**Impact of Estimation Errors:**

**Scenario 1: 10x Underestimate**

Estimate: 10K rows → choose nested loop

Actual: 100K rows

Nested loop cost:

- Estimated: 10K × 100 = 1M operations

- Actual: 100K × 100 = 10M operations

- 10x slower than expected

Should have chosen hash join:

- Cost: 100K + 100K = 200K operations

- 50x faster than nested loop

**Scenario 2: 100x Overestimate**

Estimate: 10M rows → allocate 1GB memory

Actual: 100K rows → only need 10MB

Impact:

- Wasted 990MB memory

- Other queries starved for memory

- May cause memory spills in other queries

**Scenario 3: Compounding Errors in 5-Table Join**

Each table has 2x estimation error:

- Total error: 2^5 = 32x

- Estimated: 100K rows

- Actual: 3.2M rows

- Wrong join order chosen

- Query 100x slower

**When to Investigate:**

- Estimation error >10x: Fix immediately

- Estimation error 2-10x: Investigate if query is slow

- Estimation error <2x: Acceptable

**Cost of Statistics Updates:**

Small table (10K rows):

- ANALYZE time: 5ms

- Update frequency: After every bulk change

Medium table (1M rows):

- ANALYZE time: 500ms

- Update frequency: After 100K+ row changes

Large table (100M rows):

- ANALYZE time: 30 seconds

- Update frequency: Daily or after 10M+ row changes

Very large table (10B rows):