Mixpanel Data Inspector

Explore Mixpanel schema, discover events, and analyze data structure.

Pre-flight Check

Check credentials and local data availability:

!$(mp auth test 2>&1 || echo "⚠️  No credentials configured")
!$(mp inspect tables --format table 2>/dev/null || echo "No local tables found")

If no credentials: Suggest running /mp-auth first.

Context Detection

Guide based on data availability:

• No local tables: Focus on live schema discovery (events, properties, funnels, cohorts)
• Tables exist: Offer both local exploration and live discovery
• No credentials but tables exist: Focus on local data exploration

Operation Selection

Choose an inspection operation from $1 or present menu:

Category 1: Discover Schema (Live API)

1. events - List all tracked event names
2. properties - List properties for a specific event
3. funnels - List saved funnels
4. cohorts - List saved cohorts

Category 2: Explore Local Data

5. tables - List local DuckDB tables
6. schema - Show table column definitions
7. sample - Preview random rows from table
8. breakdown - Event distribution in table

Category 3: Analyze Patterns

9. keys - JSON property keys in table
10. column - Deep column statistics
11. distribution - Property value distribution (JQL)
12. daily - Daily event counts (JQL)

---

Operation 1: events - List All Events

Discover all event names tracked in your Mixpanel project.

!$(mp inspect events --format table)

Output: List of all event names alphabetically sorted.

Filter with jq (JSON format only):

# Get first 5 events
mp inspect events --format json --jq '.[:5]'

# Find events containing "User"
mp inspect events --format json --jq '.[] | select(contains("User"))'

# Count total events
mp inspect events --format json --jq 'length'

Chain suggestions:

• Pick an interesting event → Run /mp-inspect properties <event-name> to see its properties
• Ready to fetch specific events? → Run /mp-fetch with filtered event list
• Want to see trending events? → Use CLI: mp inspect top-events --format table

---

Operation 2: properties - Event Properties

Show all properties tracked for a specific event.

Determine Event Name

• Use $2 if provided (e.g., /mp-inspect properties PageView)
• Otherwise, ask user for event name
• Suggest running events operation first if unsure

Execute

!$(mp inspect properties -e "<event-name>" --format table)

Output: Property names, data types, and example values.

Chain suggestions:

• Check property values → Run /mp-inspect distribution <property-name>
• Use in fetch filter → Run /mp-fetch with WHERE clause on this property
• Query locally → Fetch data first, then use /mp-query sql with properties->>'$.propname'

---

Operation 3: funnels - Saved Funnels

List all saved funnel definitions in your Mixpanel project.

!$(mp inspect funnels --format table)

Output: Funnel ID, name, and step count.

Chain suggestions:

• Analyze specific funnel → Run /mp-funnel <funnel-id>
• Create custom funnel → Run /mp-funnel and choose custom option
• See funnel details → Use CLI: mp query funnel <funnel-id> --from <date> --to <date>

---

Operation 4: cohorts - Saved Cohorts

List all saved cohorts in your Mixpanel project.

!$(mp inspect cohorts --format table)

Output: Cohort ID, name, count, and description.

Chain suggestions:

• Fetch cohort profiles → Use CLI: mp fetch profiles --cohort <cohort-id>
• Analyze cohort behavior → Run /mp-retention with cohort context
• Filter events by cohort → Use in /mp-fetch WHERE clause

---

Operation 5: tables - Local Tables

List all DuckDB tables in your local workspace.

!$(mp inspect tables --format table)

Output: Table names, row counts, and last fetch timestamp.

Chain suggestions:

• No tables? → Run /mp-fetch to get started
• Explore table structure → Run /mp-inspect schema <table-name>
• Preview data → Run /mp-inspect sample <table-name>
• See event distribution → Run /mp-inspect breakdown <table-name>

---

Operation 6: schema - Table Schema

Show column definitions for a specific table.

Determine Table Name

• Use $2 if provided (e.g., /mp-inspect schema events)
• Otherwise, list tables and ask user to choose
• Run tables operation first if needed

Execute

!$(mp inspect schema -t "<table-name>" --format table)

Output: Column names, data types, and nullable status.

Key columns:

• event_name - Event type identifier
• event_time - Timestamp of event
• distinct_id - User identifier
• properties - JSON column with all event properties

Chain suggestions:

• Preview data → Run /mp-inspect sample <table-name>
• See what events exist → Run /mp-inspect breakdown <table-name>
• Query the table → Run /mp-query sql with this table
• Explore JSON properties → Run /mp-inspect keys <table-name>

---

Operation 7: sample - Preview Data

Show random sample rows from a table.

Determine Table and Sample Size

• Table: Use $2 if provided, otherwise ask
• Sample size: Default 10 rows, can specify with -n flag

Execute

!$(mp inspect sample -t "<table-name>" -n 10 --format table)

Output: Random rows with all columns.

Interpreting JSON properties:

• Properties are stored as JSON: {"country": "US", "plan": "pro"}
• Use properties->>'$.key' in SQL to extract values
• Run /mp-inspect keys to see all available property keys

Chain suggestions:

• See all property keys → Run /mp-inspect keys <table-name>
• Analyze specific column → Run /mp-inspect column <table-name> <column-name>
• Build SQL query → Run /mp-query sql to analyze data

---

Operation 8: breakdown - Event Distribution

Show event counts, unique users, and date ranges for each event in a table.

