Context

In analytics work at Google, many datasets are not at the same grain. For example, a Google Ads campaign can map to multiple labels, and a conversion can be associated with multiple touchpoints. If you join these tables directly, your counts and sums can inflate.

Core question

Explain how you would handle a many-to-many join in SQL without inflating metrics. Your answer should cover:

1. How to recognize that a join is many-to-many
2. Why direct joins can duplicate rows and distort aggregates
3. Practical techniques to prevent double counting
4. How you would validate that the final metrics are correct

Scope guidance

The interviewer is not looking for obscure syntax. They want a clear explanation of table grain, deduplication strategy, and when to pre-aggregate or use an intermediate mapping table before joining. Use simple PostgreSQL examples where helpful, and frame your answer around common analytics reporting problems.