Method plate
The Survey Grid
A product is not a dashboard. It is a field of behaviours that can be named, windowed, and redrawn when the ground shifts. The Survey Grid is the house method for doing that work in App Analytics without letting a vendor schema decide what counts as true.
The grid is a contract: every cell is a verb a person could have refused.
Four cells, always
Whatever the stack, the grid asks the same four questions. What happened (the event)? Who is it attributed to (the identity seam)? Over what return rhythm (the window)? What is the Monday claim allowed to say (the plot)? Skip one and the others start performing.
Teams often arrive with fifty events and no verbs. The first survey pass is subtraction. If a cell cannot be spoken in a sentence that a designer would recognise as a human action, it comes off the grid until it can.
Isolines, not a single ridge
Retention is drawn as a set of contours: D1, D7, D14, D28 only when those days match how the product is actually used. A meditation app and a mortgage tracker do not share a week. The Survey Grid refuses to print a lonely D7 because a slide template had a hole for one.
What the grid will not do
It will not replace a warehouse. It will not certify Mixpanel, Amplitude, or GA4. It will not invent a stitch between anonymous and known users when the product has not earned one. Those limits are taught in Grid Instrumentation so that nobody leaves thinking the method is a magic layer on top of a messy log.