Field
How we pick sources for the system
Notes on what makes a source worth integrating and what makes it noise.
Member-relevance first.
The first question on any new source is whether it affects an actual member. We don't add sources because they exist. We add them because a client is exposed to what's in them and we currently can't see it.
Update cadence and structure.
A source needs to update on a knowable schedule and in a structured format we can parse. A blog without RSS is a worse source than a regulator's PDF on a known cadence, even if the blog is more interesting.
Signal-to-noise.
We watch the candidate source for two to four weeks before integrating. If it generates more noise than signal in that window, we don't integrate. The bar is high.