Keyword research for ecommerce (not blog-style keyword research)
Most keyword research advice was written for content sites: find topics, check volumes, write posts. Ecommerce keyword research is a different job. You're not hunting for things to write about - you're mapping how demand is phrased onto the catalog you sell, and finding the mismatches. The mismatches are the opportunity.
Start from your own data, not a tool
Generic tools tell you what the world searches. Your own data tells you what your demand looks like - and it's better input. Three sources you already have from Phase 3: GSC queries (demand you're already visible for - filter to non-brand terms, sort by impressions, and you have a demand map with your current position on it), site search (demand from people already on your store, in their exact words), and Merchant Center / Shopping query data if you run ads (paid queries that convert are organic targets validated with money).
External tools (Keyword Planner, Ahrefs, Semrush, whichever) then do what your data can't: sizing demand you don't yet touch, and showing the phrasing variants around each term.
Intent decides the page type - this is the actual skill
Every query implies a page type, and matching them is most of ecommerce keyword strategy. Category intent ("linen shirts women") → collection page. Product intent ("anine linen shirt navy") → product page. Informational intent ("is linen business casual") → content. Comparison intent ("linen vs cotton shirts summer") → content, close to purchase. The test is always: search the term, look at what Google ranks. If page one is all category pages, a blog post won't win it - and vice versa. Fighting the intent is the classic wasted effort.
Map, then find the gaps
The working artifact is a simple table: query family → intent → the page that should own it → does that page exist and is it any good? Three gap types fall out. Missing pages: real demand, no matching collection (create it - the taxonomy lesson's move). Wrong pages: a product page straining to rank a category term (build the collection, let the product rank its own name). Cannibalization: two pages competing for one family (consolidate).
Volume tie-breaks priority, but modifier patterns are the underrated read: recurring modifiers in your query data - a material, a fit, "oversized," "petite" - are customer taxonomy, and each consistent one is a subcollection candidate. That's the loop closing: keyword research isn't a separate SEO ritual. It's demand data feeding the same catalog structure everything else in this academy runs on.