The KPI Status Dashboard is a static dashboard I developed for an organization I used to work for with the purpose of visualizing the general trend of individual group’s (in this case, department) performance relative to others, using the widely-used Key Performance Indicators (KPI) traffic light system of green, amber, and red.
Line Charts and Bar Charts
The dashboard was designed using small multiple bar charts. We will get to the term small multiple in the next section. The purpose of using a bar chart is to emphasize the size or volume of the dimensions being measured. Due to this, it is best served when there is a need to compare dimensions among each other. Take a look at these two examples below.
It is very convenient to our eyes to note that in the chart Breakdown of Decisions, the number of people who said Yes is slightly higher than the ones who said No. Likewise, in the Sales Performance by Region chart, we can easily observe that the South region excels much higher than the rest. The nominal dimensions – the Decisions Yes, No, and Undecided; and Regions North, Central, South, and International – can be changed in any order while still preserving the integrity and the pattern of the underlying data. It is only the ordering that has changed. The charts below tell the same data in three distinct views.
The first view is a predetermined view, as in the ordering is already determined in such a way, either by the management or other pre-existing conditions. The second and third view sort the data by descending and ascending factors. Note that the data remains the same, only our perception of the ordering differentiates each of them.
If we were to present nominal data using line charts, it would make little sense. Take a look at the Breakdown of Decisions and Sales Performance by Region charts (the same ones from above) represented using line charts and placed side by side with the bar charts version.
Even though we may still observe from the line charts that certain nominal data are comparatively higher in volume that others in the dataset, our eyes are subconsciously drawn to the moving trends of the line charts. For example, when we look at the Central region in the Sales Performance by Region chart, we cannot help but feel as though it is performing better as we observe the ascending line towards the South region data. Even when we look at the chart as a whole, we get the feeling that this could or should be a sales performance chart relative to time instead of regions.
In a small number of nominal data such as these two, even though the wrong chart is used, it is still easy to compare volumes using line charts. You may notice this as you were comparing the bar charts and line charts from the same data above. However, this gets comparatively more difficult as we deal with higher number of nominal data in the same chart. Observe these same two charts below represented using both bar chart and line chart.
It is easy on the eyes to compare volumes in the bar chart. We can see Region 9 being the best while Region 4, Region 7, and Region 10 are the worst performing regions. We can somehow still observe these same data from the line chart, but it is much more difficult because our eyes are more drawn towards the ups and downs trend of the line chart. In fact, knowing this is a sales performance graph would encourage our minds to subconsciously think that this is a sales performance over time chart instead of region.
That is why, when it comes to time-series data, the line chart is the best chart to use if we are to emphasize the moving trends (ups and downs) of the measured data. It allows our eyes to easily associate the rising and falling trend of the line as indicators of whether the data is performing well or not over a time period.
Unlike the line chart, the bar chart is slightly more versatile, in that it can also be used for a time-series data. I only recommend this if the need to emphasize the data volume is higher than the trend. In other cases, the line chart works best for time-series data. Observe these two charts below of a time-series data represented using both line chart and bar chart.
The line chart focuses your eyes on the trend of the data. It leads you to time-based line of reasoning, such as “What happened in October that causes sales to dropped so much?“. On the contrary, with the bar chart, your eyes would focus more on the size (volume) of the data, causing your mind to ask follow-up questions along the line of “Why are we doing so well in March and April, and again in August and September?“.
The ‘line chart or bar chart’ choice is one of the basic ways in which we can steer the audience’s thought to what we, as chart makers, want – do we want the audience to focus on the trend or the volume? It utilizes the psychology of perception, which is one of the modules that I offer in my training programmes.
The portfolio dashboard presented a series of time-based data using bar charts. As you may guess, the reason why this is so was because I wanted the audience to focus on the volume of the KPI data. That is the message of the dashboard when it was made – recall my post on communicating messages in charts here. Before the KPI data is presented holistically in this dashboard, every department’s performance was reviewed in its own chart, without context with other departments. As such, Department 3 in the dashboard had always been perceived as problematic due to huge numbers of KPIs in the amber and red compared to others. Given the human nature of focusing more on negative events, this is hardly surprising.
