Many companies are implementing self-service reporting for the variety of benefits discussed in our previous article “Analyze, Pivot or Die: 4 Reasons Why Every Company Needs Self-Service Reporting.” At this point, we hope you have followed our advice and your business is not at risk to fail. But how do you get your BI approach to the point that you can get off “decision making life support”? And, by “life support” you know what we mean: the self-service reporting datasets are in place, and the business is free to create reports. Unfortunately, only a few reports are inherently useful. In fact, most of your reports do NOT convey information quickly and accurately AND do not allow you to quickly act on the information available on the report.
So, how do you design BI reporting that is easy to understand, visually appealing, and provides actionable insights? Read further and find out how professional BI staffs address the issues.
How can you create clear, actionable reports?
You must choose your visual type with intention. Choosing the right type of visual is key to understanding your data. The wrong type of visual can confuse the user or even result in misinterpreting the analysis.
How do you choose the type of visualization that is right for your analysis?
First, you need to have a clear question that you are trying to answer. In addition, it is equally important to understand the action that you want to result from that answer. In other words, what are you going to do with this information?
Type of Analysis – Part-to-Whole
One of the most common types of analysis is called “Part-to-Whole.” Part-to-whole analysis is a great way to see how a metric is broken out into subsets. For example, understanding how each category contributes to total sales dollars. You can think of part-to-whole analysis in this way:
How does each subset contribute to the total?
Although this analysis sounds straightforward, we often see client reports that either greatly confuse the user or produce no actionable insights.
There are many different visuals that can be used for part-to-whole analysis in order to produce actionable insights. We will take you through a few of them and discuss pros and cons of each as well as some considerations to help you be sure if part-to-whole analysis will answer your question.
Visualization Examples – Part-to-Whole
For these examples, we’re using the Iowa Liquor Sales data from Google BigQuery’s public datasets to analyze Sales Dollars by Category.
Treemap – Category
The rectangles give us a rough idea of the size of each category to the 2022 total. Adding labels for Category and the percentage of total help us quickly see the percentage of each category.
Insight – Whiskey makes up the majority of 2022 sales with Vodka being second.
Potential Action – Develop inventory reporting to make sure Whiskey is fully stocked.
Treemap – Subcategories
We can add the subcategories to the treemap to analyze the size of each category, as well as the subcategories that make up each category. One of the downsides of adding subcategory is we cannot see the total category percentage. However, you could structure the data with a hierarchy of Category and Subcategory so you can create drill-through functionality.
Insight – Although Whiskey is the largest category, American Vodka is the largest subcategory out of all the subcategories.
Potential Action – Review physical store layout and ensure there’s enough shelving space for American Vodkas.
Stacked – Horizontal
The stacked bar visual is an easy way to view how each Category contributes to the total. You can even add a reference line to quickly see proportions.
Insight – Whiskey and Vodka make up more than 50% of 2022 sales.
Potential Action – Analyze Whiskey and Vodka cost to find ways to maximize profit.
Stacked – Vertical
Stacked bars can also be vertical. Looking at the data this way allows you to see how the contribution to total changes over time.
Insight – Between 2018 and 2022 Tequila has consistently contributed more to total sales while Vodka has contributed less.
Potential Action – Analyze inventory to make sure that Tequila is adequately stocked, and Vodka is not excessively stocked.
Pie Chart
A pie chart is the first data visual many of us created, but this type of visual has gone out of favor. It’s very difficult to see which categories are larger than others unless the data is sorted by the metric. For example, we only know that Tequila sales are larger than Rum sales because of their order. If this was sorted by Category name instead, we wouldn’t be able to tell which was larger.
Many analysts strictly avoid pie charts and opt for an easier to read chart. Know your data well enough that you can use discretion. If there are only two categories, like Yes/No, it may be useful, but if there are many categories you would probably do well to choose a different chart.
Insight – Whiskey is the largest category. Without the percentage labels and sorting it’s difficult to understand how each category contributes to the total.
Donut Chart
A very similar chart is a donut chart. The same considerations for a pie chart apply to a donut chart. The only difference between a pie chart and a donut chart is the empty space in the middle.
Insight – With no labels we can see Whiskey is the largest category but are unsure of the percentage of total. The total sales dollars is helpful to get a sense of magnitude.
Considerations – Part-to-Whole
When deciding if part-to-whole analysis is the right visual type to choose, go back to your question and make sure that it is clear. Do you want to know the order of categories from largest sales to smallest? We’ve done some of that in the visuals within this piece, but there are other visuals better suited for ranking. Do you want to know the actual sales dollar amount for each category? In that case you want to choose a visual suited for magnitude rather than part-to-whole.
If you are making an interactive report, it is always helpful to add tooltips to your visuals for additional information. In this example, adding a tooltip with Sales Dollars could be helpful. You could also add drill-throughs so you can drill down into subcategories.
Finally, keep in mind the number of categories you are analyzing. In this example, Whiskey and Vodka are a clear majority and there are a reasonable number of categories, but if you have many categories or categories that make up near-equal parts, you should choose a different visual.
In the next article in this series, we’ll be talking about flow visuals. Let us know if you have any question or comments on the use of Part-to-Whole analysis. We’d be happy to work with you.
For more information on this subject or to discuss how Verstand can help you transform your visual analysis into actionable insights, contact us at insights@verstand.ai.
