Skip to content

Confusing correlation with causality

"It is all connected, isn't it?"

Confusing correlation with causality

Some time ago, there was a study on life expectations of parents. It is my favourite study on the mixup of correlation and causality to this day. Here is the basic outline:

  • Mothers and fathers generally live longer than the childless
  • Adopting a child adds three years to the lifespan, adopting two or three kids adds five years
  • People with 5 or more children die sooner then people without children

There was a lot of discussion about this study. It seemed counterintuitive. How could something as stressful as raising a child enhance your lifespan?

Later it was discovered, that the causality was the other way around. People that had the prerequisites for a longer life, were more likely to have children. And if you want to adopt a child, there are some requirements. There are things, that increase the likelihood of a long life. Many of those things increase the likelihood of having children as well. And you have to have many of them to be allowed to adopt a child, let alone three. 

When we analyse our business, we are often prone to tap into the same pitfall. We mix up correlation and causality.

The result are spurious correlations, like these, but less ovious.

How to define correlation and causality?

Correlation means, that two things happen at the same time. Causality means, that one thing happens because of the other.

Let’s take a common mixup:
Our revenue is going up means we are doing the right thing. This is a correlation. Not a causality. You could for example grow, but less than the market grows. This would mean, you are doing the wrong things in business or only some things right.

Something many people do, when they think about this, is find a way from a correlation to a causality. Something like: “But if we grow faster then the market, this means, we are doing the right things, right?”

Thats not how this works. You can’t discover causalities top down, only bottom up.

Think of correlation as two strings. You can knot or braid them together, but they are still two strings. Causality is more like a chain. Each element links to the next one. And sometimes two links are connected via a third one. Logical structures are not always linear.

How to ensure correlation?

So if you want to know, whether you are doing the right things, you have to go really low. Analyze the individual chain element. And then see, how far the chain can hold.

Make sure, that your steps are small enough. “We do customer research, therefore our customers like our product” is too big.

It’s rather:

  • We asked our customers, what happened right before they bought our product, so we have data on this.
  • 80% of the interviewed customers told a similar story, so we have a hypothesis of the pattern.
  • We designed an experiment on the hypothesis. If the experiment proves to be successful, we have reason to believe, that the pattern is correct.

We can then summarize these steps in: We did customer research. As a causality we found one trigger for a purchasing decision.

You can do business in two ways. Gut feeling mixed with luck or rational decision making. There is great data on how effective gut feeling is, but it doesn’t scale beyond yourself. If you want to grow, you have to outgrow your gut and luck.

Do you want to detect causality in customer behaviour?

Have a look at our next Mastering Jobs to be Done Online Workshop

Stay ahead on future posts and subscribe to our newsletter

Posts you may like as well

pirate digging up a treasure

Customer research for treasure hunting pirates

A tale of a young pirate learning from an old one.
man looking at survey results

Are surveys useful for product development

And if not, why?
extended competitive landscape

How to use the extended competitive landscape?

How to get your customer research skills from zero to one and beyond.