What are the most common data pipeline architecture patterns?

The formalization of pipeline architecture patterns as strategic knowledge reflects a maturation we're experiencing in data practice. What dbt Labs is articulating here is that transformation logic—how we shape raw data into trustworthy assets—deserves the same architectural rigor we've historically applied to infrastructure. I've seen teams struggle precisely because they treated transformation as an afterthought, bolted onto whatever ingestion framework they inherited. The real win comes from recognizing that semantic layers and declarative transformation (dbt's sweet spot) reduce the tribal knowledge problem significantly. When your transformation logic lives in version-controlled, documented code rather than SQL scripts scattered across databases, your team actually ships faster and with fewer midnight incidents. The streaming versus batch tension remains real for most of us, but the pattern framework helps teams make that choice deliberately rather than defaulting to batch because it's easier. My recommendation: audit your current stack against these patterns. If you're mixing paradigms without explicit reasoning, you've found your next refactoring project.