How to decide with almost no data

Cerno is a private AI boardroom that structures a decision when the evidence is thin — separating what you know from what you're assuming, and scoring confidence honestly rather than pretending to certainty. This guide is the method itself.

Why you don't have data, and why that's normal

Small businesses face first-time decisions constantly. The first hire, the first premises, the first price rise, the first big client. By definition there's no internal history, and the sample size for anything you have done is in single figures.

Meanwhile the advice available assumes otherwise. "Test it", "look at the numbers", "let the data decide" are all written for organisations with enough volume for a number to mean something. Run a test on forty visitors and the result is noise wearing a lab coat.

So the skill isn't collecting data you can't get. It's deciding well without it — and knowing which of your decisions were never going to be settled by evidence in the first place.

Five things that work

1. Borrow a base rate

You have no data. Other people have been here. Reference-class forecasting is the whole technique, and the small-business version doesn't need a study: your own last ten jobs quoted against actual, your accountant who sees forty businesses your size, three peers asked for the number rather than the story, the countable set of businesses near you that tried the same thing and are still doing it.

None of that is your data. All of it beats your estimate, because your estimate is built from a plan you already believe in.

2. Name what would change your mind

Before deciding, write down the fact that would flip you. "I'd go ahead if I knew X." "I'd stop if Y turned out true."

This does two useful things. It tells you what information is actually worth chasing — most isn't, because it wouldn't change anything. And it gives you a tripwire for afterwards: if X shows up later, you've already decided what it means, rather than rationalising around it.

If you genuinely can't name anything that would change your mind, you've made the decision already and are looking for support. Worth knowing.

3. Buy a little information, cheaply

The middle option between "decide blind" and "wait for data" is a small, time-boxed test that produces a real signal. Not a survey of whether people like the idea — a real transaction, a real conversion, a real week of the actual thing.

A pop-up before a lease. One paid project with a supplier before hiring the capability. A single client at the new price before repricing everyone. Each converts an assumption into a fact for a fraction of the committed cost, which is why the same move recurs through the worked briefs.

4. Size the bet to what you can afford to lose

When you can't reduce the uncertainty, reduce the consequence. This is what reversibility is for: a break clause instead of a fixed term, a fixed-term contract instead of a permanent hire, a lease instead of a purchase, a staged rollout instead of a flat change.

You're not making the decision more likely to be right. You're making being wrong cost less — which, when you can't improve your odds, is the only lever left.

5. Decide, and write down why

With thin evidence, the reasoning is the only thing you'll be able to review later. The outcome will be ambiguous — a good result might have been luck, a bad one might have been a sound call — so the record of what you thought and expected is what makes the decision learnable at all. That's the argument for a decision journal, and it's strongest precisely where the data is weakest.

Telling the two kinds of decision apart

Some decisions are data-poor today and will be data-rich in six months. Others will never be settled by evidence.

If data is coming, the question is whether waiting is worth its cost. Waiting is not free: it delays the upside, and in some markets the option disappears. Weigh the value of the information against the cost of the delay, and set a date by which you'll decide regardless.

If data is never coming — should I take this partner, should I go full-time, is this the right name — then waiting buys nothing. The decision is about values, judgement and what you can live with, and the honest tools are regret minimisation, inversion, and making the commitment small enough to survive being wrong. People stall on these for months hoping for a clarity that isn't available, which is the most expensive mistake in this guide.

Common mistakes

Fake precision. Building a spreadsheet with twelve assumptions and reporting the output to two decimal places. The model isn't wrong; the confidence is. If the inputs are guesses, say so and give a range.

Mistaking a small sample for evidence. Two customers said yes, so 40% will. Three weeks of sales is a trend. At small volumes the noise is bigger than the signal, and acting on it is worse than acting on judgement, because it feels justified.

Waiting as a default. The commonest failure isn't a bad decision under uncertainty; it's the decision nobody made, for months, while hoping for information that was never going to arrive. Not deciding is a decision, taken by accident, with a cost nobody writes down.

Confusing confidence with evidence. Feeling sure and having grounds are different. Where evidence is thin, your certainty is coming from somewhere else — usually from having thought about it a lot, which is not the same as having learned anything.

Where a structured board fits

What Cerno does with a thin-evidence decision is refuse to blur the line: known facts and assumptions are listed separately, confidence is scored honestly rather than talked up, and the dissent stays on the record. When you have almost nothing to go on, the most valuable thing a process can do is show you exactly how little — and then help you decide anyway, with the reasoning written down.