Categorize products in Google Sheets

Updated 28 September 2026 · Edoardo Panichi, maker of Hunch

=HUNCHPICK(text, options, question, as) reads a product title and returns the category it fits, chosen from a taxonomy you describe. Add "confidence" as the fourth argument and the same pair of formulas also returns how sure the model is, so the uncertain rows sort to the top for a person to check.

A catalog feed rarely has a category field worth trusting

Products arrive from several vendors, each with its own naming habit and its own idea of what counts as a category, or with no category field at all, only a title. Mapping every title to one taxonomy by hand works for a hundred products and stops being realistic well before ten thousand.

The taxonomy

Six categories, each described by what it holds, and other for the titles that do not belong to any of them. A general home goods catalog is shown here; a narrower store needs a narrower list, ten specific categories read better than six broad ones that argue with each other. Full syntax is on the HUNCHPICK reference.

Real outputs

Nine product titles copied straight from a catalog feed, capitalization and all.

None of the nine titles land in other, including the replacement water-pitcher filter, which reads as home and kitchen despite being a spare part rather than the appliance itself. The bathroom storage cabinet is the one worth a second look: it lands on furniture, not bedding and bath, because the option list defines furniture by object type, chairs, tables, storage, shelving, rather than by which room the object sits in. That is the option list working as written, though it is worth rereading if your team expects anything with bathroom in the title to land in the bath category instead.

Confidence for review

Same title, same options, the fourth argument changed from nothing to "confidence". It returns a probability for the option that got picked, not the option itself, run it as a second column next to the category so you keep both.

Every row in this set scores 0.84 or higher, which is not what a hand-picked set of cross-category items was expected to produce. The camping multi-tool, which could plausibly be read as outdoor gear or an electronic gadget, is the least confident of the six at 0.84, and even the kids' table and chair set comes back at a full 1.00 despite the taxonomy having no separate bucket for children's furniture. The lesson is not that these titles were easy to guess. A 0.6 cutoff, picked before looking at real data, would have returned nothing to review from this batch. Sort your own column by confidence first and read where the scores start dropping before you pick a number.

What to do with a low-confidence row

Lowest confidence first. A single low score is worth a glance; several low scores clustered around the same kind of product usually mean the taxonomy needs one more bucket, not that the model guessed badly.

Once the category column is settled, a plain count per category is the merchandising number, how many SKUs sit in each part of the catalog, without opening a single product page.

Limits

• Only the title text is read, not the product image, the price, or an existing category ID already in your feed.
• The option list can hold up to 255 entries but reads clearly under about 30. For a catalog with real subcategories, pick the top-level category first, then run a second HUNCHPICK scoped to each top-level group for the subcategory.
• A bare SKU code with no descriptive words returns a low-confidence guess or lands in other. Add the product title or description column if a feed only has codes today.
• It does not know your existing taxonomy unless you write it into the option list. If a vendor feed already carries a category string, keep it as a separate column and compare the two rather than discarding one.

Copy the template to try this on your own catalog. The expense categorization guide covers the same HUNCHPICK pattern on transaction descriptions instead of product titles, and the general classification guide covers the pattern on any list.

Questions

Reference: HUNCHPICK

More guides

• Categorize support tickets in a spreadsheet Ticket triage
• Code open-ended survey responses in Google Sheets Survey coding
• Classify text into categories with a Google Sheets formula Classification
• Tag customer reviews by topic in Google Sheets Review tagging
• AI formulas for Google Sheets: what returns a number, what returns text AI formulas compared