Visuals: signal over style

For years, data visualization on the internet drifted toward spectacle. Dashboards filled with gradients, shadows, animations, and exotic chart types promised insight by way of polish. This week, a small burst of conversation pushed in the other direction. A widely shared thread argued that most charts still live or die on a simpler question: did you pick the right form to reveal the pattern in the data? The old workhorses — line charts for change over time, bar charts for comparisons, scatter plots for relationships — keep winning because they make the signal easy to see. That argument lands because it is not nostalgia. It is the core of the field. (clauswilke.com) The reason is almost embarrassingly basic. Chart choice starts with the question, not the styling. If you want to show how something changes across time, a line chart usually beats everything else because time gives the data an inherent order. If you want to compare amounts across categories, bars work because length on a common baseline is easy to judge. If you want to see whether two variables move together, a scatter plot does the job directly. “From Data to Viz,” a widely used chart-selection guide, is built around exactly this logic: start with the structure of the data and the task at hand, then narrow to the graphic that fits. (data-to-viz.com) That sounds obvious until software gets involved. Modern tools can generate dozens of chart types with a click. Datawrapper alone offers more than 20. The abundance is useful, but it also creates a trap. When every chart is available, people start choosing by novelty instead of need. The result is often a chart that asks the reader to learn a visual system before they can learn the data. That extra effort is the real tax of flashy graphics. It slows comprehension before the argument has even begun. (datawrapper.de) The field has been warning about this for a long time. Claus Wilke’s open textbook, which readers surfaced again alongside the thread, frames visualization as a science before it becomes a style exercise. His examples keep returning to a core set of charts because those are the forms people actually encounter in research papers, reports, and news stories. He is blunt about decorative excess. Gratuitous 3D, he writes, adds a dimension that carries no data at all. It decorates the plot and makes values harder to read. The same problem shows up in bubble charts, where viewers must compare both position and area at once, a combination that weakens perception rather than sharpening it. (clauswilke.com) Once you accept that, the “signal versus noise” language becomes less like a slogan and more like a working method. Noise is not just gaudy color. It is anything that competes with the pattern you want the reader to notice. Storytelling With Data, another set of teaching materials that has spread far beyond classrooms, makes this point with almost mechanical regularity: borders, heavy gridlines, redundant labels, legends that could be direct labels, and default settings from Excel or Tableau all add cognitive burden. Removing them helps. But even that is not enough if the chart itself is wrong. A recent post on the site makes the hierarchy explicit: decluttering improves a chart, but if the chart type does not match the story, readers still will not retrieve the insight quickly. (storytellingwithdata.com) That is why the current turn back to fundamentals matters. It is not a rejection of craft. Good annotation, careful ordering, and restrained color still matter because they guide attention once the right chart is in place. But those choices are secondary. Data-to-Viz warns against spaghetti lines, overplotting, rainbow palettes, and counterintuitive axes for the same reason: each one makes the reader work harder to extract the pattern. The best visual is often the one that feels almost plain on first glance and then gives up its meaning immediately. In practice, that often means a sorted bar chart, a clean line running left to right, or a field of dots with one trend you can spot before you finish your first look. (data-to-viz.com)