Semantic Search for Gifting: Recipient, Occasion, and Budget Queries

Gift queries describe a person, not a product

"Gift for my dad who likes fishing," "wedding present under a hundred," "something for a ten-year-old obsessed with space." A gift shopper is describing a recipient, an occasion, and a budget, and often has no product in mind at all. Semantic search for a gift finder has to turn that description into a ranked set of options that fit the person.

Intent extraction is the whole feature

The search has to pull three things from the query: who the recipient is — relationship, age, interests — what the occasion is, and what the budget ceiling is. Once those are known, ranking can lead with things that suit a fishing-loving dad in that price range, drawn from wherever in the catalogue they live.

Ranking crosses categories

Unlike a normal product search that stays in one department, a good gift result for "someone who likes fishing" might be gear, a book, an experience, or a novelty. The retrieval has to range across the catalogue and rank by fit to the recipient, the same reranking-over-candidates idea as our semantic search project, with "fit to person" as the thing being scored.

Budget is a hard constraint

"Under fifty" is not a preference for a gift shopper, it is the brief. Price has to bound the set before ranking. A perfect-seeming gift at double the budget is a failed result, and it is a common way for gift search to feel out of touch.

A guided flow for thin queries

Many gift queries are vague — "present for my mum." Rather than returning a generic bestseller list, the search can ask one or two clarifying questions — interests, occasion, budget — and use the answers to narrow. A short guided step beats a weak result here.

Gift services belong in the result

Wrapping, a gift message, a scheduled delivery date, and hiding the price from the recipient are part of what makes a result useful for gifting. Surfacing whether an item supports these is as relevant as the product itself.

Where the data limits it

Interest matching depends on products being tagged with the themes and activities they suit. If the catalogue only has category and brand, "for a keen gardener" will fall back to anything with "garden" in the title. The early gift queries show which interest tags are missing.

A worked example

A shopper searches "gift for my dad who likes fishing, under 60." The finder extracts the recipient's interest and the budget ceiling, ranks across categories — a tackle organiser, a well-reviewed book on fly fishing, a flask — and leads with the options that best fit an interested hobbyist in that price range. A second shopper types only "present for my mum." The finder asks two quick questions — her interests and the occasion — then narrows from the answers rather than returning a generic bestseller list.

Measuring a gift finder

The metric that matters is completion, not clicks. Track how often a gift session ends in a purchase, how often the guided questions are answered versus abandoned, and the return rate on gift orders. A high abandon rate on the questions means the flow is asking too much before showing anything useful.

Where this stops being right

• Highly personal taste — a gift for a partner, something sentimental — is where human judgment still wins; the finder helps with the harder-to-shop-for cases.
• Experience and bespoke gifts may sit outside the product catalogue entirely and need their own handling.
• A small catalogue with a clear gifting angle may do well with curated occasion and recipient lists rather than a semantic pipeline.

FAQ

What does a semantic gift finder do differently? It extracts recipient, occasion, and budget from a natural query and ranks options from across the whole catalogue by how well they fit the person, instead of matching keywords.

How is budget handled? As a hard ceiling. For a gift shopper the budget is the brief, so out-of-budget items are excluded before ranking rather than shown as aspirational options.

What happens with a vague query like "gift for my mum"? The finder asks one or two quick questions about interests, occasion, and budget, then narrows — a short guided step gives a far better result than a generic list.

ISTRALLEN builds gift-finder search that reads recipient and occasion and ranks across categories by fit — see AI for Retail.