Recently someone sent me the bar chart below, and asked what I thought of it (as if I needed an invitation to comment!).
My immediate reaction was to ask why the author had used different shades – or “hues,” as they are called in graphic design – of black on the bars.
The heights of the bars clearly show that revenue has increased each month, so it’s redundant (and distracting) to use color hues to display the same increases. Of course it’s also true that hues often highlight variations in volumes, rates, and other measurement, but here they aren’t needed for that purpose, either.
The change in color is redundant (am I repeating myself?). Simply display the bar-graph data like this:
This is perhaps a good time to ask, “when, how, and why do hues work best in data visualization?” As I suggested above, we might want to show changes in volumes or rates with them, as on this choropleth map from the Dartmouth Atlas:
2010 Part D Medicare Enrollment Cohort
Percent Filing at Least One Prescription For a Dementia Medication
By using different hues of yellow to brown (from the lightest shade for the lowest percentages, deepening to dark brown for the highest), we can illustrate that in 2010 the percentage of Medicare beneficiaries enrolled in Part D (the prescription drug program) and who had filled at least one prescription for a dementia medication, was much higher in Southern California than in Northern California (for example).
The use of hues on this type of map helps us quickly and easily see and compare low and high values, and even to better grasp the full “what and why” behind the display. This use of color hues makes complete sense. (To see and learn more about this type of data display, take a look at The Dartmouth Atlas of Health Care.)
Hues also often work well to dynamically direct viewers’ attention to a metric that signals an urgent situation requiring immediate attention. This is particularly useful and important on dashboards or in summary reports; I’ve discussed these frequently in previous newsletters and posts.
In the issue of 15 May 2015, for example, I used a dot indicator in dark red to draw attention to those measures that fall furthest below the national comparison, then incrementally lightened the hue to match the diminishing differences between actual performance and the standard.
In another example – you can see it in our HDV website portfolio by clicking here – I used arrow-shaped indicators and a range of black tones to show changes (increases, decreases) in a hospital’s payor mix from one year to the next.
Note that the selection of a neutral color for icons (instead of more emotion-laden colors such as red or green) allows the viewer to quickly see changes in the data without conveying any judgment on the value of the change. This is especially important in a display presenting an element such as payor mix, where the same changes may be good in one situation and bad in another.
Further, avoiding colors such as red and green makes understanding the display easier for those with visual variations and inability to see certain shades accurately.
The example offered to me for comment (the first graphic, above) demonstrates a fundamental understanding of the use of color hues to show differences in volume. That’s a good thing.
But as with everything in data-viz (okay, and perhaps in life in general), true mastery resides in knowing precisely when and how to use (or not to use) a technique, so you can get your point across without distracting or losing your audience.