System Thinking

More data or better intuition? What really improves decisions

Mohit Saini · 14 Dec 2025 · 4 min read

More data or better intuition? What really improves decisions

Every few years, a new wave of technology promises that better decisions are just a data set away. First it was ERP, then business intelligence, then big data, and now AI. Storage is cheap, computing power is cheap, and dashboards are everywhere.

So here is a fair question to ask: with all the data we now have, are our decisions in business actually better?

Let us take one simple test. Most consumer goods and auto companies in India have had month-end sales spikes (the “hockey stick”) for as long as anyone can remember. Over the last twenty years, these companies have invested heavily in software, connectivity and data from the distribution channel. Has the hockey stick gone away?

In most companies, it hasn’t. Why?

The data was there. The paradigm didn’t change.

The month-end spike is not caused by a lack of data. Daily sales at the point of sale are visible in many companies today. The spike is caused by the way people are measured: every link in the chain is chasing a monthly target, so stock is pushed in the last week regardless of what the consumer bought.

If the person looking at the data is driven by a target, the data will be read through the target. A dashboard showing that retailers are overstocked does not stop a salesperson who is short of his number on the 28th.

We don’t see data as it is. We see it through the paradigms in our heads.

This is the first trap of data-driven decision making. Wrong assumptions blind us to signals that are clearly present in the data.

Hindsight makes everything look predictable

The second trap is hindsight. After a product flops or a market crashes, it is easy to build a perfect story of cause and effect. All the signals seem obvious. We wonder how the people in charge missed them.

But at the time, the same signals were buried in noise and contradictions. A story built after the event gives us the comfort that, with a little more data, we could have predicted it. That comfort is mostly false.

A simple check: take any forecasting model and use it to “predict” a past period using only the data available before that period. Most demand forecasts at item level fail this test badly. Yet we keep believing that the next tool will do better.

Then why not let the data speak for itself?

If our paradigms blind us, one could argue: let us approach the data with a blank mind. Let powerful computers find the patterns for us.

This has its own problem. Correlation is not cause. Two numbers can move together by accident, or because both are effects of a third thing. Ice-cream sales and drowning cases rise together, but banning ice cream will not save swimmers. (Summer is the common cause.)

There is a more practical problem too. Many of the things that matter most in a business can never be counted directly: a distributor who has lost interest, a salesperson who has stopped prospecting, a supervisor who is quietly holding the team together. We see only their effects, and the effects can mislead.

Slow-moving products are a good example. Imagine an analysis of last season’s sales that flags a set of products as poor sellers and recommends dropping them. The field team pushes back: those products were placed in only a fraction of the stores, and late in the season. The sales data did not measure demand; it measured placement. Given proper placement the next season, some of those “slow movers” could well become best sellers.

So we are caught in a dilemma:

  • Without intuition, we risk seeing patterns that aren’t there.
  • With intuition, we risk seeing only the patterns we already believe in.

The way out: try to prove yourself wrong

Science faced the same dilemma long ago. The philosopher Karl Popper pointed out that we can always find examples that support our belief. Seeing a thousand white swans does not prove that all swans are white, but one black swan proves that they aren’t.

So the scientific way is not to collect evidence that we are right. It is to set up tests that could prove us wrong, and see whether our idea survives them.

In a business, there is a simple way to do this: take your claim to the people who would see its effects if it were true, especially the ones who are most likely to disagree with you.

  • Production says delivery performance has improved. Ask sales: have customer complaints and expediting requests come down? Have emergency shipments reduced?
  • Sales says a new scheme increased demand. Ask the distributors: did their sales to retailers go up, or just their stock?
  • Purchasing says supplier reliability has improved. Ask the assembly line: has it stopped waiting for parts?

If the predicted effects aren’t visible where they should be, the claim needs to be questioned, without anyone feeling insulted. This is the same logic as the predicted effect check in Goldratt’s Thinking Processes: if a cause is real, it must leave other footprints, and we should go and look for them.

Better thinking, not just more data

Data is useful. Good tools help. But the quality of a decision depends far more on the assumptions we bring to the data than on the volume of data we have.

The way forward is not a bigger database. It is a habit: state your intuition clearly, predict what else must be true if it is right, and invite the people who see the world differently to test it.

As the saying goes, not everything that counts can be counted, and not everything that can be counted counts.

Related reading

Mohit Saini

Strategy, supply chain & sales transformation consultant. About · Book a discussion