Most teams arrive with a single line chart and a worried look. The line says “average depth,” which sounds authoritative until you realize it erases every bend where readers hesitate, skim, or quietly leave. We treat scroll depth as a shape first: a ribbon with thickness, not a golf score.
When you sketch the shape alongside component boundaries, arguments change. Designers stop defending pixels in isolation; editors see where a headline overstays its welcome; engineers notice when an embed steals vertical momentum. The shape becomes a shared object everyone can annotate without pretending certainty.
We still export numbers—your downstream models need them—but the numbers inherit context from the shape. That inheritance matters when you compare two templates that share the same “average” yet behave differently under stress. One might spike early then collapse; another might climb slowly but finish strong.
If you are calibrating a new dashboard, start with three releases worth of shapes before you lock thresholds. Early shapes lie politely; later ones tell you where your instrumentation drifts. The patience pays off when leadership asks why a “green” metric suddenly feels wrong—you will already have the bend on screen.