All stacks · Ingestion & ELT
Open-source ELT connectors into warehouses and lakes.
Airbyte🟡IntermediateThe ELT decision as it actually plays out: MAR pricing versus infrastructure you operate, what happens when a connector breaks, and the sources where I still write a custom extractor instead.
Airbyte🟢BeginnerGreen sync, empty destination, raw JSON versus typed silver, self-host versus Cloud, and why Airbyte does not replace dbt.
Airbyte is the open-source ELT connector platform: extract from SaaS and databases, load into warehouses and lakes, optionally transform later in dbt. You operate the workers; you do not write a new Python extractor for every vendor.
The production work is connector reliability, schema change, and not treating “the sync succeeded” as “the grain is correct.”
Fivetran if you want a vendor SLA and usage-based pricing you can explain to finance. Airbyte if you want open connectors, self-host, or sources Fivetran prices badly. Many teams run Airbyte for a few sources and Fivetran for the rest.
No. Load raw, then model. Normalization helpers are not a mart layer. Grain and tests still belong in dbt or SQLMesh.
You loaded raw without a typed silver layer. That is correct for day one and wrong as the serving layer. Keep raw, parse in transform.
