Journal · 2 February 2026
Survival curves versus classic cohort grids
A classic weekly grid answers “of the people who started in this window, what share did the retained event in each later week?” It is a good accounting picture. It is a poor picture of when leaving concentrated, because each cell is a stock, not a hazard.
A survival curve, even a simple one, asks how long until the last observed return. Product trios often find the curve more intuitive once they stop expecting a heatmap. The cost is familiarity: boards have muscle memory for coloured squares.
In the flagship studio we draw both from the same extract. If they disagree in a way you cannot explain, the definition of “return” is usually the culprit — a person who returns every 40 days will look dead on a weekly grid and merely sparse on a curve.
Neither chart is causal. A curve that flattens after a feature launch still needs a comparison group. We include this warning because students love a shape that looks like a verdict.