Give It a Spin!
How one simple rotation can transform your data viz
Charts are wonderful — they generate information out of myriad little details, unlock insights out of a seemingly unrelated tonnage of data, and to a great extent help uncover the iceberg that’s hidden beneath the cresting water. In my view, data visualization is an interesting blend of art & science that generates engagement and drives further exploration of any given topic.
As a self-professed data visualization geek, charts have always fascinated me as a great tool for communicating insights. The ground level fundamentals are obvious by now — each chart is different, and they are most effective when used in the purpose for which they were intended. But on top of it, I have always found playing with a chart and tweaking an existing graphic could be so much fun. One of such things that I find extremely interesting to do? Spinning a chart. Here’s what I mean.
Spin it by 180 degrees
Below is a line chart that represents number of average footfalls in the local metro stations of one of the major metropolis in the world.
You’d be able to easily identify that there’s a significantly large rush on the weekdays during both AM & PM working hours-time. On weekends, the rush is typically more prominent during the middle of the day. Also, notice that the peak weekend rush is only, at most, half of the peak weekday rush.
But now I’ll rotate this same graph by 180 degrees, and with some adjustments on the X-axis, I get the version below:
As compared to chart 1, the peaks here are either during mid-day or during the night when trains are not operational (in other words, the inverse of the original).
Say I change the y-axis here to “Available Space in the Station” (denoted by Space in the chart). Don’t worry about the actual unit of space; the intent is only to understand the appropriate pattern.
The first graph helped us easily visualize when are the busiest times on the Metro. However, for an Operations worker, who’s looking to utilize station space (perhaps as a source of revenue generation) or primarily wants to understand empty capacity, this is a much more direct depiction compared to Chart 1. And this will be true for any attribute that has an inverse relationship with footfalls / busy-ness. For example, it could be measuring “availability of vacant seats in train” (assuming train frequency doesn’t drop significantly during the day). So, for a person traveling with large luggage or an elderly for whom comfort is more important than actual travel time, this graph gives them a far more direct answer as to when are easier times for them to travel.
Chart 1, though, is absolutely the right answer if a decision has to be made for things that are in positive relation with footfall, such as how many ticket counters should be active at what time, what quantity of security personnel need to be deployed & when to better manage oncoming crowds or so on.
This is but one simple example from a specific area, but more generally, for two KPIs or metrics that are opposing in nature, a rotation by 180 degrees on one will almost always help to communicate the inverse message of the other.
Spin it by 90 degrees
Let’s look at this graph below:
This is great if we are looking say at a direct metric like Profitability for different products (A1, A2…A6 in the above) for a given period of time (the exact numbers are again hypothetical and not important). Very easily, it becomes clear which product is riding up north vis-à-vis which product is sliding down south. A1 and A6 seem to be driving lots of profit, while A5 may be worth revisiting for this hypothetical business!
But once again, I just can’t help myself rotating this chart. This time, however, I am going to rotate it by only 90 degrees.
The same Chart 3 would now look like lefthand side image in the graphic above, to which I am going to do some adjustments so that it becomes a better presentation as depicted in the righthand size image above.
For Chart 4, rather than representing it as Profit %, consider a new use case — there’s an inventory management system in place, which has a forecast to predict what should be the inventory level for each product. To ensure that the forecast is as close as reality, typically, operations staff would want to monitor the difference between actual and predicted levels on an ongoing basis. The larger the difference, the larger the need to fine-tune the model. And this is such an important function since whether it is under-stocking (Actual Demand greater than Forecast) or over-stocking (Actual Demand less than Forecast), both situations are either significant revenue lost or cost to organization, if the deviation is above a tolerable threshold.
In this case, if Chart 3 (un-rotated version) was used — a positive Y-axis would have indicated over-stock situations while negative Y-axis would have indicated under stock situations. See below:
Typically in a chart like that, cognitively we are so attuned to perceive a negative y-value as bad and positive y-value as good, so there’s a good chance many users may get misled attributing under-stocking to be a problem area but not over-stocking, when in reality, large deviations are most likely equally bad, regardless of direction. Precisely these are such situations when a rotated chart as in Chart 4 will come in handy. Users looking at this chart are more likely to not get biased and would tend to understand the interpretation of the bars for themselves in order to draw an appropriate conclusion.
So while Chart 3 was good for metrics where order of directions matter (i.e. one of a positive/negative axis value is actually an exclusively a good-bad situation), Chart 4 is better for metrics where neither direction is disproportionately important, case in point being bi-directional deviation from the center.
Long story short: what message you’re trying to send with your chart matters. Whenever you are creating a graph from scratch, you need to think about what that chart is going to convey, which type of chart should be selected, what would be the axes, layout, labels, etc. And if you are already following these fundamental best practices, the graphs you create will start speaking clearly to your audience. And while you are doing this, there’s no harm in giving your chart a spin — who knows, something that was confounding your audience may suddenly nail it 90 degrees later! Or you get to see something that you didn’t even think of when you started!
Saurabh Singhal is a Data Visualization & Analytics specialist who takes pride in delivering business value by generating data based insights. He is passionate about data, number crunching & everything around that. In another universe, “Data & Numbers” would have been his most sophisticated & closest ally. It would be a character whose alien musings may get unfathomable at times but would always be engrossing & enriching!
