Recipes
Five questions, start to finish. Each one gives the order the tools should be called in, what to check at each step, and where the answer actually comes from.
They are written the way the question gets asked — not the way the tools are named — because that is how you will ask it.
| Recipe | The question behind it |
|---|---|
| Set up a campaign | An offer described in a sentence should exist |
| Why is my gift not showing? | A shopper should be seeing an offer and is not |
| Did gifts lift my AOV? | Are the offers actually earning their keep |
| What text do shoppers see? | Which wording is live, and in which language |
| Can this product be gifted? | A specific gift is not landing in the cart |
The shape they share
- Start with the tool that reaches a conclusion, not the one that returns the most data.
get_setup_healthandcheck_gift_productboth hand back a verdict with reasons; reading those first usually ends the investigation in one call. - Narrow before you widen. One campaign, one widget type, one language. Results are capped at about 16,000 characters, and a broad call that gets truncated is worse than a narrow one that does not.
- Read the preview before confirming. The four read recipes cannot break anything. The first one writes, and it writes only when handed a token from a preview you were shown — that preview is the review step, not a formality.
- Take the store's word for it. Timezone, currency, and stock verdicts all come back in the result. Reading them beats assuming them.
- Say what was not checked. If a step was skipped or a figure came back partial, that belongs in the answer.