A cohort table without commentary gets read as a heat map
Cohort tables are prepared monthly. The person who looks at them interprets them, and each reading is different. People see the pattern they expect to see.
Small cohorts cause problems. A month with forty customers produces retention percentages that swing wildly. A colored cell suggests a signal that is just arithmetic.
The important comparison is seldom the one the table highlights. It is not this month against last month. It is this group at the three-month mark against all previous groups at the three-month mark. The table allows that comparison. It does not make the comparison clear.
Commentary also needs to say what changed for a cohort. A cohort that came from a discount campaign behaves differently for reasons unrelated to the product.
How the ActionFlow runs on the canvas
A PostgreSQL or MySQL node stores the cohort definition as a query. The definition stays in one place rather than being copied across reports. A Start Scheduled node runs the query once a month.
A Math node computes the retention grid and the corresponding cohort sizes. Size determines whether a difference is discussable.
A Filter node marks cohorts below the size threshold rather than hiding them, so a reader can see that a row exists but is too small to read.
A Google Sheets node provides details on what happened in each acquisition month: campaigns, pricing changes, and launches. This context turns a difference into an explanation.
A Generate Object node returns commentary: which cohorts differ significantly at the same age, the probable context, and a list of cohorts that are too small to draw conclusions from.
A Generate Text node writes the brief comment. A Create Excel node produces the table as a file. A Notion node stores both, so a month-over-month comparison does not require a rebuild.
Nodes this ActionFlow uses
- Start Scheduled: runs the report monthly.
- PostgreSQL (Data): holds the cohort definition as a single query. MySQL fits the same slot.
- Math (Util): computes the retention grid and the cohort sizes together.
- Filter (Data): marks undersized cohorts rather than hiding them.
- Google Sheets (Integration): provides the campaign and pricing context for each acquisition month.
- Generate Object (AI Core): returns meaningful differences, likely context, and cohorts too small to judge.
- Create Excel (Data): produces a shareable table.
- Notion (Integration): archives table and commentary together for month-over-month reading.
Frequently asked questions
Start building AI workflows
Create a free account, open a template or a blank canvas, and run your first ActionFlow.
Newsletter
Get product updates
New nodes, agents, and product notes. We send mail only when we have something worth opening.