"… try to present what are often quite complex datasets in a visually interesting and unambiguous way, so that users can immediately grasp what the data is indicating."
University of East London
One of the key aspects of BI is to present data which is often complex in a simple format so that time-poor users can be given the information they require in the minimum of time and with a minimum of effort on their part.
Most of the eleven projects considered what would be the best visualisation for the data they were presenting and, unsurprisingly given the diversity of the projects, arrived at a variety of solutions. All projects that created visualisations found that there was a clear need to ‘test drive’ them with users (either via a pilot or focus group) to ensure that they gave the correct information in a format with which the users were comfortable.
The ‘dashboards’ and other visualisations that your project delivers are likely to be its most visible output and, rightly or wrongly, are likely to define the perceived success or otherwise of your project. Get it right and your dashboard will sell your project for you, but get it wrong and you may find it an uphill struggle to generate any enthusiasm or appetite for the entire initiative, regardless of the underlying quality and value of what has been achieved.
The following lessons learned by the Jisc-funded projects should provide some useful pointers when it comes to considering how best to represent the data your system contains.
1. Visualisations should be appropriate for the data they are representing
A key factor in the choice of visualisation is that it has to be appropriate for the data being presented; for example, the University of Glasgow chose to present the degree of commonality between the research interests of two academics who may superficially be in different disciplines by using Venn diagrams. To highlight the geographic distribution of a research topic they used maps.
Similarly the Universities of East London (UEL), Central Lancashire and Liverpool who were all developing management dashboards chose to use bar and line graphs, pie charts, dials and ‘test tubes’ to present various aspects of key performance indicators (KPIs) and other strategic monitoring and planning information.
Another key aspect of visualisations that some projects including UEL and Liverpool considered was presenting key information to different users in the most appropriate format. Consequently many of the high level graphical visualisations also supported the ability to further ‘drill-down’ in detail so that staff at different levels of the organisational hierarchy could access the level of detail appropriate to them.
2. Esoteric visualisations are only justified where they provide additional information that cannot be easily represented in more conventional formats
The University of Huddersfield experimented with more esoteric three dimensional visualisations to represent the university’s research performance before deciding in this case that more traditional graphs were more easily understood.
Vitally, the project team decided that the 3D visualisation did not impart sufficient additional information when compared to 2D equivalents and therefore could not be justified; particularly as the users felt that decoding the 3D information was considerably more challenging than understanding traditional 2D images.
3. Using well understood visual conventions will aid understanding of complex data
Most of the visualisations that were comparing a data item to a key value used the standard red, amber, green ‘traffic light system’ to indicate ‘danger’, ‘alert’ and ‘all is well’. While this may seem simplistic it is a universally understood convention and allows very different sets of data to be easily interpreted.
In the cases of the Open University and the University of Bedfordshire the indicators were placed against individuals who were suspected of being in danger of disengaging from their programmes, while Liverpool used colour zoned dials to show such values as research income and applicant to enrolment conversion rates.
4. Road testing visualisations will help ensure that users understand the data correctly
Projects found that as well as incorporating industry standards for visualisations the target user audiences made valuable contributions. UEL, UCLan and Glasgow all found that by demonstrating their visualisations to the target audience at a reasonably early stage they received useful feedback that improved the eventual deliverable and in some cases stimulated demand into areas that had not been previously considered.
UCLan also discovered that within their target users there were distinct variations in preferences with some users preferring highly visual representations while others were happier with more ‘numerical’ expressions of the data.
5. There are a variety of ways of allowing users to customise the information they are seeing
Many of the projects considered the best mechanisms for allowing users to change the parameters of what they were seeing. The University of Bedfordshire even went so far as to agree a default set of parameters that represented a universal truth before allowing users to enter alternative values in order to explore their data.
Others were more relaxed about allowing users to enter their own parameters though inevitably this required a level of knowledge on the user’s part in order to return meaningful information. UEL implemented sliders in order to allow users to easily adjust their datasets without the need to know timeframes or coding-frames.