Data visualization is both an art and a science, and selecting the right chart can be overwhelming. My advice is to start by getting something on the page to break the logjam. You can always refine or completely change your chart later.
AI serves as a great collaborative, brainstorming assistant, and a prime example is Recommended Charts in Excel. This post demonstrates its capabilities and how to make the most of it using a straightforward example.
Insert the basic chart
To get started, click anywhere inside the dataset you want to visualize and select Insert > Recommended Charts. I’m going to go ahead and click OK on the first one recommended to me.
Did you get different recommended charts in a different order from the previous menu? Welcome to the unpredictable world of AI! Since output is generated probabilistically, results can vary for different users. In that sense, it’s random.
Excel’s recommended charts are generally a good starting point, but some suggestions may be more suitable than others. Let’s examine the chart types Excel recommended in this example.
The next three charts (scatter, line, and funnel) all mistakenly treat the first column as a measure instead of a dimension. In the following section, we’ll explore how to correct this and make other necessary adjustments.
Improving the results
Recommended charts made a good choice with a bar chart. However, as with other Excel data sources, there are ways to optimize how well recommended charts perform.
Clearly distinguish measures and dimensions
Broadly speaking, we can separate variables between measures and dimensions. Measures in data refer to numeric values that can be quantified, such as sales revenue or temperature, while dimensions represent non-numeric attributes like categories or labels, such as product names or customer names. Although dimensions are typically non-numeric, their occasional presence as numeric data can lead to confusion for algorithms when determining their roles in data visualization.
In our example, Excel might mistakenly interpret a variable labeled 1 through 5 as a sequence, making it appear suitable for a line chart. To avoid this confusion, use labels that clearly distinguish between measures and dimensions. In this case, using region names instead of numbers could be more useful:
Use labels
Including column header labels in Recommended Charts helps Excel interpret the data accurately and ensures the resulting chart is appropriately labeled for better data understanding. Go ahead and add them to the data:
After making these changes, insert a new Recommended Chart, and you’ll notice that although some poor choices like a line chart have been removed, they are replaced by other inadequate options like a map:
Excel’s assumption that this data is geographical isn’t quite accurate, as these aren’t real places.
With AI assistance, we chose the bar chart as a safe option. There’s room to improve this plot by sorting and removing gridlines. AI can help you get started, but your human expertise takes the plot from good to great.
How is this AI?
Recommended Charts is an example of AI because it uses algorithms to analyze the data you select and suggests the most suitable chart type for effectively visualizing that data. By understanding the data’s nature and relationships, the AI identifies patterns and trends, recommending the chart that best communicates the insights in a meaningful way.
Compare your file with the solutions:
Do you typically start with Recommended Charts when building a visualization in Excel? Why or why not? Do you have examples to share of the algorithm working particularly strongly or poorly? Let me know in the comments.
Leave a Reply