Very few people can remember when dashboards were not an indispensable part of business intelligence (BI). Since the Web started playing its current role in our lives in the late 1990s, dashboards became one of the focuses of BI efforts. Surely, after so many years of evolution, the science of dashboards must be perfected and common pitfalls documented and avoided, right? Wrong.
As almost anything within business intelligence domain, dashboards are still more art than science, mastered by the talented and skillful. Despite the fact that everyone wants them and the proliferation of tools and techniques, perfect dashboards stay elusive and still require as much effort to create as ever. It is important to understand why and be prepared, for both producers and consumers of dashboards. There are some factors in play as old as a concept of dashboard itself, and there are some new ones that started emerging recently.
The presentation danger
Dashboards are glamorous. The old adage that “a picture is worth a thousand words” is demonstrated through dashboards in all its glory. A well presented dashboard allows fast processing parts of our brains to understand situations and make decisions within seconds. The presentation of information and facilitation of decision making processes are the main purposes of dashboards, and as such, they have to be approached the most carefully.
One would hope that your BI efforts have been well defined and delivered, and the data behind your dashboard is reliable, clean and consistent. If it isn’t, all other dashboard concerns are irrelevant.
The most accurate and consistent data is just a bunch of numbers and there are as many ways to represent it as there are pieces of information you collect. A very common mistake is to take the available figures and concentrate on presenting them without first defining the reasons and goals for putting that data on a dashboard at all. A pie chart or a graph? Old fashioned gauge or a series of bullet charts? It is assumed that if a business already collects numbers that are supposedly relevant, reporting on them will benefit business. This approach does not necessarily lead to failure but it will not lead to great success either.
A second approach which sadly is still not the most popular is to start with the strategy, forgetting about the figures for a while. Questions must be asked about what a business is trying to achieve, what will create real value, and which numbers will help to reach that goal. The numbers themselves need to be analysed to understand if they are as accurate, relevant and meaningful as required by the strategy. Only then can candidate data be selected and its presentation defined. This approach can lead to some lengthy discussions which might at time be seen as waste of time and budget, but in a long run it will deliver better business intelligence and better business practices.
The TV effect
Anyone who ever tried to tell a story knows that there must be just the right amount of details for every genre (are you writing a novel or a newspaper column?), and if the story has too much of what is perceived as irrelevant, the meaning will be lost. Users are not supposed to spend hours analysing dashboard information, it has to be perceived quickly and effortlessly. Which necessarily leads to selection of most relevant information and often, oversimplification. I call it “TV effect” by analogy with a simplified reality TV presents to people, where real life is reduced to a set of labels and the viewers do not choose which labels to apply.
One outcome of oversimplification is reliance on top level picture and neglect of lower level details and underlying causes. If the indicator for sales in one region is green and red in another, we are immediately called to action by red light and soothed by green. But what if the first region is missing some large opportunities while still meeting its sales targets? One answer is to develop more indicators, thus increasing complexity. Another answer is to review the lower level details and periodically re-evaluate criteria for relevancy and accuracy.
Another outcome of oversimplification is a tendency for early conclusions and premature changes of course. It is unnerving to be faced with red indicators for any prolonged period of time. An almost flat or slightly falling real time graph of new product sales might lead to a decision of abandoning the product earlier than it would happen with a different reporting approach. True, this can be a blessing and allow a business to minimize its losses. But what if an original decision would yield positive results given more time? One answer can be more data, another – setting relevant goals and keeping them in mind when making decisions.
Business decision makers have to remember that dashboards are only a tool and any tool is only good if it serves its purpose.
To be continued.