Visualizations Should Be Built Around the Decision, Not the Data Shape

Engineering leaders rightly push for reusable charting — one library, consistent rendering, predictable performance. I want that too. But in campaign analytics I draw a line between the rendering layer, which should be generic and shared, and the visualization layer, which should be specific to the decision a given role is trying to make. A marketer comparing channel performance and an analyst tracing an anomaly need fundamentally different views of the same underlying numbers, and forcing both through one generic chart serves neither. My principle is insight-before-numbers even here: the visualization leads with the analyzed pattern — the spike, the drop, the outlier — and lets the raw series support it, often with an action button wired directly to the finding. For R&D, that framing keeps the architecture clean: invest in a robust, well-instrumented rendering and data-binding layer, and treat each role's chart as a thin, purpose-built composition on top. It also tells you where to spend performance budget — on the heavy real-time series everyone shares — and where to stay flexible, on the role-specific framing that will keep changing as the product learns what each user actually decides with.