What Actually Makes Good Charts Useful
Scott Berinato's book Good Charts is a practical guide to making data visualizations that people can actually understand. It's not about fancy tools or advanced statistics. The core argument is that most charts fail because the designer hasn't thought clearly about what the viewer needs to take away from them. Here's the thing people miss when they're starting out: the chart itself is only half the problem. The other half is the audience's mental model. If you make a beautiful chart that your stakeholders can't process in three seconds, you've wasted everyone's time.
Good Charts Scott Berinato
The book breaks down chart design into two main categories. Simple charts that show one thing clearly, and complex charts that organize multiple relationships. The distinction matters more than most people realize. Most of the time, you should be trying to collapse your data into the simple category. Complexity is a failure of design, not a feature. I spent years building executive dashboards where the requirement was always "show everything." The result was charts nobody looked at for more than five seconds. Once I started cutting content ruthlessly, engagement actually went up. Stakeholders preferred getting three clear messages over thirty data points.
The Framework
The practical method starts before you open any visualization tool. You need to define three things: the viewer's question, the single insight you want them to leave with, and the chart type that makes that insight unavoidable. If you can't articulate the insight in one sentence, you don't have a chart, you have a spreadsheet with gridlines. This is where most people get stuck. They start building and then wonder why the result is confusing. The fix is to write the insight first on a sticky note. If it doesn't fit, go back to the data. Chart selection follows a simple logic. Use a bar chart for comparing categories. Use a line chart for showing change over time. Use a pie chart sparingly and only when you have fewer than five slices that add to a meaningful whole. Scatter plots work for relationships between two variables. Tables are fine when precision matters more than pattern recognition.
Get the Full Details
Common Mistakes I See Constantly
Three-axis charts. People love adding a third dimension because they think it shows "more." It shows confusion. If you have three variables to display, use color, size, or facets. Don't layer them all on the same axis system and hope the viewer figures it out. Starting every axis at zero when it doesn't matter. Berinato argues for this contextually. If you're comparing market share percentages, starting at zero makes the differences look tiny and meaningless. If you're showing absolute revenue in dollars, the scale matters. The rule is simpler than people make it: ask whether the origin point changes the interpretation. If it does, show it. If it doesn't, don't. Decorative elements that add zero information. Gridlines everywhere. Shadows. Gradients. 3D effects on pie charts. These are almost always distractions. A good chart looks boring. If yours looks interesting, you've probably added something unnecessary.
My Experience With a Tough Case
I once had to present regional sales performance across twelve territories with quarterly data over three years. The initial instinct was a heat map with small multiples. It was technically accurate and visually dense. Executives stared at it silently for about twenty seconds before asking if I could "make it simpler." The fix was to break it into two separate charts instead of one complex one. The first showed the overall trend across all regions as a single line, establishing the baseline. The second used a bar chart ranking the twelve territories against each other for the most recent quarter, with color indicating whether they were above or below the baseline. This cut the viewing time from about fifteen seconds to maybe four, and the discussion actually focused on the gaps instead of trying to read individual cells. Another edge case involved presenting survey data with Likert-scale responses. The natural instinct was a stacked bar chart showing the distribution. But stacked bars make it hard to compare the positive end across categories because the baseline shifts for each one. I ended up using a diverging bar chart centered around the neutral point. It took longer to build but made the comparison unambiguous.
What the Book Doesn't Cover Well
Good Charts focuses heavily on static visualizations for business presentations. It doesn't address interactive dashboards, which require a different set of decisions around filtering, drill-down paths, and responsive design. If you're building things in Tableau, Power BI, or Looker, the principles apply but you'll need additional guidance on interaction design. The book also predates the current emphasis on accessibility in data visualization. Color blindness affects roughly eight percent of men. Relying solely on color to encode information excludes a significant portion of your audience. Always pair color with shape, label, or pattern where possible. This isn't just a compliance issue, it's a readability issue that affects everyone.
Practical Workflow
Start with the insight you want. Write it down. Choose the simplest chart type that communicates it. Build it in a tool, review it against the insight, remove anything that doesn't serve it, and test it on someone who hasn't seen the data before. If they can't tell you the insight in their own words within ten seconds, start over. This process usually takes me about twenty minutes for a standard chart and up to an hour for something more involved. The alternative is spending two hours building something that gets ignored, which happens more often than I'd like to admit. The best resource after the book itself is practice. Build charts regularly, get feedback, and notice which ones people actually engage with. Patterns will emerge that no framework can predict. Your audience's expectations shift depending on context, culture, and the specific decisions they need to make. No single technique handles all of that, but the discipline of starting with the insight and working backward gets you closer than most approaches do.
