Grid Method

Three layers, one argument you can take to finance


Push Notification Performance Analytics fails when teams flatten everything into a single open rate. The Grid Method keeps delivery, queue, and cohort on separate axes until they earn the right to be combined.

Server racks suggesting queue infrastructure

Delivery layer

Accept, retry, display, collapse. Vendor “delivered” is only the first cell. OEM battery rules and dual-SIM token decay belong here, not in creative retrospectives.

Queue layer

Worker age, backoff, and the gap between your app’s “sent” row and the platform accept. If latency lives in your house, stop blaming user fatigue.

Cohort layer

Who still taps after week four, under which attribution window, and after which quiet-hour policy. This layer is useless if the first two are fiction.

Analyst working on a laptop with charts

Why we refuse a single score

Composite “health scores” look premium in a board pack and hide the only question that matters: which layer broke. A beauty retailer in our case notes looked excellent on a blended index while Android display was quietly collapsing.

The Grid Method is taught in full inside the Delivery Intelligence Lab. Shorter briefings cover only the delivery layer so leadership can hear vocabulary before budget season.

We do not claim the Method replaces your warehouse. It is a reading order and a set of labels. If your event names are a mess, the grid will make that embarrassment earlier — which is the point.

See how programmes use it

When not to use it

If you send fewer than a few thousand pushes a month, or you only use in-app messages, the Grid Method is overhead. Start with a Desk Seat or wait until push is actually a system, not an occasional announcement.

Ask whether the Method fits your stack