What’s so Special About Waterfall Charts?
The Rise of Waterfall Charts
If you have worked on a consulting firm (or close to anyone who has), you know what a waterfall-chart is, but if you haven’t, in a nutshell, a waterfall-chart is a representation of the composition of change over time.
It may sound complicated, but waterfall-charts are just a variation of bar charts, the most common type of chart that we all use and adore. These charts were popularized by the consulting firm McKinsey & Company. Their use became even more frequent after the book “The McKinsey Way” was published, where some of the advantages of using the chart are explained.
Using the chart can often be somewhat frustrating as setting the chart up requires some additional steps, and the plug-and-play versions of the chart in most data processing tools are not very good.
Waterfalls are now widespread in many industries including tech — and while I was skeptical at first — this chart has grown on me. I’ll do my best to convince you to use it a little more too.
Composition and Change — The principles Behind.
Usually, charts and graphs respond to a communication problem; “how to illustrate… distribution, change, composition, position, etc.”. There is usually a clear choice for most of these but the intersections can be trickier.
For composition, where the goal is to show the parts and proportion of an element, a stacked-bar chart does the job quite well most of the time. Pie charts are often a suboptimal solution but in some very rare cases, they are okay.
Similarly, for change, we try to show the degree of change of a given element over time, you don’t need to go beyond a line chart to nail it.
Waterfall charts solve for these two principles; this chart shows the composition of change over time. Ultimately it answers the inevitable business question:
“Why did X change Y between A and B?”
Cognitive Elements + Real Estate
In addition to answering the question “Why did X change Y between A and B?”, the waterfall charts are a fantastic example of good uses of cognitive principles embedded in data visualization. Here are just some:
- Intuitive reading flow: at least in the western world, we read from left to right, waterfall charts follow a natural progression from left to right.
- It provides an anchor: the first visual element of the chart is our starting point. It’s a mean of comparison that we will use to give magnitude to the following elements.
- Allows for visual conventions: it’s common to see waterfall charts with different colors according to the additive or subtractive nature of the components, but can also use principles of proximity to group subcomponents. These are also called “preattentive features.”
- Provides tons of real estate: this chart has a lot of negative space, and while it gives the chart an airily and decluttered feel, that space is also precious real estate to add annotations.
- Delivers information gradually: its horizontal arrangement and gradual delivery of information help your reader digest the content.
- Tells Micro Stories: a good story takes us on a journey from A→B, and this is precisely the heart of this chart.
Case Study
Getting to the decomposition of a change can be a little tricky, especially with non-linear multiplicative relationships, but I’ll assume we already have the data for the purpose of this post.
I’m a huge advocate for the praxis, so let’s assume that we work for a really big online streaming company, we will call it… Webflix.
So during March, we had 10 thousand active users to our service, quite average to be honest. During April, that number grew 40%, an unprecedented growth rate in a single month.
Here is where we ask ourselves;
“Why did X change Y between A and B?”
Or filling the blanks in…
“Why did subscribers grow 40% between March and April?”
Luckily for us, the monthly active subscribers have additive elements, and we know that for each month the Active Subscribers are defined as follows:
Active Subscribers = Previous Month Subscribers — Lost Subscribers + New subscribers.
At the same time, new subscribers can be split into paid and organic acquisitions, so the numbers of active subscribers for April look a little bit like this:
Granted, a summary table is a practical yet less appealing way to show the data. We can deliver the same information with a better visual.
Remember how we talked about the real estate in the graph? This time we used this free space to provide our reader with comments that might help them understand the numbers a little better. The annotations behind each module will gradually tell the story of how the subscribers increased by that much.
The red segments usually represent loss and the blue segments gains in order to leverage preattentive features that make the data even easier to understand.
How to Make a Waterfall Chart
The most common data manipulation tools usually have an out-of-box solution to make waterfall charts. However, I find them quite inflexible and hard to set-up correctly.
The secret is to get creative with stacked-bar charts (not 100% stacked) and use “invisible” segments to raise the next stacked element to the right level:
It requires a little bit of math, but nothing out of the ordinary. I like to reshape my data like this with a row for each segment color so it is more readable. Remember that the purpose of that table is just to help you plot the visualization and not an adequate way to deliver the data!
After plotting the data I just used Google Slides to add some guidelines, annotations, and titles since the editing there is more user-friendly.
You can apply the same principle in Excel, Tableau, Gsheets, DataStudio, or even Seaborn and Matplotlib.
Here’s a link to a sample in Gsheets
I really want to keep these series short and accessible so I’ll stop it here, but I hope you learned one thing or two.
