Last week, I stepped on the fancy scales at my gym for the first time in four months.

You know the ones. They don’t just tell you your weight. They tell you your muscle mass, bone density, basal metabolic rate… Basically, everything short of your star sign.

The results were horrifying. Everything had gone backwards. I’d even managed to lose 400 grams of muscle.

I’d been training five days a week and made no major changes to my diet. I was lifting heavier weights than ever before.

None of it made sense.

For two days, I was miserable - and mildly worried I might be sick.

Then I started asking questions.

How are these numbers actually calculated?

It turns out the scales measure only two things directly, and everything else was calculated using formulae. So if one input was off, it would flow through to everything else - hence, the correlation in the metrics.

Could something have affected those inputs?

As it turned out, yes. Every other time I’d used those scales, it was first thing when entering the gym. This time, it was at the end of an intense training session.

The data wasn’t wrong, exactly. But the conditions under which it was collected were different enough to call the outputs into question.

The next day, I climbed on the scales again. This time everything was as expected.

I’d spent two days being miserable about a measurement error.

Here’s the thing…

The same thing happens in organisations every day - with far higher stakes.

If, at face value, the data looks reliable - and the conclusions drawn come from a seemingly credible source (be it AI or a well-meaning analyst) - decisions can get made before anyone thinks to ask: do these numbers actually make sense?

The most important habit a data professional can develop is to slow down and question the numbers.

It could save you days of misery.

Talk again soon,

Dr Genevieve Hayes