Join our "Building Machine Learning Models" interactive demo with Amazon SageMaker in New York July 18th

Register Now
Tips and Tricks

When Charting Data Discoveries, Less is More

As data tools evolve and become more accessible to citizen data scientists, processes are also going to have to adapt. Periscope released data discovery for business with the intention of bringing business knowledge into data analysis, but this move also comes with the charge of preparing these new analysts to communicate their data-based insights.

A few weeks ago, I shared my best tip for first-time data analysts to help anyone new to the analytical process dig into data and come away with valuable insights. But that initial discovery is only half the battle — insights need to be visualized in a way that tells a clear story for other stakeholders.

To help citizen data scientists chart their discoveries, I wanted to share a very important principle for communicating with data visualizations: maximize the signal, minimize the noise. When it comes to finding insights in data, you may have heard the terms “signal” and “noise” before. The signal is the important information, the data that illustrate the insight that you discovered. The noise is everything else, the content that clouds the communications and makes it difficult to see the important information.

Once you’ve found an insight and selected the right type of chart to tell your story, there are a lot of things you can do to increase the strength of your signal. Here are a few of the easiest ways to reduce noise and focus on your discovery.

5 Tips for Removing Noise from Charts

1. Show only one insight per chart — The easiest way to reduce the noise in your chart is to boil the finding down to its most basic element and limit the chart to show just that. The more information you decide to include in a chart, the more opportunities there are for noise to creep into the picture. If you have several related phenomena that you want to illustrate, build a dashboard that includes one chart for each. Like an author who writes a book of several chapters, good data storytellers break their insights down into bite-sized pieces of information and chart each one individually to minimize noise.

2. Only show one series at a time — Similar to the first tip, the best way to reduce clutter is by focusing a chart on one thing at a time. It might be tempting to try comparing several similar series of data on a single chart, but they’re often easier to digest (and far easier to understand) if they’re plotted on charts and placed side by side. When using several charts with different scales, this is especially true. Charting more than one series at a time often means you also need to include clarifying text or a key to explain all the information, so the number of assets on the chart starts snowballing. In the case that you’re comparing series across the same time span (year-over-year, quarter-over-quarter, etc.) it may make sense to stack those series on one chart, since they’ll share the same scale.

3. Use color to emphasize your insights — If the goal of a chart is to direct attention to a certain insight or piece of information, color is often the best tool for that. It can be used both as a means of increasing attention to specific data points or decreasing attention from distracting data points. In a dashboard, color can be used to show similar data points across several different charts to reduce noise and add continuity to your data story. As an alternative, secondary information can be shown in gray to reduce attention to data that might be distracting.

4. Reduce the amount of text in your chart — When it comes to the copy on your visualizations, less is always more. There should be a short, descriptive title and the axes should be identified, but any copy beyond that (like labels or annotations) is likely unnecessary. Anyone looking at your chart will read all of the words you’ve included, so be judicious about the amount work you’re giving them. Alternatives to removing text entirely are reducing the size of the copy, graying it out or making the text more transparent, but these tactics should only be done in rare circumstances. If the text is worth de-emphasizing, it’s probably worth removing entirely.

5. Limit the range of data you’re illustrating — When you’re determining how much of your data to show (how far to extend your axes), more information doesn’t always mean a more complete picture. Widening the range of information displayed is an easy way to include outliers and reduce the attention to the signal you’ve found. As your range gets larger, the individual points take up less space, so you should be judicious with the bounds that you decide to put on your information.

For more tips about data exploration and communicating data-based insights, check out our How to Chart Your Data Discoveries guide.

Want to discuss this article? Join the Periscope Data Community!

Christine Quan
Christine spends a lot of time thinking about data visualization theory and building tools to empower data teams. When she is not constructing SQL queries or building visualizations in R, Python, or Javascript, she can be found dissecting Taylor Swift lyrics through text analysis or analyzing emoji use in surveys.