Category keywords are the easy layer. "Running shoes" and "duvet cover" are obvious, expensive, and shared with everyone who sells the category. The queries that name a model, a size, a fit problem, or a job-to-be-done are where a catalog of 1,000 SKUs can be specific. They are also the queries a generic brief never lists, because the person writing the brief does not have the variant table open.

Start from the product, then write the way a buyer who already knows something would search. Verify every line afterward. Discovery without a volume check is how invented phrases enter an ad group.

Four lists to draft before you open Planner

Work one collection at a time. For each, write four lists and keep them labeled so they do not collapse into one keyword dump:

  • Product and model: the name on the page, the manufacturer code, the capacity, the dimensions buyers quote.
  • Size and variant: the size systems your customers use, including the local one if you are selling abroad. A US size and an EU size are different queries.
  • Problem and use: the situation the product is bought for, in everyday words. "Sheets for night sweats" is a different query from the fabric name.
  • Competitors and substitutes: other brands and the generic alternative, parked for a later decision about whether you want that campaign at all.

The scan of what you actually sell has to come first. If the model is out of stock in the sizes people buy, discovering its keyword is not a reason to advertise it. Read the live catalog the way the product and collection scan describes, and only draft queries for what you can ship.

Everyday language beats the spec sheet

Buyers mix the spec with the shelf name. They type the problem when they do not know the spec yet. A coffee store will see "espresso machine" and also the exact portafilter diameter, and also "gift for someone who makes coffee." Those are different pages and different promises. Gift queries in particular need attributes a non-expert can use, which is the point of gift-intent product data.

Write the phrases a shopper would type, not the phrase your merchandiser prefers. Then let Planner delete the ones that are only internal vocabulary. Local retailers are a better source of everyday nouns than a translation of your domestic title. Use both, and measure them apart.

Leave each list in its own bucket

Do not merge the four lists into one ad group called "keywords." Model queries want the product page. Problem queries may want a collection or a guide that still sells. Competitor queries want a decision, a separate budget, and no competitor name in the headline. The intent split is where these lists go after they survive the volume check.

A useful discovery set for a single collection is often a few dozen lines, not a few hundred. You are looking for phrases with a page behind them. You are not trying to name every adjacent noun.

How Kopa drafts the phrases

In Kopa, Smart Campaigns drafts everyday buyer phrases, exact product and model names, sizes, use-case queries, and competitor terms from the store, then sends them through Keyword Planner before they can become keywords. The labels stay intact so a problem query is not mixed into a model ad group by accident.

What good looks like

  • Every priority collection has model, size, problem, and competitor lists written separately.
  • Phrases use buyer language and local shelf names, not only the internal title.
  • Out-of-stock models do not get a keyword just because the name is famous.
  • Each surviving phrase has a destination that answers that phrase.
  • Nothing goes live on the draft alone. Volume and intent filters still run.