Determine Table Name

• Use $2 if provided
• Otherwise, list tables and ask user to choose

Execute

!$(mp inspect breakdown -t "<table-name>" --format table)

Output: For each event:

• Event name
• Total event count
• Unique user count
• First and last occurrence dates

Chain suggestions:

• Analyze specific event → Filter in /mp-query sql with WHERE clause
• Build funnel → Run /mp-funnel with discovered events
• Time-series analysis → Run /mp-inspect daily <event-name>

---

Operation 9: keys - JSON Property Keys

Show all unique JSON property keys in a table, optionally filtered by event.

Determine Parameters

• Table: Use $2 if provided, otherwise ask
• Event filter: Optional, use -e flag to filter by specific event

Execute

All events:

!$(mp inspect keys -t "<table-name>" --format table)

Specific event:

!$(mp inspect keys -t "<table-name>" -e "<event-name>" --format table)

Output: List of property keys found in the JSON properties column.

Using property keys in SQL:

-- Extract property value
SELECT properties->>'$.country' as country FROM events

-- Filter by property
SELECT * FROM events WHERE properties->>'$.plan' = 'pro'

-- Aggregate by property
SELECT
  properties->>'$.country' as country,
  COUNT(*) as events
FROM events
GROUP BY 1

Chain suggestions:

• Deep dive on property → Run /mp-inspect column with property path
• Check property distribution → Run /mp-inspect distribution <property-name>
• Use in query → Run /mp-query sql with properties->>'$.key' syntax

---

Operation 10: column - Column Analysis

Perform deep statistical analysis on a specific column.

Determine Parameters

• Table: Required
• Column: Required (can be JSON path like properties->>'$.country')

Execute

Simple column:

!$(mp inspect column -t "<table-name>" -c "<column-name>" --format table)

JSON property:

!$(mp inspect column -t "<table-name>" -c "properties->>'$.country'" --top 20 --format table)

Output:

• Null count and percentage
• Cardinality (distinct values)
• Top values with counts
• For numeric columns: min, max, avg, median, stddev

Chain suggestions:

• High cardinality? → May need grouping or binning in analysis
• Low cardinality? → Good candidate for segmentation/breakdown
• Many nulls? → Check data quality, may need filtering
• Use in segmentation → Run /mp-query segmentation --on <property>

---

Operation 11: distribution - Property Distribution

Analyze value distribution for a property using JQL (JavaScript Query Language).

Required Parameters

• Property name (e.g., country, plan, device_type)
• Date range: --from and --to flags

Execute

!$(mp query jql - --from <from-date> --to <to-date> << 'EOF'
function main() {
  return Events({
    from_date: params.from_date,
    to_date: params.to_date
  })
  .groupBy(['properties.<property-name>'], mixpanel.reducer.count())
  .sortDesc('value');
}
EOF
)

Output: Property values with counts and percentages.

Chain suggestions:

• Fetch filtered data → Run /mp-fetch with WHERE clause on high-value properties
• Local analysis → Fetch data, then use /mp-query sql for complex queries
• Segmentation → Run /mp-query segmentation --on <property> for time-series

---

Operation 12: daily - Daily Event Counts

Show daily event counts over time using JQL.

Parameters

• Event name: Optional filter (use -e flag)
• Date range: --from and --to flags

Execute

All events:

!$(mp query jql - --from <from-date> --to <to-date> << 'EOF'
function main() {
  return Events({
    from_date: params.from_date,
    to_date: params.to_date
  })
  .groupBy(['day'], mixpanel.reducer.count())
  .sortAsc('key.0');
}
EOF
)

Specific event:

!$(mp query jql - --from <from-date> --to <to-date> << 'EOF'
function main() {
  return Events({
    from_date: params.from_date,
    to_date: params.to_date,
    event_selectors: [{event: '<event-name>'}]
  })
  .groupBy(['day'], mixpanel.reducer.count())
  .sortAsc('key.0');
}
EOF
)

Output: Daily counts showing trends and patterns.

Chain suggestions:

• See detailed breakdown → Run /mp-query segmentation with property breakdown
• Fetch period data → Run /mp-fetch for date range with high activity
• Retention analysis → Run /mp-retention to analyze user retention

---

Next Steps

After inspecting your data:

• Discovered interesting events? → Run /mp-fetch to get them locally
• Ready to analyze? → Run /mp-query to execute SQL or JQL queries
• Need funnel analysis? → Run /mp-funnel for conversion analysis
• Need retention analysis? → Run /mp-retention for user retention curves
• Want a report? → Run /mp-report to generate comprehensive analysis

Troubleshooting

No Credentials Configured

If mp auth test fails:

• Run /mp-auth to setup credentials first

No Local Tables

If mp inspect tables shows no tables:

• Run /mp-fetch to fetch data first
• Or use live discovery operations (events, properties, funnels, cohorts)

Event/Property Not Found

If inspection fails with "not found":

• Verify event name is correct (case-sensitive)
• Check property exists with /mp-inspect properties -e <event>
• Use /mp-inspect events to see all available events

JQL Operations Slow

If distribution/daily operations timeout:

• Reduce date range (Mixpanel API limit: 60 days recommended)
• Use local SQL queries after fetching data with /mp-fetch
• Check Mixpanel project size (large projects may be slower)