Handling Missing Values in Production

Context

Missing values are common in operational data and can silently break downstream reporting, Tesla fleet telemetry analysis, or KPI dashboards if they are handled inconsistently. Interviewers want to see that you can combine SQL techniques with production judgment.

Core question

You are asked to explain how you would handle a dataset with missing values in a production PostgreSQL environment. Describe how you would distinguish true NULLs from invalid placeholders, how you would profile missingness with SQL, and how you would decide between filtering, imputing, defaulting, or escalating data quality issues. You should also explain how you would make the logic reproducible for downstream consumers.

Scope guidance

Keep the answer grounded in SQL & data manipulation rather than generic ML preprocessing. A strong answer should cover practical PostgreSQL patterns such as CASE WHEN, COALESCE, validation rules, and using CTEs to separate raw, cleaned, and audited outputs.