You're supporting a data pipeline that feeds customer-facing workflows, and a customer reports behavior that looks wrong in the product. Before treating it as an application bug, you want to determine whether the issue comes from incorrect, missing, duplicated, or delayed data moving through the pipeline.
What would you do if you suspected a customer issue was caused by bad data rather than a product bug?
Raw source payloads for the affected customerIngestion timestamps versus event timestampsDuplicate keys such as external_record_idSchema drift, null spikes, and rejected rowsDifferences between raw, transformed, and served tables