Data visualization is great – isn’t it? With almost no learning curve, anyone can create a scatterplot, a pie chart, a heatmap – or any of a hundred different types of data visualization. The appeal is immediate. We are visual creatures – we like pictures, and particularly when they have lots of color and nice clean layouts. But here comes the rub.
Creating a visualization is around one per cent of the work. Interpreting it is another two per cent. The other ninety seven per cent is actually doing something about it.
Say a visualization shows that product A is not selling well in regions X, Y and Z. Modern data visualization tools will give you this information at the click of a mouse – maybe a couple of minutes work. Then you have to decide whether the data are significant. After all random variations in sales, or any other activity for that matter, happen all the time. Is the variation worth investigating further. Some statistics would probably help, but data visualization tools are essentially dumb – they are not going to help in this respect. And besides statistics is hard – so maybe we settle for a subjective judgement. Flip a coin – ignore the data, or decide it’s worth going further. The data visualization isn’t going to tell you why there is an anomaly either, but data mining might. Well, data mining is hard too, so maybe you decide to get the local sales reps in and tell them to try harder. If you had mined the data maybe it could have told you that people who live in dry parts of the country just don’t want to buy your de-humidifiers.
So there are two routes. Look at the pretty pictures and make an educated guess as to the story the dots and bars are telling – or do some statistical analysis to see if there is anything significant. If there is something significant, we can scratch our head and try to figure out what is happening, or we can mine some data and see if any patterns emerge.
Finally there is the thorny problem of turning insight into action. If you use a business rules management system, then it might not be all that difficult. If you don’t then it’s probably going to mean training, meetings, more meetings, workflow changes etc, etc.
The moral of this story is that data visualization is just the first step in the continuous improvement that all businesses should engage in. It isn’t the whole story – it’s about 1% of the whole story. So please – no more books on data visualization, no more endless twitter streams showing the populations of salmon in various rivers, and please, please, please – no more posturing as if a data visualization is a ticket to a Nobel Prize. Enough said.