SQL for Product Managers
Speaking the Language of Data Teams
Learn these terms precisely, and be openly comfortable about everything else. It's the same credibility move as with an ML or engineering team.
Term | Plain English | Why you care |
|---|---|---|
Schema | The structure of a database: what tables exist, what columns each has | The map you need before writing any real query |
ETL / ELT | The process of moving data from source systems into a warehouse, transforming it along the way | Explains why the number in the warehouse might lag the live product by minutes or hours |
Data warehouse | A database built specifically for analysis, separate from the live product database | Where you'll actually run most of your queries, not against production |
Source of truth | The one table or system officially trusted for a given metric | Prevents the "why do these two dashboards disagree" conversation |
Grain | What one row in a table represents (one user, one order, one event) | Getting the grain wrong is the root cause of most fan-out and double-counting errors |
Materialized view | A query's result saved and refreshed on a schedule, instead of recalculated every time | Explains why a dashboard might not reflect the very latest data |
Data freshness | How current the data is compared to real time | The honest answer to "is this live" |
dbt model | A defined, version-controlled transformation that turns raw data into a clean, reusable table | The modern way most data teams build the tables you'll actually query |
Three sentences that earn real credibility fast:
"What's the grain of this table, one row equals what?"
"Is this the source of truth for this metric, or is there another version somewhere?"
"How fresh is this data, when did it last refresh?"
YOUR MOVE THIS WEEK
In your next conversation with a data engineer or analyst, ask about the grain of one table you use often. If you've never asked that question before, there's a real chance you've been assuming the wrong thing about what a row represents.