The 9-Box Grid
A powerful heuristic diagnostic tool.
Background
I first encountered the 9-box grid in my consulting work in HR Analytics. It’s an elegant device to classify the workforce along 2 orthogonal dimensions — Performance vs Potential. Each dimension, or axes, is divided into 3 levels — High, Medium, and Low. Based on an employee’s classification along these 2 dimensions, a set of recommended development actions would be applied. The diagram below summarises it.
There are those who love the simplicity of the 9-box grid, and there are those who hate it. Of the latter, there are many valid reasons to criticise this approach to categorise the workforce, from definition and measurement errors to action mis-alignment. For example, how should one define potential without incorporating performance, so that the 2 dimensions are independent? This article isn’t about the merits of the 9-box grid as a talent management tool, but rather, the use of the 9-box as an effective heuristic segmentation methodology.
And so I dedicate my 66th article to discussing the use of the 9-box grid approach to support diagnostic analysis.
(I write a weekly series of articles where I call out bad thinking and bad practices in data analytics / data science which you can find here.)
Diagnostic Analysis
The 9-box grid sits in the toolbox of Diagnostic Analytics. In my 42nd article (How To Perform Diagnostic Analytics), I wrote about what Diagnostic Analytics is and isn’t. It is an investigative method that starts with data sensemaking (the ability to apply a frame or perspective to interpret m eaningful information from the data) and from which a clear intent is formed to test certain hypotheses or operating assumptions. The 9-box grid does exactly that — it creates a framed perspective, it throws up both intuitive and (hopefully) counter-intuitive patterns, leading us to form hypotheses.
At the heart of it, the 9-box grid is a distribution-cum-classification technique. Firstly, it is a distribution along 2 axes. Unlike a scatter plot, which is also a distribution along 2 axes but with the primary intent on seeking correlation, the 9-box grid’s intent is about the concentration mix. There is intuition as to what that concentration mix should look like, and that serves to confirm that things are as they should be. On the flip side, if the concentration mix is counter-intuitive, it might suggest that there are unperceived opportunities or threats.
The 9-box grid is also a classification technique. The purpose of the 2 dimensions / axes is to create a more sophisticated way to label the population. In the original use case of HR Talent Management, combining different levels of Performance and Potential gives us a way to label an employee as a Leader (high performance x high potential), a Critical Resource (high performance x low potential), or a Liability (low performance x low potential). It’s an intuitive and heuristic way to reduce data complexity.
Now, the nature of this classification technique is arguably an action-oriented one. Because of the nature of the 2 axes, you can navigate the 9 boxes incrementally like you would playing chess. This allows you to articulate how you would invest resources to achieve a different concentration mix in each of the 9 boxes in a more guided and strategic manner.
Cognitive Load
The 9-box grid also hits the sweet spot when it comes to cognitive / information processing load. In 1956, Harvard professor George Armitage Miller, a cognitive psychologist, published the paper “The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information” wherein he conjectured that the human brain had certain information processing limits (sometimes interpreted as working memory); the human brain could only process 7 +/- 2 pieces of information. The paper became one of the most highly cited papers in psychology, and the conjecture became popularly known as Miller’s Law. Miller’s Law has been used in the telecommunications industry (landline telephone numbers were limited to 7 digits excluding regional codes), in UX design (form filling) and in management (presentation bullet points). While there are detractors to this conjecture (based on different definitions of working memory), there is neuroscientific evidence that Miller isn’t too off the mark; that the human brain has a cognitive bandwidth limit.
A 4-box grid (i.e. 2x2 segmentation) does not place any material strain on human cognition, but it also doesn’t provide sufficient range for classification variety and choice of actions; it’s too broad and simplistic. A 4x4 segmentation gives us 16 boxes which leads to immediate information overload. So a 3x3 combination comes off as “just nice”. What about 2x2x2 segmentation (which is 8 boxes)? Introducing another dimension / axis leads makes it harder to interpret the combination and derive appropriate labels.
Construction Principles
So how should one go about constructing a 9-box grid? Firstly, the 2 axes of choice must have an interplay that gives intuitive meaning but are not derivative of each other. For example, Sales vs Profit, Success vs Risk, Vintage vs Performance.
Secondly, the axes variables should be ordinal in nature. This is so that we can derive 3 increasing levels such as Low, Moderate, High.
Lastly, to determine the 3 levels, we either apply domain knowledge as to what constitutes H/M/L, or we use the rule-of-thumb (heuristic) of 20/50/30 — top 20% for High, bottom 30% for Low, and the remaining 50% for Medium / Moderate.
Conclusion
I love heuristic approaches to problem-solving and data analysis. They reveal information signals in interesting ways. The 9-box grid capture simplicity, practicality and complexity as an all-in-one diagnostic analytics tool — simple because it’s easy to communicate with, complex because of the thoughtful choice of axes that can frame and re-frame a problem or phenomenon, and practical because it can used as a roadmap for resource planning and execution. So let’s start having some practice with it!
