Data Science

How To Create Better Charts

Pinterest LinkedIn Tumblr

Visualization gives you answers to questions you didn’t know you had.

Ben Shneiderman, Distinguished University Professor Emeritus (UMIACS)

In times of big data, it’s often increasingly difficult to see the (data) wood for the trees, never mind leveraging data to obtain any relevant insight in order to make optimum decisions.

Data visualization is all about the representation of information or data in a chart, graph or other type of visual format. It makes it possible to see relationships between data through images, so patterns and trends can be identified. There are a number of techniques, plus some great tips, and we have chosen just our top six to kindle your data analyst spirit.

Data is the new oil? No, data is the new soil.”

David McCandless, British data-journalist

💡Tip #1: We Plough The Fields And Scatter…

We all had a dabble at school with the odd scatter plot, also known as scatter diagram, scatter graph or correlation chart. It’s a tool used to analyze relationships between two variables to determine how closely the two variables are in fact related. We plot one variable on the horizontal axis and the other on the vertical axis. With scatter plots, we can easily display two sets of data to see if there might be a correlation or a connection between them. But there is another use to give us more insight—the quadrant plot.

As the name implies, a quadrant plot is a scatter plot divided into four equal quadrants or sections. With the data in separate quadrants, data points which are similar and those which are not are easy to identify, based on specific characteristics.

To analyze for example your marketing performance, you’ll need data on the efficiency by customer acquisition costs (CAC) and on the quality by customer lifetime value (CLV) per campaign.  In this quadrant chart, CLV is shown on the horizontal line and CAC is found along the vertical line. The four quadrants are designated Bad (upper left), Mixed (upper right, lower left) and Good (lower right). Place each of your campaigns in the appropriate box based on where they rank in CAC and CLV. Of course, you are looking for an inverse relationship between your CAC and your CLV, with your CLV being the higher of the two numbers. The less it costs you to acquire a single customer, and the higher the total value this customer represents, the more profit you can make.

For Insiders 🤫

To move quadrant plots to the next level, start by using color as the third dimension instead of assigning color to the quadrants. Let’s take an example from MLB: The following scatter plots show the interplay between a pitcher’s performance on the mound (X-axis), his team’s performance at home plate (Y-axis), and how this interplay is translated into winning games (color).


💡Tip #2: The Small Multiples To Success

Rather than one big chart, why not use loads of small charts? Small multiples are the name given to a series of small charts in a table layout.

They fix over-plotting data and make comparison easier and work best when each chart is just a simple image. As an interior design store might want to see sales of chairs, against the rest of furniture and other design items by state, so you use 50 of the same pie-chart in one table, rather than just one big pie-chart. Our brain and eyes can rapidly detect one thing is not the same as the other, so we get context blisteringly fast.

💡Tip #3: A Moment Of Clarity Please!

A lack of clarity could put the brakes on any journey to success.

Steve Maraboli, Behavioral Scientist

For any data project (or any project for that matter!) to be successful, clarity is crucial. If you can secure clarity, then you can create clarity. Your project is doomed to failure, plus frustration and disappointment to boot, if your stakeholder or client has not clearly defined the questions the project should answer through your work.

💡Tip #4: Time Series Variation Fully Under (A) Control (Chart)

If sales drop by 15% in one week woeful cries of dismay and alarm a followed by the analysist’s attempt to query a whole load of random items to find the cause of the terrible affair. How on earth can an analyst ever hope to uncover the mystery of exactly which factors caused sales to decrease from the week before? Answer—with great difficulty (better we keep to impossible).

There is a magic in graphs. The profile of a curve reveals in a flash a whole situation — the life history of an epidemic, a panic, or an era of prosperity. The curve informs the mind, awakens the imagination, convinces.

Henry D. Hubbard(1870-1943), member of the U.S. Bureau of Standards

Help is at hand to pinpoint whether the 15% decrease is of real importance or perhaps it’s just a random variation based on historical results. The helping hand is the control chart.

Data are plotted in time order. The control chart contains a central line for the average. Then there is an upper line for the upper control limit and finally a lower line for the lower control limit – typically two standard deviations above and below the central line. Current data are compared to these lines and conclusions can be drawn on whether a variation is really unusual or more just a random variation. Using the sales percentage to week-over-week, you can clearly conclude whether the 15% increase was just plain normal or plain not.

💡Tip #5: A Time To Design… A Time To Format

The greatest value of a picture is when it forces us to notice what we never expected to see.

John Tukey, American mathematician

The whole idea behind data visualization is communicating to your target audience, inspiring trust and being downright credible. If your insight is worth sharing, then presentation truly matters. Investing in design and formatting leaves a wake of impressed followers. It’s such a waste if the data and analysis are fantastic, but they never resonate with anyone else but you.  You need to be engaging and insightful in your data visualization quest, not just a number-cruncher with a display. In short, calculate some time for design!

💡Tip #6: Must-Have Analyses

For all those who wouldn’t know a candlestick chart if it lit up in our faces, and those with little time, RetentionX is one of the smartest solutions on the market. Non-specialists can easily reap the benefits of a fool-proof data-analysis system with gorgeous, meaningful data visualization. RetentionX uses AI-driven data analysis and can be up and running within 120 minutes and fully integrated with existing systems. It reveals hidden value in data, helps to predict future outcomes and provides recommendations for action.



Shelby Temple, „5 Amazing Tips for Data Visualization“

SaaS expert operating at the interfaces of business, data and technology – Head of Implementation at Personio, Europe's leading HR software.

Write A Comment