Golden Path

A golden path is the supported, opinionated route through a platform for a common task: creating a service, shipping a pipeline, publishing a dataset. It is a specific set of choices — this language, this pipeline, this deployment target — packaged so that following it is the fastest way to get the job done.

The idea answers a question every platform faces: how much variety to support. Supporting everything means supporting nothing well, since each combination needs its own documentation, testing, and on-call knowledge. Supporting exactly one option makes the platform an obstacle the first time a team has a genuine reason to differ.

A golden path resolves that by being a default rather than a mandate. It is the route that is documented, templated, monitored, and known to work, so choosing it costs nothing. Teams remain able to go elsewhere, and the platform simply does not promise the same support when they do.

That framing matters because it changes what a departure means. When paths are mandatory, going around the platform is a violation and therefore hidden, which is the worst outcome — the organization now has unsupported systems it does not know about. When the path is a default, a departure is a signal, and a recurring one is evidence the path is missing something.

Two properties keep one healthy in practice, and this part is a working judgment rather than a definition. It has to stay genuinely the easiest option, which means maintaining it as the underlying tools change rather than letting it decay into the oldest supported versions. And there should be a stated answer for what happens off the path — who supports it, what still applies, what does not — because the alternative is discovering the answer during an incident.

The same logic applies to data platforms, where the golden path is how a team gets a dataset ingested, governed, and published. It is discussed in context in Data Platforms and Developer Platforms.