Intuition needs feedback

An architect comparing a bridge model with measured drawings in a sunlit workshop

I trust experience more when it can explain what would make it change its mind. That is where intuition becomes useful in a business: it gives you a starting point, and evidence gives you a way to examine it.

A good architect can see possibilities in an empty space. Proportion, light, movement, the feeling a room might have. Calculations and material constraints then put demands on that idea. They can reveal a weakness, suggest a better solution, or confirm that an initially strange design deserves to be built. I think decisions benefit from a similar exchange.

In discussions about data, people often defend intuition as though measurement threatens their experience. I can understand the reaction. A number presented without context can be a poor substitute for years spent doing the work. But experience also deserves scrutiny. Repeating a decision for ten years gives you ten years of practice; how much you learn depends on the feedback you receive.

If you never find out why customers left, your explanation of customer loyalty can remain confidently wrong. If every successful sale is credited to your judgement and every lost sale to the market, you have made your judgement almost impossible to improve.

Make the belief specific

Consider a manager who believes a new product will sell particularly well in one city. There may be something valuable in that instinct: conversations with customers, knowledge of the area, an observation nobody has yet put into a spreadsheet.

The useful next step is to make the belief specific. Which customers? At what price? Over which period? What would count as a disappointing result? Those questions give the intuition enough shape to test.

A small pilot can then produce evidence that a confident presentation cannot. Perhaps interest is high but purchases are low. Perhaps the product sells well only with a discount that removes the margin. Perhaps the manager saw an opportunity that the existing data missed because the company had never served those customers before.

Each result teaches something different. A single headline number can conceal all three.

I like this way of working because it gives both expertise and analysis a demanding role. The experienced person helps identify what is worth investigating. The analysis has to produce information that could actually affect the decision. Everyone should be able to say which finding would change the plan.

Give the numbers the same treatment

Data can also acquire an authority it has not earned. A model inherits choices about the records it receives, the outcome it is trained to predict and the errors someone has decided to tolerate. Automation can repeat those choices very consistently. Consistency alone tells us little about whether they are sensible.

A recommendation system, for example, may predict which product someone will click. Whether those clicks produce a satisfied customer is another question. If the company cares about repeat purchases, it needs to examine that outcome as well. The objective deserves as much attention as the model.

There are also situations where the data is thin or the environment has changed. Experience can help identify a missing factor, but uncertainty should remain visible. I would rather make a limited commitment, observe the result and retain room to adjust than pretend a precise-looking forecast has settled the issue.

This is why I find the idea of educating intuition compelling. You can write down what you expect, act, compare the outcome with the expectation and revisit the explanation. Over time, you get a better sense of which situations your judgement handles well and where it needs help.

For a founder, that practice is especially useful. We have to act before we know enough. The danger is allowing the urgency of the first decision to become a permanent excuse for avoiding the evidence that comes afterwards.

My preference is to keep a simple record: the belief, the decision it led to, the result, and what I would do differently. It makes a much better conversation possible a few months later. Memory is remarkably good at making our earlier confidence look justified.

Revised from my Medium article of 4 September 2023.