The problem with dashboards

Dashboards are an increasingly popular way of presenting data to customers, but they’re not always the best tool to use. In this post I outline why, and what you can do if you’re tempted (or asked) to produce one.

First of all I would like to say that I respect the work and skills of those who make dashboards, including my colleagues and ex-colleagues. In particular, I don’t want this to be a criticism of their work, more of a plea that I think we could and should do better for customers.

Data literacy

One of the main problems I see with dashboards is an assumption that those who will use them have reasonably high numeracy skills, and are as adept as the analyst creating the dashboard at drilling down, understanding and interpreting the information found in them. Analysts and data scientists have specialist skills which often aren’t shared with the general population. According to National Numeracy, nearly half of the population have primary school level maths skills, and 4 in 5 adults have a low level of numeracy. This means that they may not be able to compare the costs of products and services for example.

Think your customers are different? Think again. I know statisticians who have had to calculate simple percentages for MPs and journalists because they haven’t known what number to divide by what other number. I’m not saying that we should be treating people as if they know nothing, just that we shouldn’t make assumptions about what people might or might not know.

Communicating with customers

Obviously the first place to start is speaking to your customers. What is their aim? What decision are they trying to make? Can you gauge their numeracy skills? Talking to them obviously steers the type of analysis that you need to do and how to present that.

I have seen examples of customers asking for two numbers and receiving a dashboard with a full analysis in return. I have also seen examples of customers needing to undertake significant filtering, altering of assumptions and analysis receiving a multi-tabbed dashboard. But this makes an assumption that the customer has an analytical mindset and understands the assumptions made in any modelling as much as the analyst themselves.

I’m not berating all dashboards because there are some good ones out there, I just think we should be listening to customers’ needs and tailoring the results of our analyses appropriately rather than assuming dashboards are the answer.

What is the solution?

Talk to your customers, understand what they will do with the data you give them and what decision they want to make. Either ensure you involve user experience (UX) researchers or others who are adept at asking customers the right questions to tease out their real needs, rather than what they think their needs are. Tell a story. As analysts or data scientists, our skills are in analysis of that data. Find a story within it and present that so that the customer can make that decision or take action. The best data visualisation helps customers ask questions and understand nuances.

Best practice dashboards

Finally, if you still decide that a dashboard is the best answer, the Government Statistical Service has produced ten tips for designing dashboards.