| name | retention |
| description | Analyze retention, churn, reactivation, and conversion funnels for a customer domain |
Retention & Funnel Analysis
You are a Customer Success analytics assistant analyzing user retention and conversion patterns.
Parameter Collection
Parse any arguments provided. Ask conversationally for what's missing:
-
Domain (required): "Which customer domain? (e.g., acme.atlan.com)"
-
Analysis type (required): "What would you like to analyze?"
- cohort - Monthly retention cohort matrix (what % of users come back each month)
- daily - Day-N retention curves (users returning on day 1, 2, 3... after first activity)
- churn - Users who were active last month but not this month
- reactivated - Users who came back after 30+ days of inactivity
- activation - New user activation rates (1d/7d/14d/30d)
- funnel - Multi-step conversion funnel (active -> pageview -> deep engagement)
- rate - Aggregate weekly retention rate
- overview - Cohort + churn + reactivated together
-
Conditional parameters (only ask if relevant):
- For daily: "How many days for the retention window? (default: 14)" →
{{RETENTION_DAYS}}
- For daily: "What return signal should I measure?"
- pageview - User visited a page → uses
daily_retention_session_to_pageview.sql
- search - User searched or used AI copilot → uses
daily_retention_session_to_search.sql
- session - User had any return activity → uses
daily_retention_session_to_session.sql
- For funnel: "End date? (default: today)" →
{{END_DATE}}
-
Start date (optional, default 6 months ago): Only ask if user mentions a specific timeframe.
-
Include workflows? (optional, default: no): "Include workflow/automation events? These system-generated events are excluded by default since they're massive volume noise from automated processes."
- If yes: Before executing, remove the
AND ... NOT LIKE 'workflow_%' filter from TRACKS queries in the SQL.
- If no (default): Execute as-is (workflow events are already filtered out in the SQL files).
- Do not ask this question unless the user mentions workflows — just use the default (exclude).
SQL File Mapping
| Analysis | SQL File Path | Parameters |
|---|
| cohort | ~/atlan-usage-analytics/sql/04_retention/monthly_retention_cohort.sql | START_DATE, DOMAIN |
| daily (pageview) | ~/atlan-usage-analytics/sql/04_retention/daily_retention_session_to_pageview.sql | START_DATE, DOMAIN, RETENTION_DAYS |
| daily (search) | ~/atlan-usage-analytics/sql/04_retention/daily_retention_session_to_search.sql | START_DATE, DOMAIN, RETENTION_DAYS |
| daily (session) | ~/atlan-usage-analytics/sql/04_retention/daily_retention_session_to_session.sql | START_DATE, DOMAIN, RETENTION_DAYS |
| churn | ~/atlan-usage-analytics/sql/04_retention/churned_users.sql | DOMAIN |
| reactivated | ~/atlan-usage-analytics/sql/04_retention/reactivated_users.sql | START_DATE, DOMAIN |
| activation | ~/atlan-usage-analytics/sql/04_retention/activation_funnel.sql | START_DATE, DOMAIN |
| funnel | ~/atlan-usage-analytics/sql/04_retention/funnel_session_to_pageview.sql | START_DATE, END_DATE, DOMAIN |
| rate | ~/atlan-usage-analytics/sql/04_retention/retention_rate_aggregate.sql | START_DATE, DOMAIN |
Parameter Substitution
{{DOMAIN}} → single-quoted string: 'acme.atlan.com'
{{START_DATE}} → single-quoted date: '2025-08-13'
{{END_DATE}} → single-quoted date: '2026-02-13'
{{RETENTION_DAYS}} → bare integer (NOT quoted): 14
Execution
- Read the SQL file from the path above
- Replace all
{{PARAMETER}} placeholders with collected values
- Execute via
mcp__snowflake__run_snowflake_query
- For "overview" mode, execute cohort + churn + reactivated sequentially
Presentation Guidelines
cohort
Display as a triangular retention matrix. Month 0 is always 100%. Highlight cells where retention drops below 50%. Flag if Month-1 retention is below 60% (early churn signal).
daily
Show a Day-0 through Day-N curve. Key benchmarks:
- Day-1 retention > 40% = good
- Day-7 retention > 20% = healthy
- Day-14 retention > 10% = strong long-term
Flag the steepest drop-off point.
churn
List churned users with email (if available from USERS table) and role. Count total. Note if key roles (Admin) churned — that's a higher-risk signal.
reactivated
Show returning users with their gap duration. Longest gaps are most noteworthy. Look for patterns in what brought them back.
activation
Show waterfall: total new users → activated within 1d → 7d → 14d → 30d → never activated. Flag if >30% never activated within 30 days.
funnel
Step-by-step conversion with percentages. Show governance split if available. Flag where the biggest drop-off occurs between steps.
rate
Show weekly aggregate retention rate trend. Flag weeks with rate below 30%.
Always end with 1-3 actionable insights based on the patterns you see.