ETL and ELT

ETL and ELT are the plumbing of the data economy: moving data from where it is produced into where it is analysed — extracting, transforming, loading, in either order. Unglamorous, ubiquitous, and the place where a large share of data quality and data protection problems quietly originate.

What It Is

ETL (Extract, Transform, Load) pulls data from source systems, reshapes it (cleaning, joining, standardising), and writes it into an analytical store — classically a data warehouse. ELT inverts the last two steps: raw data lands in the destination first (increasingly a cloud warehouse or lake with compute to spare) and transformation happens there, keeping the original available and making transformations re-runnable as logic changes. The choice is architectural rather than doctrinal; both live in pipelines that run on schedules or events, and both are where source data becomes report-ready.

Why It Matters to You

Executives encounter ETL as the explanation behind three chronic phenomena. 'The numbers don't match': divergent figures between operational and analytical systems are usually transformation logic — the pipeline silently dropped, duplicated or re-defined records. 'Stale dashboards': pipeline schedules and failures govern how current the truth is. And 'why is our data protection exposure in the warehouse?': every extract copies personal data into a new location with its own obligations — pipelines are where data-protection scope quietly multiplies, which is why regulators treat data maps, not intentions, as the fact of record.

When You'll Meet It

In data platform reviews, dashboard disputes, and data-protection audits. When two reports disagree, the answer is almost always in the transformation layer — ask for the lineage before the argument.