When a rule needs a tool, not a template

This blog has already listed the four tests a document has to clear before it earns a place on the launch shelf — rebuilt often, expensive when wrong, mapped to a primary source, maintainable. None of those four tests says anything about format. They don't decide whether the thing that clears them ships as a fillable template or as one of the shelf's own browser-only tools, and that decision turns out to run on its own separate test — one this blog's own Craft posts keep accidentally demonstrating, one branching regulatory fork at a time.

A template is a list; a fork is a question

A template works because the thing it encodes holds still: a document-control SOP names the same revision-history fields for every user, a CAPA log needs the same columns whether the finding came from a complaint or an audit. The moment the underlying regulation branches on a fact specific to the device in front of you — is this device significant risk, does this business count as small, is this accessory's own risk lower than its parent's — a static template stops being able to show the logic at all. It can list the branches in prose, the way a checklist item might read “if X, do A; if not X, do B,” but it can't walk a specific reader through their own answer. That's a different job, and it's the job an interactive tool is built for.

The shelf's own Craft posts are a map of where the forks are

21 CFR 812.2(b)'s significant-risk determination forks a study onto one of two tracks before an application ever gets drafted. Section 737(13)'s affiliate-control test forks a device sponsor's user-fee bill depending on gross receipts a spreadsheet, not a checklist, is built to add up correctly. An accessory's own classification under Section 513(b) forks on the accessory's own risk rather than its parent's. Every one of those posts exists because a team kept reaching for a checklist to answer a question a checklist can't actually resolve — and every one of them is a candidate for the shelf's interactive side rather than its template side, for exactly that reason.

The test runs on the logic, not the topic

Two documents can share a regulation and still land on opposite sides of the split. A document-control SOP under Clause 4.2.4 is the same fixed set of controls for every quality system, so it's a template. A QMSR transition record — which of the old DMR, DHF, and DHR content maps to which incorporated clause — is a template too, because the mapping itself doesn't change from one reader to the next; the same crosswalk applies to every file built against the old Quality System Regulation. But a study's own significant-risk call depends on facts about that specific device that no two sponsors share, run against a four-part test with a catchall clause at the end — and that's a tool's job, because the branches, not the topic, decide the format.

A tool still owes the same sourcing discipline a template does

Nothing about running client-side loosens the mapped-to-a-primary-source test this shelf holds every template to. A classification decider that asks about significant risk still has to cite 812.3(m)'s own four tests, the same way a template would, and it goes stale exactly the way a template does the moment a cited section moves. This blog has already covered the one place the two formats genuinely diverge: what happens to a tool's output once it gets promoted from a working draft into the literal record a predicate rule requires. Right up to that point, the discipline is identical on both sides of the split; only the interface changes.

Where the line actually falls

The distinction isn't a preference for one format over the other — the shelf ships both, on purpose, and neither is the fallback version of the other. It's a test worth running before either one gets built: does the regulation ask the same question of everyone, or does it ask a different question depending on the answer the reader gives it first? A resource built on the wrong side of that line either flattens a branch that mattered or dresses up a fixed list as a decision it never had to make. If you've found a branch our own tools should be asking about and aren't yet, the shelf's interactive side takes that correction the same way a template submission does.