Talk about your next project
Dagster case study

The lesson

Write a case study for technical buyers as an engineering decision log: the old setup, the specific pain list, the tools you were compared against and why, then the migration details.

The move

Under the title, a jump list runs from "About Catalyst Cooperative" through "The early days of the PUDL project", "The limits of a DIY Python approach", "The right orchestrator: imperative or declarative?", "Migrating to Dagster" and "Building with Dagster". The story opens on the mission of the PUDL Project, then describes the original Python, pandas and nightly cron setup with a "Pre-Dagster architecture" diagram. A bulleted list of six challenges follows, then the evaluation of Airflow, Prefect and Dagster with a link to the team's GitHub issue. The migration section starts in September 2022 and covers partitioned EPA data, Pydantic config mapping and help from the Dagster community.

Why it works

  • Our readThe before diagram and the six bulleted pains let a data engineer check whether their own stack has the same problems before reading about the fix.
  • Our readNaming Airflow and Prefect, with the customer's own reasons for passing on each, answers the comparison question a technical buyer is already asking.
  • Our readAdmitting that mapping Pydantic classes "required some effort" and that partitions "presented challenges" makes the migration account more believable to engineers.

What not to copy

Don't copy the PUDL energy-data mission, the named agencies (EIA, FERC, EPA) or the Prefect comparison quote, which reads differently now that the banner says Dagster is part of Prefect. Keep the jump list, the before-architecture diagram, the bulleted pain list, the named-alternatives evaluation in the customer's words, and the honest migration section.

More from Dagster

More case studies

Steal it for your brand.

Tell us what you want to make. We’ll talk through how to put these ideas to work for your brand.