The next tip in my dashboard makeover series (which began here) shows how information can be skewed when you alter the scale. One of the features introduced in Reporting Services 2008 was the ability to add scale breaks like this:

scale-break

The purpose of the scale break is to prevent a category with a small value, like France in the chart above, from virtually disappearing from view because other categories like Canada and United States have significantly higher values. The problem is that your brain doesn’t easily interpret those scale breaks. A glance at the chart might lead you to think that sales in France are roughly half of sales in the UK which in tern are roughly half of the sales in the US – which is not the case.To interpret the chart correctly, you have to carefully review the values on the scale and then mentally calculate the difference between two categories rather than rely on the size of the column. And that violates the rules of dashboard design. If the viewer must start doing math to understand the message, you’re not doing it right!

scale

When the scale break is removed, you can more easily discern the correct relative sizes between categories. You can see that Canada sales are under half of US sales and that the difference between France sales and US sales is significant. Sure, you can’t see the actual value here but remember that the purpose of the dashboard is to communicate information visually. You should understand what you need to know immediately, and in this case, the information we glean from the revised chart is where sales are high and where they are low.

As a side note, I also cleaned up the chart a bit by removing the gridlines and reducing the emphasis of the non-data points. That is, I used a softer color (gray) for the labels and the axis lines because they are merely supporting features and not the data points themselves. The data points here are represented by the columns and pop out from the chart. Your eye focuses on the column sizes and is not distracted by extra lines that contribute no meaning.

To communicate sales values, the new chart is better than the old chart. But from a dashboard perspective, are these values what we expect to see? Are these good values or bad values? For use in a dashboard, there is more improvement that should be made to enhance the usefulness of this information, but I’ll save that discussion for a future post.

P.S. You can introduce a similar problem with skewing the message when you start the scale for a column or bar chart at a value other than 0. The relative size of the columns and bars are not reflected accurately when you do this.

Continue on with Dashboard Design Tip #3: Provide Context.