There are 5 very subtle but deliberate features that make the graphs of your performance measures speak so clearly that it can feel like a very focusing slap in the face:
1. Use line charts as your default.
Line charts are the best visual method for displaying changes through time, and that’s exactly what performance measures – most of them anyway – are for. We want to see if performance is getting worse, getting better or not changing at all. Bar charts are okay if you don’t have much historic data, and less than a dozen values of your performance measure time series to display. Virtually all my performance measures, including my New Signups measure which tracks new subscribers to my email newsletter week by week, are displayed in line charts.
2. Include at least 20 points of historic data.
To really draw confident conclusions about trends or changes in performance over time, you need more data than you think. You need at least 20 values of your performance measure time series. There are statistical reasons for this, and it boils down to the fact that there is always some level of natural variation in our measures’ values which is always there even when nothing has changed. You need to focus on the patterns through time, not the difference between this month and last month. In my graph of New Signups, having the historic values clearly shows the sudden growth in the level of signups since August 2009.
3. Add a line to highlight the current average level of performance.
Because the difference between this month and last month virtually never means anything, because of natural variation, it’s much more meaningful to focus on the average level of performance. I’m not talking about the overall average of your 20 measure values. I’m talking about the average level that current performance exhibits. In my graph of New Signups, the average level of 115 per week only accounts for the values that belong to the new trend since August 2009. Before then, the average was more around 25 per week. The current level predicts the average number of new signups I’ll likely continue to get if nothing significantly changes.
4. Keep it clean and uncluttered – let the data dominate.
I certainly don’t want to stop you from having fun with your dashboards (I know many people feel numbers are boring and they need all the help they can get to be interesting). But I do want to warn you about when decoration distorts and distracts you from what your performance measures are really trying to say. What stands out most about my New Signups graph? The data does. I’ve used colour to highlight the line in the chart, and everything else is subdued. Incidentally, I use colour to flag at a glance which measure belongs to which of my business processes: Marketing (red), Sales (purple), Service Delivery (green), Product Development (blue).
5. Use the graph title to stay focused on the goal.
This one’s optional but I love it. You state your goal in the title of the graph, so it makes the performance measure clear but also the target you’re trying to achieve. My graph’s title is “New Signups are at least 250 per week” and that encourages me to always check if my average level of current performance is moving closer to that target – or not!
Please appreciate that this is not an exhaustive list of how to design fabulous graphs. But it is a fabulous start!
Why not create yourself a graph template that satisfies these 5 features, and test a few of your business performance measures using this new template? See if you get some clearer insights into what’s really going on with your business performance.
ABOUT THE AUTHOR
Stacey Barr is the Performance Measure Specialist, helping strategic planners, business analysts and performance measurement officers confidently facilitate their organisation to create and use meaningful performance measures with lots of buy-in. Sign up for Stacey’s free email tips at www.staceybarr.com/202tipsKPI.html and receive a complimentary copy of her renowned e-book “202 Tips for Performance Measurement”.
Great post! I fully agree. And for those who want to test their Graph Design IQ visit: http://perceptualedge.com/files/GraphDesignIQ.html
Stephen Few’s info is fabulous, Karel – I’m glad you thought to include his blog in your post. Thanks!