The Visual Analytics Alphabet Series – B is for Best Practices!
- A is for Aggregated Data and APIs
- B is for Best Practices (this article!)
Today, is brought to you by the letter B… which is for Best Practices.
No discussion about building reports is complete without discussing best practices: what you should and should not do (I’m looking at you pie charts) when you build reports.
Creating a good report in Visual Analytics isn’t just about dropping charts onto a page. A good report is thoughtful, focused, and easy to use. It’s designed with the audience in mind from the very beginning.
Before you even jump into Visual Analytics, you need to have a plan, know what you want to show, and who you’re showing it to. From there, it’s about picking the right visuals, focusing on what actually matters, and organizing everything so it’s easy to follow. Finally, you need to test the report to ensure that it functions the way you intend and it’s easy to interpret.
Let’s consider a scenario. I would like to build a report that analyzes horror movies, how horror has evolved over time, how individual movie franchises perform, and what makes a horror movie successful.
Draft a Plan
First, we need to select the data we want to use for the report, think about the audience, craft our data story, and sketch a design of the report.
For the data, we will use our HORROR_MOVIES table that contains a description of various horror movies, the tagline, the release date, original language, genre, the collection, a popularity score, budget, revenue and runtime. We also want to include information about kill counts, so we have joined the data with our KILLCOUNTS table that contains information about the total confirmed kills for the movie.
Note: Details about joining the data will be discussed in a future article.
For more details about the data and the data preparation steps undertaken, please see The Visual Analytics Alphabet Series – The Data.
The audience for the report will be horror fans worldwide. Because the report is intended for a general audience, it’s a best practice to use simple graphs (like key value objects, bar charts, and tables) rather than advanced objects (like statistical models, forecasting, or text topics). In addition, when less common objects (like box plots or treemaps) are used, we will explain the content of the object or present that information in an alternative format (like with tables or crosstabs).
For the data story, we want to show the audience how horror has changed over time, which franchises lead the way, and showcase the movies that really stand out, as well as those that fall short.
This story gives us a good idea of the design of the report. We will have one page that focuses on the evolution of horror (by providing kill counts by decade), one page that compares franchises (by providing horror metrics by collection), and one page that shows the top and bottom performers by each horror metric (by displaying top 10 and bottom 10 performers for each metric).
Choose the Best Chart
Next, we choose the best chart for displaying the data. Let’s consider each page separately.
Evolution of Horror
This page showcases how horror movies have evolved from earlier films to modern releases. We can illustrate this evolution by looking at kill counts by decade.
Key value objects can be used to highlight one important fact; in this case, the average kill count by decade. It’s interesting how average kill counts spiked in the 1980s (thanks to the release of Return of the Living Dead which saw over 4,000 kills) fell off a bit in the 1990s and 2000s and rose again in the 2010s (thanks to the release of a relatively unknown movie, Dude Bro Party Massacre III which also saw over 4,000 kills).
Box plots can be used to show variations of a metric, like the average response, the minimum and maximum values, and any outliers. Because many people are unfamiliar with box plots (and our report is targeted to a general audience), we will add an explanation of the chart (using a text object).
Note: For this boxplot, outliers are ignored (which can be seen by the higher averages for some decades than the box portion). We chose to ignore outliers for display purposes.
Franchise Showdown
This page compares franchises (collections) across subgenres in terms of movie metrics like revenue, ratings, and kill counts.
To show the portion that each subgenre/franchise contributes to the whole genre, we can use a treemap. The size of the tile represents the total revenue for that subgenre/franchise and the color indicates the average kill count. It looks like standalone films (not part of a franchise) and iconic horror movies (like the Halloween franchise) have the highest total revenue, while zombie movies (unsurprisingly) have the highest kill count.
Because treemaps may be difficult for some users to interpret, we will display the same information in an interactive crosstab.
Note: A hierarchy is used to add drill-down functionality to the crosstab. Hierarchies will be discussed in more detail in a later article.
