ThinkingDisfluency – the key to insight

Disfluency – the key to insight

Do you ever get the feeling that you’re surrounded by well presented data these days but actually not making any better decisions, or taking any better actions, because of it? Well you wouldn’t be alone. It turns out that often the easier data is to read – well designed graphs and tables – the less you’ll end up doing with it. Instead, it turns out that the secret to gaining insights from data is to actually play with it – get dirty with it in all sorts of seemingly inefficient ways – for you to actually understand it enough to take any real decisions from it.

Have a listen to this great 538 podcast “What’s the point?” – which is a podcast all about data. In this episode they interview Charles Duhigg, the author of Smarter, Faster, Better. You can listen on the link below (it should work straight on your browser or mobile phone if you simply click it – and don’t be put off by the 30 second promo at the start for another program):

It turns out that people only understand data sufficiently if they’ve had to work at it. In Duhigg’s terms, people understand data better if you introduce “disfluency” – if you actually make it harder to interact with.

Duhigg gives a great example – kind of relevant this week as the NAPLAN results start coming out – of teachers in the Cincinnati elementary school system. For ten years these teachers had been given all sorts of data on their students in well presented graphs and tables. This data told them a lot about individual student needs as well as overall trends. Great data to use if you want to lift results. But the results didn’t lift. Nothing shifted – despite all the data.

Then the state launched the “Elementary Initiative”. Teachers were required to create a “data room” in their school. They had to write out by hand on index cards the student information, and put it up on the walls (I hope they locked the door to this room!). In short, they had to rework the data: in a seemingly completely inefficient way. To the outsider this was a complete waste of time. The data was already presented to them in pre-prepared reports. They were adding no new information. And yet – the results for the students started to lift. Why? Because forcing the teachers to have to interact with the data – introducing disfluency – meant that they remembered it – meant that they had to make sense of it – meant that they were thinking about the student whose marks they were writing down. And all that interaction and thought then affected the way they taught and what they did in the classroom.

Apparently you get similar results with those scales that send your weight to your phone and graph it for you. All that beautifully presented data turns out to have no effect on losing weight. What’s better? Introduce disfluency. Every Sunday, graph by hand that beautiful data on your phone – create a less beautiful graph yourself. And what happens? As you do this you start to think. “What did I do Monday that meant my weight goes up on Wednesday? What can I learn from this?” And you get the feedback in a visceral way – because you’re playing with the data – and it starts to give you knowledge and insight.

It makes me wonder about all those board reports I see. Beautifully presented data that the board members simply need to read – all completely fluent. What insights is the board missing because they don’t have to work at the data themselves?

I’d love to hear from you as to your experiences of fluency and disfluency in data. Is Duhigg on to something?

As always, if you’ve enjoyed this post, sign up to my weekly email by dropping me an email at, and I’ll add you to the list.