What the documents look like
Marketplace advertising adjusts bids against traffic and conversion in near real time, so the operating rhythm differs from platform promotion sign-up, which turns mainly on campaign review. The working materials for this category are the delivery plan, keyword report, audience package description and conversion attribution table, together with store transaction data. In practice these are read daily, and on some days hourly, by delivery specialists, brand marketing teams and agency media buyers.
For copy drafting, these documents sit at the front of the chain. The delivery plan fixes the region, category and date range. The keyword report supplies observed click and conversion figures for named terms over a named period. The audience package description states the permitted scope of that package. The attribution table links outcomes back to delivery. Copy that departs from these materials removes the traceable starting point for later bid adjustment and attribution checks, and the visible consequences are wasted spend, distorted return on investment, and budget cuts to the campaign.
Constraints at the workflow routing step
The form input node is a user interaction node. When it triggers, the conversation enters an interaction state, the workflow state is recorded, and execution continues after the user completes the interaction. When the user fills the required information and submits, the node collects the submitted form values and passes them to later nodes. The user selection node is also a user interaction node with the same interaction-state behaviour. When the user makes a choice, the node evaluates that choice and executes the "yes" branch.
Both nodes can be added repeatedly and both trigger execution. Neither node performs business verification, approval, database writes or external synchronisation on its own. Publishing a finished copy line, writing an order, backfilling a document or synchronising an external state requires a separate interface or a later process.
Hypothetical paper exercise sample. The source material consists of three sentences. First: "This delivery plan covers the summer sunscreen category in East China, with a delivery period from 1 June 2025 to 30 June 2025." Second: "The keyword report shows that for 'sunscreen sensitive skin' between 1 June and 15 June the click-through rate was 2.1% and the conversion rate was 0.8%." Third: "The audience package description states that this package is used only for the above delivery plan and does not mark it as extendable to other categories." The draft line to be assessed is: "During the East China summer delivery period, this product performed outstandingly among sensitive-skin audiences and is suitable for long-term priority promotion." The comparison checks the subject, period and field of the draft line against the source sentences. The material gives click-through and conversion figures only. It gives no benchmark against which "outstandingly" could be measured, and no authorisation for "long-term priority promotion."
Configuration and implementation checks
| Configuration item | Setting or check relevant to this page | Relation to the sample |
|---|---|---|
| Form input node | Enters interaction state on trigger and collects required information | Carries submission of the delivery plan and keyword report material |
| User selection node | Executes the "yes" branch after the user makes a choice | Used to confirm whether the draft line proceeds to the next step |
| Interaction state check | State recording is behaviour of the interaction nodes | Verifies that state continues after the sample is submitted |
| Output handover check | Collected information is passed to later retrieval, judgement or interface nodes | Compares the draft line against the material field by field |
| External action split | Order writes, document backfill and external sync connect separately | Publication after copy sign-off sits outside these nodes |
Three common errors
- Writing "outstanding conversion performance" directly from a conversion rate of 0.8%. The material provides no benchmark value for a comparable category, so the comparison object is absent and the statement remains unsupported.
- Ignoring the stated scope in the audience package description and using a package limited to this delivery plan in copy for another category. The material marks the package as not extendable, so cross-category use exceeds the stated boundary.
- Treating the "yes" branch of the user selection node as an approval. The node evaluates the choice and executes a branch; approval and publication require a separate business interface or a later process.
How to verify the setup
- Input the three source sentences and the draft line, then compare subject, period and field. If the draft line contains a comparative conclusion that the material does not state, record it as unsupported and log the gap. The condition is met when every element of the draft line traces to a stated element of the material.
- Input the sentence "this package is used only for the above delivery plan" from the audience package description, then compare whether the draft line applies that package to another category. If cross-category use appears, record it as beyond the authorised scope. The condition is met when the draft line keeps the package within the stated plan.
- Input the selection result from the user selection node, then compare branch execution against the output of later nodes. If the result is "no" while a later node still carries the draft line forward, record a branch connection error and check each connection in turn. The condition is met when the executed branch and the downstream output agree.
A relevance score is not a category, entity or permission filter, and retrieval does not perform authorisation. Documented retrieval limits use their documented unit, and same-model retrieval association applies only where the documentation states it. Empty-reference output and context limits keep their documented conditions. Where a test would need a number, model, rule or identifier that the supplied documents do not provide, the field remains unknown and the gap is recorded rather than filled.