A product manager at a retail firm outlines a new AI capability: 'On a customer's order history screen, display three recommended items we believe they'll appreciate, accompanied by a brief justification explaining the personal relevance of each item.' An AI pipeline engineer needs to convert this high-level business concept into a precise input/output specification. Which of the following specifications MOST accurately reflects this requirement?
Answer & explanation
Correct answer: B
WHY B is correct: A good specification maps the business goal to concrete inputs and structured outputs. Personalization needs the customer's purchase history together with the catalog to choose from, and the request for a brief justification per item means the output must carry a natural-language reasoning for each of exactly three products, not just identifiers.
WHY NOT A: A global bestseller list ignores the personalization requirement: every customer would see the same three products whatever their history. Returning names alone also leaves out the explanation of why each product suits this customer.
WHY NOT C: A paragraph about the customer's tastes is not a recommendation. The product manager asked for three specific products; a taste profile may be a useful intermediate step in the pipeline, but it is not the final output described.
WHY NOT D: Demographic targeting is not personalization from the customer's own behavior: it does not use what this customer actually bought. The output also omits the personal explanation the requirement calls for.
WHY NOT E: Raw cosine similarity scores are an intermediate computation, not something to show on a customer's screen. The requirement is a named product with a readable explanation, and using a single recent purchase discards the wider purchase history needed for good personalization.