Previously I’ve advocated the use of probability distribution curves to describe uncertainty more clearly. It helps us get away from the binary success/fail of traditional risk listing, and instead allows a much wider understanding of the likelihood of different gradations of outcome. For example, not just whether we’ll make our predicted profit, but different degrees to which we might be over or under our target figure.
But this week I read a great piece by Ruth Fisher about using game theory to make effective decisions in uncertain situations. And along the way she used an alternative visualisation: a bar chart.
Adapting an example from a previous article let’s suppose we’re considering the danger of a security breach. We might say…
- There’s an 80% chance we won’t have a security breach this year;
- There’s a 5% chance we will have a breach costing us up to £200,000;
- There’s a 10% chance we will have a breach costing us £200,000 to £1m;
- There’s a 5% chance we will have a breach costing us over £1m.
And suppose we also considered some kind of specific remediation—say, educating our workforce—which we estimate changes those figures to 85%, 4%, 8% and 3%. We might illustrate those two situations with two probability curves, but here’s how we might do it with bar charts:
This shows us the estimated effect of our education, squeezing the likelihood and impact of a breach. We can then discuss whether we are happy with this, or if we’d like to explore further actions.
Although one visualisation is not wholly better than the other, I like this as an alternative. Probability curves carry a bit more information (and therefore require a bit more information) but can be harder to interpret, and I’ve only ever used them in one-to-one conversations where I am sure of my audience and can discuss any concerns or confusion. They are also tricky to generate with good accuracy. By contrast these stacked bar charts are easier to interpret and easy to generate. I will be looking to use them at the next opportunity.