Scale matters

Charts have become an invaluable way to communicate information and articulate data. Charts help us not just convey data, but to also “tell a story”. With Charts, we can highlight certain points and help the people who need to act on the information we provide quickly understand it. Many considerations go into selecting appropriate charting types for your information. But in almost any conceivable chart type you plan to use, you will need to be thoughtful and plan how to address the scaling issue.

When I think about what can go wrong with chart scaling, I imagine a massively obese and happy guy standing smiling on a scale, next to a sad and stunned thin guy, who is also on a scale. Both of their scales only go up to 120lbs. You could argue this is great for the fat guy and bad for the skinny guy, but the fact is this is terrible for both of them, because they are both looking at a wrong reflection of their reality that will most certainly guide them to make the wrong choices.

One piece of advice you should consider up front and save yourself a lot of headaches down the road: if you plan to have any sort of charting capabilities in your BI application, report, dashboard, etc., make sure you understand how chart scales are handled. Most tools handle scaling automatically, and do not provide a lot of flexibility in terms of scale options. This would probably be ok most of the time, but I guarantee you will run into a data situation where these automatic settings would simply not be the right way to handle the picture. So, all you can do is try to be prepared.
Here are a few examples for some of the considerations you need to take into account when thinking about your charts scale.
Consider a relatively simple requirement: you are asked to create a report that will depict performance across regions for a national company. The company has presence in all US regions at this point, and is growing rapidly. It recently expanded into the Southwest. Performance is measured using two key measurements: revenues and margins.
You might start out with collecting the needed information and placing it in a table.

This is pretty good. It’s a clean simple table, it contains all the relevant numbers management would be looking for in a concise format. However, it does not tell a story. It’s difficult to analyze. It provides information, but it is not information management cannot ignore. You may decide at this point that a chart would do wonders to bring this information to life. 

This is the first example of how scale, when not handled, can turn into a major problem. This chart plots both the revenues and the margin across the same axis. Furthermore, the scale was hard coded, based on prior months results, to go from 50-150, so the Southwest region is simply not represented in the chart and the Southeast is cutoff. The picture depicted here is incorrect and misleading.

The next attempt is much improved. Instead of trying to plot revenues and margins on the same scale, you add another axis on the right, and you increase the scale limits to accommodate the data. The Southwest is not in the picture. Nice. However, something is still a miss. It is represented as a “blimp” compared to the other regions, and management looking at this depiction may not notice that the ratio of revenues to margins in this new region seems to be different than in other regions.

This final version plots revenues on the left vertical axis and margin percentages on the right vertical axis. In this picture, the newly added Southwestern region immediately draws attention. It dominates the margin picture. While revenues are still relatively low, something very interesting is happening down there, helping grow profits, and management will have a much easier time not being able to ignore this picture.
BusinessObjects Xcelsius charts have a good variety of scaling options, including primary and secondary axis, linear and logarithmic scales, and the ability to calculate min and max values for the Y (and Z) axis. This is a relatively understated feature of the product, but in fact, not many charting tools have this capability. It’s an invaluable tool and you should always keep an eye for the right choice of scale in your charts.

Controlling your charts scales will also allow you to come up with new charts types, overlay different types on top of each other to create interesting combinations. In a prior post I explored such concept of a Step Chart, made out of several synchronized scale charts. You will be able to handle a wide variety of data and filtering combinations that would otherwise result in misleading pictorial depictions.

This entry was posted in BI At Large, Xcelsius and tagged , , , . Bookmark the permalink.