Therefore, when all the data was presented as a whole as in the dashboard using bar charts (to emphasize volume) instead of line charts, we can observe that Department 3 was being measured on much more KPI (volume) than other departments, with Department 1 and Deparment 6 being the least measured in terms of the number of KPI. Of course, this was a fact that was widely known even before this dashboard was made. However, by putting all the data in one view, any emotional biases could be put aside and the pattern in the data could be observed holistically. As you can see, Department 1 was the worst performer as of this dashboard’s time period, with an overall Red KPI indicator.
In short, remember this guideline:
- Bar Chart emphasizes volume and size. Use it for time-series data and nominal data.
- Line Chart emphasizes trend (ups and downs). Use it strictly for time-series data.
Small Multiple Charts
Small multiple charts are a great method when we need to compare multiple dimensions in a data set. Instead of using only one chart to represent the entire data set, it is broken down into small parts by dimension, all of them using similar chart elements, which is repeated for each dimension: hence the name small multiples. In the portfolio dashboard, there are two dimensions that utilize small multiples – department and KPI status indicators. The department dimension has six items – the six departments – and the status indicator has four – the traffic light status and one for the status Not Reported. In total, there are 24 (6 x 4) bar charts, all displayed in one screen as a dashboard.
Observe the dashboard containing stacked bar charts below, which is one of the many ways that displays the same data without the benefits of small multiples. We want to retain the emphasis on volume therefore bar chart will remain the display medium.
Observe that comparing the Green KPI between each department is easy, since their value starts at the bottom of the y-axis, the value zero. Try comparing the other KPI indicators between departments. It would be significantly harder since their values do not start at the same level. We can try to fix this by displaying the dashboard using clustered bar chart, as below.
Now, we can better compare individual bars with each other due to all of them starting at the same level. However, the data now looks cluttered. We can hardly compare the Green KPI between departments without the other KPIs interfering in our view. Small multiples would eliminate all this disadvantages.
Although we may call the two example dashboards above as small multiples themselves, each chart however contains four pieces of information, namely the four KPI indicators. Small multiples work great when there is only one piece of information in each chart, as you see from the original dashboard. If you need to use more than one information in a small multiple, ensure that the data is not cluttered. In the original dashboard, the Green KPI small multiples contain two information – number of green KPIs and target – that together, do not look cluttered.
Small multiple charts use similar chart elements that are duplicated throughout all of the charts. In the portfolio dashboard, you will observe several recurring elements. The x-axis on all the charts represent each month of the year, from January till December. The y-axis represents a scale from the value 0 to 34, which is duplicated in all the charts. The y-axis scale is not actually visible in the chart and was deliberately omitted in order to save space. However, all the charts were plotted on the same y-axis scale (0 to 34).
Therein lies the small multiple’s strength. It allows us to efficiently compare the charts with each other without the clutter of seeing all the data in one chart. Since the similar elements (the x-axis and y-axis) are the same for all the charts, we need to remember and understand this fact only once in our working memory, and the rest of our focus can be efficiently utilized to compare the individual charts with each other.
Say for example, one of the similar elements were not duplicated, the y-axis for instance. Look at the chart below to see the results of several modifications of the y-axis. Click on each image for larger view.
Note that in each of the charts, I decided not to display the y-axis scale, just like in the original dashboard. The data labels within each chart should provide the context of the y-axis scale.
In the first chart, chart , you would note that fair comparison cannot be made between each chart, since each of them uses its own y-axis scale. The dashboard made it seem as though Department 3‘s Green KPI almost the same size as Department 2‘s Not Reported KPI, even though the former uses a y-axis scale from 0 to 35, and the latter 0 to 10. You will find this is a common problem in both charts  and  as well.
In chart , each department’s KPI share the same y-axis, granting fair comparison to be made within the charts under each department. However, fair comparison between departments would be unfair considering they use different y-axis scale. In this regard, it is better for each department’s KPI status to be displayed in its own page separated from the others, such as in the example below for Department 1. Five more displays can be made independently for the other departments.
