People ask me how three opinionated people can write a book like The Big Book of Dashboards together. Didn’t we disagree on things? How were we able to work out our differences? I can’t speak for Jeff Shaffer and Andy Cotgreave, but I’m very glad I found two fellow authors who would challenge every assertion I had, as it made for a much better book.

And why did it work? Because we had one overarching goal in common: clarity.

When I am asked about the process I think of a band breaking up because of “artistic differences”. That didn’t happen with the three of us because we weren’t trying to create art. For certain, we wanted dashboards that were beautiful, but more than anything else we wanted dashboards that allow the largest number of people to get the greatest degree of understanding with a minimum amount of effort.

Let me take you through a case study on how the Churn dashboard came to fruition and how following the approach we used could also help you to make better dashboards.

I had just finished presenting the third day of three days’ worth of intensive data visualisation training when an attendee showed me a data set similar to that in Figure 1 of subscribers gained and lost over time within different divisions.

I asked the attendee what she thought she needed to be able to show and she said it was important to know when and where things were really good (ie, many more people signing up than cancelling) and where and when things were really bad (ie, more people cancelling than signing up). She also stressed that management would insist on seeing the actual numbers and not just charts.

Here’s a famous quote attributed to car designer Alec Issigonis: “A camel is a horse designed by a committee.” Which is one way of saying that you will run into problems if you attempt to incorporate many people’s opinions into a single project.

This was not the case with the Churn dashboard as I received more input from more people over a longer period than any other dashboard in the book – and it resulted in a much better product than if I had just gone at it alone.

Let’s look at the evolution of the dashboard.

Take a look at Figure 2 which is one of my first attempts to show what was happening for Division A.

Starting with the left side of the top chart, we see the starting point for the month (0 for January), number of people subscribing (the grey bars going up) and number of people cancelling (the pink bars going down). It’s easy to see that I had more people subscribing than cancelling in January, and more people cancelling than subscribing in February.

The second chart shows the running sum over time.

Most of my attempts were fashioned around some type of GANTT/Waterfall chart, but one chart that showed promise for a small group of reviewers was a juxtaposed area chart, dubbed the “mountain chart” by one client who was kind enough to give me feedback (see Figure 3).

While some people “got” this, most had a problem with the negative numbers (the cancellations depicted as pink mountains) being displayed as a positive. The idea was to allow people to see in which months the negatives exceeded the positives and you can in fact see this easily (February, May and especially July). But most people were simply confused, even after receiving an explanation of how the chart worked.

In addition, superimposing a second chart (in this case the running total line) almost always invites confusion as people must figure out how the axes work (eg, do the numbers on the left axis apply to the area chart or to the line?)

I made some modifications to the GANTT chart and thought I had a winner, but Andy wasn’t buying it. It was then that I realised that I had lost my “fresh eyes” and what was clear to me was not clear to someone else, even someone as adept at deciphering charts as Andy. Andy explained that he was having trouble with the spacing between charts and the running totals. It was just too hard for him to parse.

I took the feedback to heart and came up with a solution that I think works well.

You may recall that one of the requirements is that people using the dashboard would need to see the numbers driving the chart. I suggested adding a text table (also known as a cross tab.)

When I showed this to Jeff there was a long pause, and then I recall him saying that he didn’t think this added much to the analysis. By this time I had worked with Jeff for well over a year and I knew that “I don’t think this adds much” was Jeff’s way of politely telling me that he hated that component of the dashboard.

I started to argue with him that there was a stated demand by the audience to show the actual numbers driving the charts when I realised that Jeff was, in fact, correct. Just showing the numbers didn’t add much and there was a better way to meet the requirement and provide additional insight: use a highlight table (also called a “heat map”).

I got a great deal from reviewing the dashboards other people submitted to the book and now wish I could go back in time and borrow some techniques from those dashboards and apply them to my own. Indeed, there isn’t one dashboard that I built for the book that I wouldn’t like to modify and that is certainly the case with the Churn dashboard.

Figure 4 – Churn dashboard with BANs (Big-Ass Numbers) – is the dashboard I would submit now. There are BANs along the top and these elements can do a lot to help people understand key components of a dashboard, they can be conversation starters (and finishers), provide context to adjacent charts, and serve as a universal colour legend.

If I could only make one recommendation on how to make better dashboards it would be to find people that can give you good, constructive feedback on whether what you’ve built is as clear as you think it is. Gird yourself for a lot of revisions and be prepared to add refinements, but it will be more than worth it.

Wexler, Shaffer and Cotgreave present a series of business scenarios and develop dashboards for each, so that people can take the most relevant one as the basis for their own scenario.

Options include:

• What-if analysis (wage increase ramifications)
• Ranking by now, comparing with then
• Multiple key performance metrics
• Year to date, year over year – at the same time
• Sentiment analysis – showing overall distribution
• Sentiment with net promoter score
• Showing rank and magnitude
• Showing churn or turnover
• Show actual versus potential utilisation
• Economy at a glance

Wexler uses two examples to demonstrate different churn/turnover dashboards: the growth of a subscriber service and airport passenger flow.

For the subscriber service, his dashboard monitors fluctuations in subscriptions over time; comparison of overall gains and losses by division; a way of identifying where losses outweigh gains; the details for each month, and the ability to spot best and worst months’ performance. The various merits of waterfall diagrams, spark bars, mountain charts and heat maps are compared to see which works best for the dataset (Figure 4 also shows examples of waterfall, spark bar and heat diagrams).

Raw tabulated data arranged by month also records gains, losses, net and running total (a waterfall diagram). Data is presented month to month in each ‘step’ of the waterfall, with colours to differentiate the gain and loss columns. Wexler also uses a sparkbar in the top left of his dashboard to make easy comparison of monthly gains and losses, while also highlighting the months with the greatest gain/loss.

In an airport flow diagram, Wexler uses a sparkbar arrangement without a waterflow chart as a better measure. He layers multiple bars on top of one another to show inbound and outbound flights against the number of planes on the ground, with this bar changing colour when a certain number of planes is exceeded. This is plotted over time on an X-axis to show change throughout the day.

• Data visualisation in presentations
  Business & Management Magazine, Issue 259, November 2017

The ICAEW Library & Information Service provides full text access to leading business, finance and management journals. Further reading on producing effective data visualizations is available through the articles below.