Starter content is an editable start; samples stay out of the record

What it is

Putting something into an empty product at first use is two different jobs. Sample data exists so people can see the filled shape; it is not their record, should leave as a batch, and should stay out of search, counts, and export. Starter content is an editable beginning: a seed project, a draft, a templated document they will keep changing, a real record from the first moment. Treating both as one pile of “demo data” makes people delete a start they should keep, or keep fake rows as finished work.

Why it happens

Display wants revocable scenery; a start wants a continuable object. If scenery is written as a real record, deletion is destruction and stats are polluted. If a start is written as scenery, one-click “clear samples” wipes the work they just began to edit. Identity must fork at write time: samples live in their own namespace; starters are ordinary objects whose origin is “template.” Which road to take depends on the task: dashboards and inboxes that must be understood before people load their own stuff fit samples; documents, boards, and automations whose first artifact needs a structured draft fit starters. Both can coexist, but they cannot share a clear button or an unmarked look in the list. Once a starter is edited, origin may still say “from template,” but identity is already the user’s—and must not be swept by “remove demo.”

Where it stops holding

A read-only industry demo account is scenery all the way: no starter, and no “become mine” unless an explicit copy. Legally required system objects (a default admin, an undeletable root folder) are neither sample nor starter; keep them out of any clear or “start editing” pack. The first item a user creates from blank is not a starter; do not stamp template origin on it after the fact.

Applying it

• Mark every auto-created object as “display only” or “editable start.” Forbid a third, vague “demo data.”
• Display-only objects stay out of true-record queries. Editable starts enter true records, but offer “switch to blank” before first save without a delete.
• Do not handle both kinds with one “clear samples” button.
• Verify by “remove the scenery” and by “change one word on the start, then search.” The first must leave no fake rows in counts; the second must find the edited word as a real record that “clear samples” cannot wipe.

Related

• Within the group: H2.08.2 Starter templates must look unlike user-created work · H2.08.3 People often read starter content as personalized output · H2.08.4 Starter content must resemble real work or it fails as a model
• Adjacent: H2.06 Sample Data · H2.02 Empty-state Guidance · H8.01 CRUD operations
• Search terms: starter content · seed project · sample vs starter