This week in The Legibility Letter:
• The brewer’s rule: modern statistics was invented by a brewer working with three samples of barley — it exists because someone couldn’t afford to wait for more data (and you’re not going to get more either)
• From the inside: starting something turns into an obligation to maximise it — full calendars, sessions taken to the limit, projects we keep pushing out of habit (sometimes you don’t need to finish the bottle)
• A quote we like: I convinced myself an initiative would work and started reading every signal in its favour — the problem isn’t what the data says, it’s when you stop reading it and start choosing it
The brewer’s rule
In 1899 Guinness hired a twenty-three-year-old fresh out of Oxford. His name was William Gosset and his job was to make every pint taste the same.
That meant measuring the barley, the hops and the yeast. Measuring cost money, so the samples were small. Sometimes three.
The statistics of the day were built for large samples, and with three they didn’t work.
Three samples of barley
Gosset had two ways out. Wait for more data, which was never going to happen because analysing barley cost money. Or build new mathematics that worked with what he already had in front of him.
He did the second.
In 1906 Guinness sent him to London for a year, to Karl Pearson’s laboratory, to solve the problem. In 1908 he published the result: a method for working out how much you can trust a small sample.
It’s the t-test, and it’s still used today in medicine, in physics and in every laboratory in the world.
Guinness gained something very concrete. They could decide from three samples whether one barley was better than another, instead of buying blind or paying for enormous trials. Gosset went on to build the brewery’s statistics department.
And the method doesn’t carry his name. Guinness banned its employees from publishing because it didn’t want other brewers knowing how it worked. So he signed as Student, and who he really was only came out when he died, in 1937.
They hid it because it worked.
Modern statistics was invented because someone couldn’t afford to wait for more data.
You’re not going to get more either
A small business doesn’t generate thousands of cases. It generates five lost clients, three rejected quotes, seven sales conversations a month. Those are your numbers and they’re going to stay your numbers.
And the sentence that always turns up is the same one: we don’t have enough data to decide yet.
It sounds prudent. In practice it means postponing the decision until a month that isn’t coming.
The trap
Three samples can lie too.
If three clients tell you your price is too high, your price might be too high. Or you might have spoken to three tight-fisted clients. The same three data points, two opposite conclusions, and from outside they look identical.
That’s exactly what Gosset worked out: how often three samples mislead you.
Which is why the second step matters as much as the first. You decide with what you have, and you write down beside it how sure you are.
Think about the decision you’ve been putting off for months for lack of data. Write down exactly how many data points you have. If it’s five, it’s five.
And here’s the part that usually goes unseen. The data you’re waiting for doesn’t arrive on its own. The decision produces it.
You won’t know whether that price works until you charge it. You won’t know whether that channel works until you put a client through it. Waiting doesn’t gather information, it prevents it.
Gosset didn’t get more samples by waiting. He ran more tests.
Decide with five. And write down beside it how sure you were.
In six months you’ll have two things you don’t have today. A result, and a sixth sample.
The bottle
I grew up partly in Nigeria. My father was Nigerian, my mother Swedish. The standard large bottle of beer in Nigeria is 60 centilitres, bigger than anything you’d normally get in Europe.
I remember watching my father, when I was young, not finish the bottle. It struck me as odd, so I asked him why. And he said something that stayed with me for decades.
“Sometimes you don’t need to finish the beer.”
He didn’t make anything of it. It wasn’t a lesson. It was simply how he saw things. But that sentence has travelled with me my whole life, and over the years I’ve understood that what it held went a long way beyond a bottle of beer.
There’s something deeply lodged in us that says if you start something, you have to finish it. If you open something, you empty it. If you start a project, you take it all the way. If there’s time in the calendar, you fill it. If you’re training, you go to the limit.
Starting turns into an obligation to maximise.
And that obligation takes us to places we don’t need to visit. Calendars full of things we don’t need to do. Sessions that wreck us when what the body was asking for was to move gently. Projects we keep pushing when instinct is telling us they’ve already given what they had to give. Working relationships we hold on to because we started them, not because they still serve.
There’s something freeing in understanding that not everything you start needs to be maximised. That you can taste something for the taste and leave the rest. That you can fill your day with less and have that less be enough. That you can train three days this week instead of five, and that isn’t weakness, it’s intelligence.
My father didn’t leave beer in the bottle out of discipline or self-control. He left it because he didn’t need it. There was no internal struggle, no willpower involved. There was simply enough, and enough was enough.
That’s the hard part to learn. Moderation as clarity rather than as virtue. Knowing when you already have what you came for, and being calm enough to leave the rest.
Sometimes you don’t need to finish the bottle.
Edward
“It is a capital mistake to theorize before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts.”
— Arthur Conan Doyle
I like this quote because it points at a self-deception that turns up constantly in business: falling in love with a hypothesis and then bending reality to prove yourself right.
We often think we’re deciding with data.
But usually we already have the conclusion before we look, and we use the data to confirm it.
We pick the metric that flatters us, ignore the signal that contradicts us, read a bad result as “it’s still early”. We’re not reading reality. We’re negotiating with it.
And that’s more dangerous than having no data at all, because it gives you the confidence of believing you’re deciding on evidence when you’re deciding on what you want to be true.
I’ve done it. I convinced myself a new initiative would work and, without noticing, started reading every signal in favour of the idea.
Early signs became promising. People not moving became a lack of vision.
I wasn’t lying to anyone. I was lying to myself, one favourable reading at a time.
The problem isn’t what the data tells us. It’s when we stop reading it and start choosing it.