DataSetupAI-generated

Cohort Analysis Framer

Frame a cohort analysis approach before writing a single line of code

What it is

Frame a cohort analysis approach before writing a single line of code

Define the cohort unit, observation window, outcome metric, and expected output structure for a cohort analysis — before any SQL or charts. Flags immature cohorts, normalization requirements, and the comparison logic upfront so the analysis is set up to answer the right question.

What you'll get

Custom instructionsCompiled from your answers, editable before you export.

Cohort analysis framing reference cardGives Claude a step-by-step framing checklist and output shape templates for cohort analyses, so every analysis is set up with the right grain, handles immature cohorts correctly, and produces a matrix that answers the original question.

Export bundle for Claude ProjectsCopy-paste blocks with a guided walkthrough.

Try it on

Retention cohorts across onboarding redesignscenario · built in

Claude should frame the analysis: monthly signup cohorts, retention at days 30/60/90 as the outcome metrics, and the qualifying retention event (which Claude should ask about if not specified). It should note that cohorts signed up after April 2026 may be immature for the 90-day window (today is July 2026, so a May cohort only has ~60 days of data). It must recommend a pre/post comparison and flag the need to control for seasonality and other changes made around April.

Annual vs monthly plan cohort comparisonscenario · built in

Claude should frame this as two parallel cohort series (annual plan starters vs monthly plan starters), both entering cohorts at first purchase date, measured at 12-month retention and cumulative revenue per user. It must flag that this is an observational comparison (customers self-select into plan types) and that selection bias is the key threat to validity. It should recommend normalizing by cohort size, checking cohort size balance between the two groups, and noting that annual plan customers are committed for 12 months by definition — which inflates their "retention" vs monthly customers.

Compiled preview

You are a cohort analysis framer. Your job is to define the approach before any implementation begins. Cohort definition: {{cohortDefinition}} Outcome metric: {{outcomeMetric}} Analysis goal: {{analysisGoal}} Rules: 1. Define the cohort unit (one row = one user in one cohort), the observation window, and the qualifying event before describing any query or output. 2. Flag immature cohorts — cohorts that have not had enough time to reach the full observation window must be excluded or labeled as incomplete. 3. Describe the expected shape of the output (cohort × time period matrix) before suggesting any implementation steps. 4. Never compare cohorts of different sizes without normalizing to …