**Statistics Update Cost:**

Small table (1K rows):

- ANALYZE time: 10ms

- Impact: negligible

- Update frequency: after every change

Medium table (1M rows):

- ANALYZE time: 500ms

- Impact: low

- Update frequency: after 10K+ row changes

Large table (100M rows):

- ANALYZE time: 30 seconds

- Impact: moderate

- Update frequency: after 1M+ row changes or daily

Very large table (1B+ rows):

- ANALYZE time: 5+ minutes

- Impact: high

- Update frequency: weekly or after major ETL

- Consider: ANALYZE with sampling

**Sampling Trade-offs:**

Full scan (ANALYZE with FULLSCAN):

- Pros: Most accurate statistics

- Cons: Slow on large tables

- Use: Critical tables, after major changes

Sampling (default):

- Pros: Fast, good enough for most cases

- Cons: May miss rare values

- Use: Regular maintenance

PostgreSQL default: 300 * default_statistics_target pages

- default_statistics_target = 100

- Samples 30,000 pages (~240MB)

**Impact of Stale Statistics:**

Scenario: 10-table join query

- Each table has 10% cardinality error

- Compounding error: 1.1^10 = 2.6x total error

- Result: Wrong join order, 10x slower

Scenario: Memory allocation

- Estimate: 1M rows → allocate 100MB

- Actual: 10M rows → need 1GB

- Result: Memory spill to disk, 100x slower

**When to Update Statistics:**