Every figure in ENTEL is computed from archived source documents. No figure is entered by hand, every number opens the document behind it, and anything ENTEL judged rather than read is published as a rule you can check.
The corpus
Twenty seven sources run on their own cadences, from every three hours to monthly, and every document they return is hashed and archived on arrival. When a page changes under us the old version is still there, which is the only way to answer "what did this say when you told me that." The archive holds over 1,500 documents and grows every day.
Eighteen further sources are disabled, each with a recorded reason, and we keep that list. A source that was probed and failed is itself a finding. An empty feed and a broken feed look identical unless somebody recorded which one it was.
Every record that arrives carries one of five evidence classes, A to E. The classes and what each one takes are set out on the about page.
Judgment
A procurement lane opens on a project when its lifecycle type and current stage put it inside a window. ENTEL works that out rather than reading it off a filing, so the rule it used is published rather than described. Here is one in full.
Derived New build, stage: site secured
Site is permitted but no construction application is filed. Site characterisation and early works scope firms up first, which is why civil moves ahead of hardware here.
Disagree with the rule and you can say exactly where, because the rule is a published object you can open.
What this does not yet cover
Everything above describes how project movement is graded. The federal lane, meaning DOE and M&O recompete forecasts and the buying that follows them, is a different evidence object: a solicitation is not a project-stage event and rendering it as one would be wrong. It runs on its own ladder, and that ladder is not published yet.
The federal ladder ships with the beta on October 1st. Ask us for it then.
Check any count against its source.
The live demo renders the source table and the grading table off the corpus, and every source with a public index links out to it, so any count on any page can be checked against the source that produced it.