Dashboards are often the default for presenting organisational data, but they are only useful under specific conditions. They generally bring together data from different sources and present them in one view using graphs, tables, and infographics. As with all other data analyses, they can be done well, and they can also be done poorly. At worst they are data displays that are easy to ignore.

What good dashboards do

Starting first with what good looks like, these dashboards are designed to support decision making. For example, organisations will start with a series of key metrics that they want to track on how well the business is operating. These metrics are good measures of what the organisation wants to track and the underlying data are valid, reliable and updated in a timely way. The metrics are actionable, with changes in trends over time pointing directly to the need for changing business decisions. Bringing different but related metrics on one dashboard enhances visibility across a whole system, enabling decision making and changes in the right place at the right time.

An example of a good dashboard in public health is a communicable disease outbreak dashboard, with live updates. These will typically show the number of daily tests being conducted, cases detected, test positivity, number of people hospitalised, number of people who have died and can also include syndromic surveillance data from primary health care. Trends over time in these key metrics will also be presented. The combination of these metrics will provide information on how well the outbreak is being controlled, the morbidity and mortality from it and point to where action needs to be taken.  Data change at least on a daily basis, and are sourced from established data systems for pathology, communicable disease notifications, hospital admissions, primary health care and mortality data systems.  A combination of high numbers of presentations meeting the syndromic case definition alongside a fall in testing and case numbers suggest a need for further tests. Declining test positivity, maintenance of high testing levels and lower case numbers on the other hand suggests improving outbreak control.

It should be noted that even with well-designed dashboards, that good data literacy is necessary for interpretation. Users should understand something of the underlying data, the system they come from, the usefulness of the metric being presented and how data from different systems can be interpreted as a whole. These underlying principles are perhaps even more important with a dashboard, as data are presented without methodological information or context.

What bad dashboards do

There are also many examples of poorly designed dashboards. These tend to start from the position of presenting data that the organisation has, which is not the same as presenting data that relates to a problem the organisation wants to track. Metrics used are often stable over time and may be based on less robust data, or data from one system only. There is a lot of information but not much of it is meaningful or even actionable. These are more like static displays than dashboards. Static displays have the additional problem of being ignored over time because the user is accustomed to nothing changing, making it even less useful.

An example of a less useful dashboard, for example, is a dashboard on chronic disease surveillance. If we took a dashboard on diabetes, we might see measures for:

• Case numbers: prevalence data are reasonably stable over time
• GP visits or hospitalisations: these presentations are stable over time
• Even if there was some change, it is not clear what actions can be taken because of the change.

This kind of dashboard can be improved by including disease control outcomes such as average HbA1c levels, blood pressure control or primary health care screening for diabetes. Changes in these measures could inform new actions. Given these data move slowly, it is still the case they are probably not suited to a dashboard. These data might best be presented in a periodic report, with in-depth analysis of trends over time. Trends in sub-populations or geographic zones would be possible, allowing for a more nuanced interpretation of the progress of diabetes prevention and care in the population.

Conclusion

Dashboards are best placed for rapidly changing indicators that are meaningful and actionable. They bring together information from different data systems to provide whole of system insight, supporting timely decision making. Careful consideration of when a dashboard is the right approach ensures data are used to best effect, not just displayed.