction, but horribly important. You Have Been Warned, perhaps?