The 10 Commandments of Data Visulizations
Thou shalt not take thy data source for granted.
The first step of every data visualization should be to figure out where it is from and how often it updates. Ideally, this should be information available to the end user as well. Even more ideally, the dataset should be able to tell you if it isn’t up to date so you know what/when to find a fix.
In addition, whatever data visualization platform you’re using, it will have limitations for data access. You should be familiar with them and be able to find ways around them, like APIs, automated connectors or even email updating. No matter what, your approach to the rest of the project will be dictated by these limitations, so try to choose a platform that minimizes them.
Thou shalt not use conflicting/meaningless colors.
It might seem obvious that colors should be clear and distinct within a chart. It might also be obvious that these colors should match company branding where possible, especially if they are being shared widely. However, a typical data visualization project will involve multiple charts, and there for you should match colors across charts. If you’re using blue to represent dollar volume on one page, that should carry over to all other pages. You don’t know how the user might export and remix the information you’re giving them, so planning for consistency ahead of time will avoid presentation issues down the line.
If you’re feeling overwhelmed by selecting a color palette, choose colors for your 3 most common variable from the start and choose the rest as you go or come back and unite them at the end of the project.
Thou shalt not use more than two visualizations for the same information.
Look at your dashboard. Even great data communicators will notice cards that represent the same information. It’s okay to have some duplication across pages to reduce searching and clicking around for related info, but if you were on a roll making charts, you probably made two of the same right next to each other.
When you’re in the planning stage, you should have datasets related to the trend values and present values. If this is the case, examine whether your duplicate chart might work better as a trending data display.
Thou shalt not use misleading scales
Scales can be annoying and it might be a good idea to stick with the defaults most of the time. Don’t be tempted to manipulate scales to make it look like the data is “doing something.” Further, don’t confuse statistical and practical significance (see point 7 in this article on misleading graphs). Just accept it if you have two values that are the same (but don’t forget to double check that’s actually true!)
Thou shalt not expect the audience to notice patterns on their own
Information that is related should be grouped to tell a story on the same page. Most people can’t hold relevant numbers in their mind while flipping across pages. True math nerds have a tendency to forget that use mere mortals can’t do this!
It’s helpful to find out what the end users/executives want to find so you know what visualizations will answer that business question.
Thou shalt not suffocate the insight
Let the data breathe, and don’t put too much information in one chart. This might seem like a contradiction to not making duplicates, but if you are using a good platform, this should be achievable 9 out of 10 times. For instance, a bar chart with an overlapping line should still be legible, eliminating the need to show a second bar chart for your second metric. It should always be clear what action to take on the insight as well, with goal lines or ranges marked clearly to make the insight useful.
Thou shalt always pass the squint test
There should be enough white space available that all data labels can be read easily. This is known as a squint test, which basically dictates that the user should be able to squint at your visualization and roughly understand what it’s trying to convey.
Thou shalt not use chart types most people haven’t seen since middle school
Box and whisker plot, we’re looking at you. Lots of analysts or technical users may feel tempted to vary their dashboards by using different chart types. While varying chart types can keep your page looking fresh and exciting, if it comes at the expense of legibility it isn’t accomplishing the goal of actionable insights. Be sensitive to your audience and stick to something easy to read wherever possible.
Thou shalt label everything
The user should not have to click, page down or take any action to see the labels of the crucial insights on your chart. It’s okay to have extra information in descriptions or on other pages, but don’t leave room for interpretation on bars, pie wedges or axes, especially where the data may be shared in the form of a screenshot.
Thou shalt not provide too much information
Don’t forget about data security after you’re done making your visualization. Now that you’ve done all the work keeping your audience in mind, you need to make sure the correct audience is accessing the correct information. This is both for ease of use and for privacy. Some tools offer user-based or group based permissions, so be sure to take advantage of these features to enhance the experience for the user and declutter information based on need.
Talk to us about our data visualization philosophy and tools you can use to organize your data into actionable insights.