In addition, key value objects can be used to show the number of movies, the average revenue, the Average IMDb rating, and the average kill count for each selected subgenre/franchise.
Key value object showing horror metrics for selected franchise
The Best…and the Rest
This page showcases the top and bottom performers in terms of movie metrics like kills per minute, revenue, return on investment (ROI), and ratings.
Bar charts are useful to compare values and, when sorted, are great for showing rankings (like top 10 and bottom 10). In addition, bar charts are widely recognizable, easy to interpret, and perfect for a general audience. We can see that the two movies with the highest kills per minute are the same ones we identified that increased the averages for our kill counts by decade: Return of the Living Dead and Dude Bro Party Massacre III.
Top 10 and bottom 10 movies by kills per minute
Focus on What’s Important
After you have chosen the best chart to display your data, you want to ensure the content is organized, easy to navigate, versatile, and attractive.
For our report, we will organize content by adding multiple pages to the report, one to focus on each part of the data story (the evolution of horror, comparison of franchises, and top/bottom 10 movies).
We can ensure the report is easy to navigate by including a table of contents, adding descriptions for each page, and providing instructions for all features.
The table of contents will display a summary of the report (and motivations and conclusions that can be drawn from the report), as well as a link to and description for all pages.
Each page will also contain a header that provides the title of the page and a summary.
Interactive objects will include titles that indicate the actions that can be performed and will enable users to better understand the interactivity of the report.
To ensure that the report is versatile we will add ranks, prompts, and actions. Ranks can be used to show the leaders (top 10) and laggers (bottom 10).
Page prompts enable viewers to narrow the top and bottom movies by a specific subgenre. Notice that text is used to explain the purpose of the prompt.
Actions add interactivity to the report and enable viewers to take part in the data discovery process. For our report, we will enable viewers to see metrics for a selected subgenre/franchise.
Finally, we use styles to make the report more attractive. Specifically, we use colors that evoke emotions related to horror movies (red for kill counts to evoke danger and fear, gold for revenue to evoke success and achievement, and neutral colors for ratings for objectivity).
Consider the Layout
For each page, you want to arrange pages and visuals to communicate your data story. Our report contains a limited number of pages (four including the table of contents) and each page contains a limited number of objects. This enables viewers to focus on the data for easier understanding.
To enhance the visibility of the report and direct the viewer’s attention to the main part of the story, the most important objects on each page are the largest objects. All other objects provide supplemental information without detracting from the focus of the page. On this page, because the box plot is the largest object, the audience will spend more time studying it and the key value objects are used to provide additional details.
Objects are placed to provide a focal point and to drive comparisons. On this page, bar charts are placed in columns to indicate that the objects should be compared. Containers are also used to group objects and organize content. Specifically, a stacking container is used show the same details (top 10 and bottom 10) for various metrics.
When a page contains actions, the objects should be arranged in a logical order: the source object should be placed on the left (for cultures that read left to right) and the target objects should be placed on the right. On this page, selecting a tile in the treemap updates the values displayed in the key value objects.
Objects with actions are arranged in a logical order
Test, Test, and Test Again
Finally, you will want to test the report to ensure everything works the way you intended. You can test the report by evaluating the performance, conducting a peer review, viewing it on multiple devices, or exporting content.
Note: Evaluating the performance of the report will be discussed in a later article.
At the end of the day, building a great report isn’t about using every feature available, it’s about using the right ones: the ones that are right for your story and your audience. The best reports are those that are clear, focused, and thoughtfully designed. Because in the end, the goal isn’t to have a report, it’s to ensure that your audience understands your data.
For more information about best practices when building reports, check out these resources:
Tricks for SAS Visual Analytics Report Builders
Ready to see what SAS Viya Copilot can do?
Visit the Tips & Tricks page for setup guidance, demos, and practical examples that show how Copilot supports your workflows.
