Data viz Colors: The Good the Bad (and the Ugly)
Recently, I watched a great data visualization debate between two data viz masters — Andy Kirk and Andy Cotgreave, hosted by Tableau. They debated on interesting and relevant data visualization topics (if you haven’t watched I strongly recommend). There was one topic there that specifically urged me to write this article.
The specific debate topic was: Blue Orange color palette: Love or Hate?
The purpose of this round (6th out of 6 rounds) was to argue for and against Blue-Orange color pallets as indicators of good and bad KPIs as an alternative to Green\Red Color pallets. Personally, I like Blue-Orange but this isn’t the point (the votes for and against were equally split).
The point is that there was a strong agreement that Green-Red is not recommended, mainly because of the fact that color-blind people can’t distinguish between them.
As a matter of personal taste, I think traffic-light pallets also don’t look good and add unnecessary color noise to the viz.
The sad truth I have come across, is that the vast majority of business managers I worked with, and as a result their analysts and designers, tend to favor the use of Green-Red Pallet to indicate good and bad.
I have also heard of professional data viz courses and UI\UX designers that recommend the use of Green-Red.
To make things even worse, I have seen vast usage of Traffic-light (Red-Yellow-Green) pallets to indicate good bad and what’s in between (as if the in-between is so important that it deserves a flashy color of its own).
When I try to convince them to consider using other color pallets, I get answers like “This is what management is used to and wants”. I say that our job as data visualization providers is not to give managers what they want. We should give them what they need and what is considered as a best practice.
On the left-hand side of the below image there is a typical diverging Traffic-light color example to indicate sales profit ratio (%). On the right-hand side there is an illustration of how Green-Red color-blind (Deuteranopia) people see it:
One may argue that color blindness is only 5–10% of the population. The percentage of people with significant difficulties in physical functioning is about the same (National Health Interview Survey, 2018), but making accessible parking lots and public places for them is a consensus. So why is accessible data visualization isn’t a consensus?
What are the alternatives?
Here is an example of Blue-Orange color pallet (Orange — low profitability, Blue — high profitability):
Don’t like Blue-Orange? that’s fine. Here is another pallet to use, which is color blind friendly (Red- high % of discount, Gray- Low % of discount):
This approach highlights only the bad KPIs with the assumption that the bad KPIs are those which will probably require an action and thus need to stand out.
Another approach is to use size and positions to indicate Good and Bad KPIs, for example Up and Down arrows. The issue here is that Up isn’t always good, for example when it indicates year over year cost.
Here is a summary of the the common alternatives to represent accessible Good and Bad through colors along with my personal favorites which uses Red Blue\Turquoise- one less saturated\pastel and the other more saturated:
This is a call out to all managers, analysts and designers: Do Good for your data consumers: Stop using traditional Traffic-light and\or Green-Red Pallets (they are Bad or at least not a recommended choice, so experts say). Consider using common accessible Blue-Orange or Gray-Red pallets or use my “got to” pallets shown above instead. Beyond the fact that it is color-blind friendly, I truly think it also looks better and gives a quieter and cleaner experience, as opposed to the (Ugly- This is really a matter of my personal taste) noisy traffic-light.
Here is the link to the debate: