Failure is data, except when it’s not
Aphorism aside — by default, failure isn’t data. If you want to learn from failure, you have to do these three things.
Hey friend 👋
By default, failure isn’t data.
I’m flouting traditional wisdom today. Perhaps it’s because I’m in Paris, and the revolutionary spirit of the French has infected me. Or maybe I’m just a contrarian.
Either way, I come armed with a solution. Here’s how to ensure you always learn from failure.
Read time: 3 minutes.
“Failure is data.”
It’s a common retort to unpleasant news. I’ve certainly said it. It comes from the idea that we can only learn by being wrong before being right. Otherwise, we were right the whole time, and there was nothing to learn.
My honest opinion? Failure is amazing.
It is the foundation of the very science upon which the computer on which I’m writing this was developed, as well as the advances in medicine and healthcare that afford me entering my late thirties with my best years still ahead.
It’s taking us back to the Moon as a springboard to Mars.
But let’s not romanticize failure: it’s uncomfortable, personal, & often public. And, you know what? It should sting, and we should try to avoid it.
Failing sucks!
But therein lies the rub: it’s the only way we can learn.
Unfortunately, we don’t usually learn much from our failures.
It’s not just due to cognitive dissonance, wishful thinking, and post-hoc rationalizations (thought they certainly play a role in the psychology of decision-making).
Mostly, it’s because data is retrospective. It’s proactive.
We get far better data — in quantity and in quality — if we ask the kinds of questions in advance to which need answers, then we do if we just try to interpret a negative result.
In other words, you getter data if you think about failure as failed experiments, and therefore design your behavior as experimentation.
Paradoxically, designing for failure also means we fail less.
It’s a pretty straightforward process. Here are three steps you can follow to fail less, always learn from negative outcomes, and use failure as a ladder of progress.
Step 1: Start with a question.
It’s the simplest thing you can do, but we rarely do it.
When most founders want to try something — a marketing campaign, a social post, a message on a website hero — they do it, intending only to see what happens.
The problem with seeing what happens is that you’ll always succeed — in seeing what happens. But you’ll never learn anything, because you took failure off the table.
You’ll get a number back, but you won’t know what it means, nor how to interpret the result. Is it a good result or a bad result?
Eh… I’ll just go with my gut.
Asking a good question requires specificity:
What’s the one thing you’re testing?
What are the details: who, what, how, and when?
What does failure look like specifically?
Science calls it an hypothesis.
Before setting out to do anything, write down exactly what success looks like, and allow failure to take place.
If you set no bar, you can never miss it, and you never have to face it.
Step 2: Reflect on the failure.
When you fail, ask: “what does this mean”?
The failure is that our expectation was incorrect, but to learn from the miss, we need to dig deeper:
Why did we get the result we did?
What does the change in expectation affect?
How big of a deal is the deviation?
We often won’t know the answers to these questions, but often the next experiment we want to run — the next question we want to ask — comes from this reflection.
It’s not enough to experiment and fail.
If you don’t understand what the failure means, you can’t do anything about it.
Step 3: Transform the data into a change in behaviour.
This is the most important step
Now the part too many skip: what are you going to do differently? Do we pivot or persevere? Do we try something new to achieve the goal, set a different goal, or change the path we’re on entirely? If you don’t do anything differently, there’s no upside, and you’re choosing to fail the same way again in the future.
Bottom line? When it comes to trying new things, never just “see what happens”, because you’ll always succeed…in seeing what happens. But you’ll never learn anything.
This process might sound familiar: it’s build-measure-learn.
Step 1: Measure. Ask a question and run an experiment.
Step 2: Learn. Ask “what does this mean”?
Step 3: Build. Transform the data into action.