Search

Search works as soon as your store is online: there is nothing to set up. Tuppence indexes each product online from its title, brand, type, attributes, collections and description, and keeps the index current when you change a product, its price or its stock.

How a search is answered

• Words and meaning. A search matches the words in your products and what the shopper means, so “gift for a runner” finds running gear without those words in its title.
• Filters. Longer searches, or ones with a price, size, colour or who it is for, become filters from your products’ own attributes. Anything that isn’t one of your values is dropped.
• Never an empty page. When the filters leave nothing, Tuppence relaxes them one at a time and says which. When nothing matches at all, the shopper sees your best sellers and can ask the shop.
• Did you mean. A misspelt word is matched to the words in your catalogue.

Shoppers on your site use /search?q=, which works without JavaScript. Your own front end calls the Storefront API:

curl -sS -X POST "https://<your store>/storefront/v1/search" \
  -H "Content-Type: application/json" \
  -d '{"query": "blue mug", "filters": {"price_max": 2000}, "limit": 10}'

filters takes category, price_min and price_max (minor units) and attributes ({"colour": "blue"}). Prices are the all-in prices your store shows.

What shoppers search for

Store › Overview shows the ten words shoppers searched for most and the ten most searched that found nothing, over the last 28 days. Those are products to add, or words to put in a product’s search fields. Email addresses and phone numbers are taken out of a search before it is kept.

Over MCP and the assistant

Tool	Scope	Does
get_store_search_insights	sites:read	The most searched words and those that found nothing, over days (28, up to 90)

Next

• Products
• Subscribe and save