In chart , all the horizontal charts share the same y-axis scale. As such, we can fairly compare the Green KPIs of Department 1 and Department 2, but comparison between KPIs would not be fair since, again, they use different y-axis scale. In this case, just like in chart , each KPI indicator can be displayed in its own page, like in the chart below for the Green KPI. The same can be done for the other KPI indicators.
In this section, we have learned two things. One, small multiples are a great way to holistically compare charts without the clutter of seeing all data in just one chart. Two, if small multiple charts are used, all the charts must align together. In the dashboard, this was achieved by having all of them share the same axis scales and the same medium (in this case, bar chart).
Design of the KPI Dashboard
Now that we have explored the charting methods of the dashboard, I would like to conclude by explaining some of the design principles of the dashboard.
The first of this is the use of dashed bar to indicate forecasts, predictions, and projections (call it what you will). This is a highly effective method to draw the audience’s attention that dashed bars indicate something that is not solid. In comic, for instance, dashed balloons are used to indicate whispers, a non-solid form of communication, if you get my meaning. The same principle is used in charts since our brain is so used to associating dashed data with non-solid items. Look at some of the charts below for examples of this technique in effect. The first two charts show forecasts where the months demarcate each other perfectly, while the last two caters for charts with partial data in a month.
Other ways to display forecasts include using different colours and data labels, such as in the charts below.
Usage of lighter hues of a primary colour – in this case, blue – would also serves the same effect to our brains to associate the lighter hue as being non-solid. In my opinion, both techniques work well and it is up to individual chart maker to make the decision. Only remember this: be consistent in your choice throughout your charts. Do not use both techniques in the same display.
The second design principle is the data label for dates – in the dashboard, this is located at the top section, with each month labelled using its initials, J (for January) to D (for December). In a dashboard, it is important to abbreviate and be succinct since space is so limited. Abbreviations that are used must also be instantly recognizable by the audience, otherwise their thought patterns may be disrupted. In this case, using the months’ initials is acceptable because it is commonly used in our daily lives.
The third design principle is the highlighting of the current month through the use of a light colour. In the dashboard, this is presented in the month of August using a light grey colour, with a legend in the bottom left indicating this colour representing the current month. This works well with the forecast technique since any bars following the current month utilized dashed bars. As such, our minds would quickly make that association when studying the dashboard.
The fourth design principle is the use of colours green, yellow, red, and dark grey to represent the KPI indicators. Since the KPI indicators are themselves named after colours (except Not Reported), choosing appropriate colours for each of them was an easy matter. In other dashboards and charts, we need to be careful when choosing the colours green, yellow, and red. Our brains are so used to associating green with good, yellow with caution, and red with danger. This association comes from traffic lights in the real world.
The fifth design principle is the use of black bars within the Green KPI to indicate target. Usage of such bars facilitates the mind to associate them with targets. How is this so? Sports. In sports, it is common to see targets – goal posts, for example – presented as between two points. The link between these two points are perceived as a line that needs to be crossed in order to win. In the dashboard, only the Green KPI needs the target to be displayed. Since this is a KPI performance dashboard, each department should strive to have its total KPI in the green at all times.
The last principle is the summary of each department’s performance. In the dashboard, this was presented using the calculation in the bottom right corner of the dashboard and each department’s summary performance is displayed with a round button right under the department’s name. How the summary is calculated bears little relevance to data visualization. However, displaying the performance summary does. In dashboards, always have summary information. If your audience have little time to study the whole dashboard, he or she at least can glance at the top of the dashboard to view the summary information.
The KPI dashboard presented in the portfolio did not take account of the different priorities of the individual KPIs. As some of you would know, some KPIs are more strategic to the organization than others. The same dashboard techniques presented here can be cascaded to other areas, such as strategic KPI and operational KPI covering divisions and business units, even right down to individuals. As always, no matter how high or how deep you represent the dashboard(s), remember to be consistent in your display of data.