Use when recommendation slots are filled by an algorithm and headed with nothing a shopper can act on.
You are writing the copy around product recommendation slots. You are not choosing the recommendation algorithm.
Slots and where they appear: {{SLOT_LOCATIONS}} (for example product page, basket, order confirmation)
Anchor products and what is currently shown alongside them: {{PAIRING_DATA}}
Why each pairing exists, where anyone knows: {{PAIRING_RATIONALE}}
Attach rate per slot: {{ATTACH_RATES}}
For each slot in {{SLOT_LOCATIONS}} produce:
- A heading under 6 words naming the relationship, not the mechanism. "Needed to fit this" beats "You may also like".
- A supporting line under 20 words.
- The per item line shown beside each recommendation, under 12 words, written as a reason to add it.
- The condition under which the slot should not render at all.
Then SUPPRESS: pairings in {{PAIRING_DATA}} that should not be shown, with the reason.
Rules:
- Every reason must come from {{PAIRING_RATIONALE}}. Where none exists, write a plain listing heading and flag the pairing for review rather than inventing a relationship.
- In a basket or confirmation slot, recommend additions only. Never a replacement for what the shopper already chose.
- Rank the slots by {{ATTACH_RATES}} and say which to rewrite first.
- No em dashes. No "customers also bought" unless that is literally what the data is.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When should I change the copy rather than retune the algorithm?
When the slots are filled reasonably and headed with nothing a shopper can act on. Retuning is the expected answer and costs engineering time. Compatibility and consumables slots reliably earn their space; taste based ones are argued over, and the attach rate per slot is the only way to settle it for your catalogue.
What do I need to know about the pairings?
Where the slots appear, the anchor products with what currently shows beside them, why each pairing exists where anyone knows, and attach rate per slot. The rationale decides the output: where none exists the prompt writes a plain listing heading and flags the pairing rather than inventing a relationship for it.
What comes back per slot?
Per slot a heading naming the relationship, a supporting line, the per item reason and the condition under which the slot should not render, then a suppress list. The no-render condition matters most on the order confirmation, where a badly matched slot reads as a fresh sales pitch after payment.
Which part of the output gets skipped?
Leaving the suppress list unimplemented, particularly for accessories already in the basket. It is the commonest complaint about these slots and costs nothing to fix. In basket and confirmation slots, check as well that nothing recommended is a replacement for the item the shopper has just chosen.