Most arguments about data and intuition are staged as a contest. On one side, the spreadsheet: objective, repeatable, easy to defend in a meeting. On the other, the gut: fast, experienced, impossible to audit. Pick a side.

It is a false choice, and an expensive one. Both failure modes are familiar in digital business: the dashboard nobody questions, and the conviction nobody tests. The useful work happens between them, where a number makes you question a feeling and a feeling makes you question a number.

What each side actually is

Intuition is not magic. It is pattern recognition built from experience: many small observations, compressed into a sense that something is off, or that something will work. It is fast because the analysis has already happened, slowly, over years.

That is also its weakness. Intuition is trained on the past you happened to see. It cannot tell when the world has changed, and it feels exactly as certain when it is wrong as when it is right.

Data is not the truth either. It is a record of what someone decided to measure, in the way they decided to measure it, filtered through whatever broke along the way. Every dataset quietly carries the questions of whoever designed it.

A number is an answer to a question someone already asked. The skill is noticing which questions it can’t answer.

A forecast shows this clearly, because a forecast is mostly a structured way of writing down assumptions. Here is the heart of the deliberately simple method behind the Forecasting Playground. Notice how much of it is a choice rather than a fact:

// Project the trend h months ahead, damped by phi.
// phi = 1 keeps the growth going; phi = 0 assumes it stops.
function dampedSteps(h: number, phi: number): number {
  if (phi === 1) return h;
  if (phi === 0) return 0;
  return (phi * (1 - phi ** h)) / (1 - phi);
}

Whether growth continues is not in the data. It is a judgement about the future: intuition, written down as a parameter. The honest move is to make that judgement visible and adjustable, rather than burying it inside a model and presenting the output as a fact.

A conversation, not a contest

The more useful habit is to let each side interrogate the other.

DATA SAYS YESDATA SAYS NOGUTSAYS YESGUTSAYS NOAgreement. Ask whatwould make both wrong.Look for whatyou aren’t measuring.Look for whatthe data can’t see.Stop, unless a cheap testcould change either view.
Four situations, and the question each one asks.

When they agree, the risk is a shared blind spot. Both may rest on the same assumption, and that assumption is exactly where to look. When they disagree, the disagreement is the information. Either the data is missing something experience has noticed, or experience is out of date. Both possibilities are worth knowing, and neither is settled by whoever argues more confidently.

Small experiments settle arguments

The best resolution is rarely more analysis or more conviction. It is usually a small, cheap test that could change somebody’s mind. Build the rough version. Show it to a handful of people. Run the forecast with different assumptions and see which of them would actually change the decision.1

This is where analytical and creative thinking stop being opposites. Designing a good test is a creative act: you have to imagine how you might be wrong, then find the quickest way to find out. Reading the result is analytical. The gut proposes, the spreadsheet checks, and with a bit of luck both of them learn something.

Footnotes

  1. If no plausible assumption changes the decision, the forecast was never the hard part. ↩