Farm Data Harvest: Understanding ETL vs ELT. Welcome to Green Valley Farms, data farmer. Farmer Ted is drowning in data — field sensors, daily harvest logs, and weather feeds — and wants to turn it into decisions: which fields yield most, where irrigation is wasted, when to plant. To get there, you need a data pipeline, and the first decision is the classic one: ETL or ELT?
This is an open-response design brief — no auto-grader. Lay out your pipeline and justify your approach in the editor, then reveal the worked solution to compare.
ETL vs ELT, in one breath
- ETL (Extract → Transform → Load): clean and reshape the data before it lands in the warehouse. Classic for on-prem, strict-schema, or PII-sensitive flows.
- ELT (Extract → Load → Transform): land the raw data first, then transform inside a cloud warehouse (often with dbt). Flexible when transformations evolve and storage is cheap.
The data to integrate
- Field sensors — soil moisture, temperature, streamed continuously.
- Harvest logs — yield per field, dropped as daily CSVs.
- Weather feed — pulled from an external API.
Your task
- Choose ETL or ELT for Green Valley Farms and justify it (think warehouse, volume, evolving transforms, governance).
- Sketch the pipeline — sources, how each is ingested, where raw data lands, and where transformation happens.
- Layer your transforms — staging → intermediate → marts — and name the business-facing models (e.g. yield per field).
- Call out data tests — freshness, uniqueness, volume — that keep the pipeline trustworthy.
How to work the project
Write your pipeline design and justification in the editor below. When you're ready, reveal the worked solution to compare. The in-page tutor can pressure-test your ETL-vs-ELT reasoning.