“Our performance last month was 46.”

Oh, you don’t have to thank me, I was just doing my job. Not very well, I might add. 46? 46 what? Or 46 who’s?

Without context, 46 is just a number, just data. In context, perhaps that’s 46 out of 48 (not too shabby), or 46 out of 100 (shabby), or 46 out of 10,000 (fire that guy) (unless it’s a defect rate). Maybe it’s a ranking (#46 of the 50 States) or just a cardinal number (46 escape attempts), a velocity (spacecraft: 46 kilometers per second), a volume (46 tons of toxic wastes), or a change from the prior data point (478 to 524). Without context it could be any or all of the above.

Providing context is generally pretty straight forward, but an often overlooked aspect of communicating information. The initial obvious candidates for context are magnitude/direction (positive or negative), ratio or percentage, and simply labeling the units, for both the numerator and denominator if applicable. If the scale is not implicit, as with a percentage, then it needs to be made explicit (i.e. X out of 180). Perhaps the scale needs a bit more elaboration, expressed as a range, with high and low boundary conditions.

After you’ve got the basics settled, additional context can help you make value judgments about the data: is 46 good, bad, average, or impending doom? Perhaps the context now takes the form of a dial, slider, thermometer, or traffic light, with color-coded red/green/yellow thresholds, visual context to help make the value judgments explicit. But if we're going to apply thresholds, then just having the data isn’t good enough anymore, we need data about our data. Where do these thresholds come from? How do we know they reflect the right value judgment vis-à-vis our objectives? Typically there are two methods for validating your thresholds: external benchmarks, and statistical analysis of your internal data. With these approaches you can set meaningful thresholds and confidence intervals and know that your interpretation is accurate and the context is aligned with your goals. What’s the right trigger to signal a change from GREEN to YELLOW – anything under 100%? +/- 10%? Do you have an academic bias that might assign anything above a 70 as passing whether or not there is any support for that basis?

Context is what allows you to lie with statistics, or not, as the case may be. Do you simply let the application automatically set the range for the bar graphs or the line chart, or do you force the appropriate scale to appear each and every time? I can make my growth rate look spectacular on a bar chart as it doubles from Q1 to Q2 and then again for Q3, unless the axis scaling calls me out by showing how insignificant that 0.1 to 0.2 to 0.4% sequential growth really was when compared to my 12% target.

No need to stop there, though. Your data might only be one number, but you can keep piling on the context to increase its value. Do you manage your operation geographically? Then show the number(s) on a map. Do you want to manage each region in a consistent manner? Then make sure the same dozen KPI’s (key performance indicators), in the same order/layout, appear on the screen no matter which region you click on. Or is physical layout a more significant context for your operation, such as a hospital, in which case you don’t necessarily want the exact same consistent data categories for each department; you want the data categories to be context-driven, showing only the factors and KPI’s relevant for that department. The key metrics for the emergency room are likely to be somewhat different from those for the lab or for ObGyn or for oncology. Or do manage the operation as a process flow, where you want to investigate the bottleneck at step 4b that is holding things up on both the BLUE and ORANGE production lines?

Lastly, think of your enterprise as an ecosystem, where everything is related to or impacts everything else in some form or fashion. Context in this context means relationships and correlations and cause-and-effect and hierarchies and feedback loops. A “46” in this case might be a leading indicator, an amphibian or plant species particularly sensitive to changes in the environment that bodes ill for other components of the ecosystem in the future. Or that “46” is part of a collection where if any 3-out-of-10 of these leading indicators are RED simultaneously then the entire ecosystem may be at risk. (“Can analytics help do that”, I hear you ask? Absolutely).

There is of course no right answer to providing the right or the best context in any particular situation or organization. Not only is it different for each organization, and even within different functions of that organization, what’s best will evolve over time as better metrics substitute in for poorer measures and new visualization techniques are applied to the ecosystem. But what should be clear by now is that the initial bar for a high-value scorecard or dashboard has been set pretty high; and that by comparison, spreadsheets and standard management reports with their dearth of context are rather low-value approaches to providing business intelligence.