What is more important? Data (ETL, data modeling) or data visualization?

⚠ The visual part of business report development is often heavily underestimated (by developers and their customers – business users). Data preparation is important; data modeling is important. But the only thing the end users will see is a visual representation of the data 📊 . The visual representation can hide important patterns and trends and exaggerate what is not important; it can be misleading and lead to wrong decisions; it can be hard to understand and hard to use; it can introduce errors into perfectly clean data.

➡ Good data visualization requires knowledge, skills, and experience. And it’s not as straightforward as 2 + 2 = 4. There are some rules/best practices, but there are many ways to reach the goal. In Power BI, we often need a lot of tricks and workarounds (data modeling tricks, overlays of multiple visuals, repurposing available format options, e.g., using error bars to display additional data series) to build good data visualizations, and there are a lot of things to learn (calculation groups, Deneb and Vega/Vega-Lite, SVG, HTML/CSS).

➡ Also, there is no such thing as a perfect data model that can be used for good dataviz without needing to write more DAX and even to modify the data model (in many cases, you need tricky data modeling solutions to enable proper data visualization). Often, during attempts to visualize data, you will understand what data you need, what data errors have to be fixed, and whether you should exclude outliers from the report or highlight them. In this case, you have to go back to the data source, make changes there, then adjust the data model.

➡And in my practice, when someone says, “I have a 100% complete data model; I just need someone to make a good report, it shouldn’t take much time,” they are deeply underestimating what they have (they often don’t have what they think they have; I’ve heard “we have DWH” when there was just a few flat tables with dummy data), and they are deeply underestimating what it means to build a good report.