Why E-commerce Managers need this
Your category tree says one thing and your crawl logs say another. Size, colour, price and sort filters multiply into a near-infinite set of URLs, and Google's own guidance warns that crawlers then have less time for new, useful pages, so a launch collection can sit undiscovered. Meanwhile the pages you do want ranked are often a grid of products with no copy, the same item sits under three category paths, and page 2 onward is canonicalized to page 1 or hidden behind a load more button. Before fixing any of that, decide which categories deserve to exist as indexable pages at all. Keyword Research starts from one seed such as a product type, pulls live autocomplete from Google, Bing, DuckDuckGo and YouTube, and returns a scored list with intent labels and topic clusters. For an e-commerce manager it works as a demand map: which product types shoppers actually name, which modifiers they add, and which phrases belong on a category page rather than as a filter combination. It reports relative indexes, not monthly search volume.
E-commerce Managers specific guidance
Seed with the product type, not the brand
Open the tool and type the plain product-type noun a shopper would use, for example trail running shoes or standing desk, then set Language and the Country that matches your store. Run Standard depth first; Deep adds the digits 0-9, which helps if you sell model-numbered goods. Within ten minutes, copy the full list and paste it next to your current category list, then mark every phrase that has no matching category URL. Questions and co-occurring modifiers show how shoppers narrow a search, which tells you whether a modifier such as waterproof deserves its own page or can stay a filter. The result does not show whether any of your pages already rank, and it cannot see your catalog, so a phrase with strong demand may already be covered by a page you forgot you had.
Read demand, difficulty and opportunity as relative
Sort the priority targets by opportunity, but treat the Demand Index as a relative score built from autocomplete weights and agreement across engines, never as monthly volume. Difficulty is a heuristic from phrase length, intent and demand; it does not look at who ranks, so a low number does not mean an easy category to win. Check the intent label on each row: commercial phrases suit category pages, while question phrases usually fit a buying guide. Intent labels come from word lists and can be wrong, so click through to the live search for your top five phrases and see what kind of page appears. Then write the page type beside each shortlisted phrase. None of this tells you commercial value or margin.
Turn the list into an indexable-page plan
Use the topic clusters as a skeleton: each pillar is a candidate category, each spoke a subcategory or a filter. Keep a combination as a crawlable page only when it appears as a distinct phrase with its own demand; otherwise leave it as a filter and plan to keep those URLs out of the crawl, as Google's faceted navigation guidance describes. Then open your platform's category editor and add a unique intro of a few sentences to the two or three top categories instead of a templated paragraph. Give every paginated page its own URL and a self-referencing canonical. The tool does not read your site, robots.txt or canonicals, so verify those separately, and a strong idea list proves nothing about indexing.
Example scenarios
1. Standing desk store with 6,000 filter URLs
- Situation
- Illustrative example: a store with 400 SKUs has one Desks category, and the height, width and colour filters have produced about 6,000 crawlable URLs while the newest collection is slow to appear.
- Input
- Seed: standing desk | Language: English | Country: United States | Depth: Standard
- Reading the result
- A typical run would return a long list of ideas grouped into clusters such as electric, small and corner desks. Phrases that recur across engines with a high opportunity score are candidates for their own category pages. A colour-plus-size combination that appears nowhere in the list is a filter, not a page.
2. Boots category hidden behind load more
- Situation
- Illustrative example: a boots category lists 432 products across 18 pages of 24, every page canonicalizes to page 1, and products past the first 48 are reachable only through a load more button.
- Input
- Seed: ankle boots women | Language: English | Country: United Kingdom | Depth: Standard
- Reading the result
- The clusters would typically separate subtypes such as chelsea or heeled boots. Each subtype with solid demand becomes a linked subcategory, so deep products are reachable in two clicks instead of relying on page 14. The result says nothing about whether your pagination markup is correct.
3. Manufacturer copy on 120 bottle listings
- Situation
- Illustrative example: 120 insulated bottle products use the supplier's description word for word, so the same 90 words appear on dozens of competing stores.
- Input
- Seed: insulated water bottle | Language: English | Country: Canada | Depth: Deep
- Reading the result
- The questions and modifiers sections would typically surface shopper wording, such as how long a bottle keeps drinks cold or whether it fits a cup holder. Use a few of those phrases to rewrite the first two paragraphs of your ten best sellers. The tool will not detect duplicates itself.
Common mistakes
- Reading the Demand Index as volume. A score of 60 is a relative position within one run, not 60 searches or a forecast. Plan budgets with a real volume source.
- Building a page for every filter combination. Strong phrases deserve pages, but most colour, size and sort combinations are not searched as phrases. Turning them all into pages recreates the crawl waste you were trying to escape.
- Seeding with a brand or SKU code. A narrow seed leaves the expansion little room, so you miss the product-type phrases a category page should target. Start broad and let the long-tail variations appear.
- Trusting the intent label blindly. Labels come from word lists and can misjudge a phrase. Click the row to check what actually appears in the live search before choosing a page type.
- Treating the list as a content audit. The tool cannot tell whether a page of yours already ranks or whether two categories compete for the same phrase. Check your existing URLs before creating new ones.
Frequently asked questions
How do I do keyword research for ecommerce category pages?
Start with the plain product-type noun as the seed, choose your market, and run the tool. Then work through the priority targets and topic clusters, and match each strong phrase to one category URL. If two categories seem to answer the same phrase, merge or differentiate them before writing copy. Click a phrase to see the live search and confirm the intent label. The output gives relative demand and a heuristic difficulty, so it narrows your choices rather than guaranteeing that a category page will rank or draw traffic.
Can keyword research fix faceted navigation SEO?
No, it informs the decision but does not change how Google crawls your store. What it does is show which filter combinations shoppers actually type as phrases, such as a colour plus a product type, so you can decide which deserve a crawlable page. Everything else stays a filter. Applying that decision means canonical tags, noindex or robots.txt rules in your platform, which are separate checks. A short, evidence-based list of allowed combinations keeps the decision concrete.
Does the Demand Index equal monthly search volume?
No. It is a relative 0-100 index built from autocomplete weights, suggestion position, agreement across engines and breadth of variations. Scores of 60 and 30 tell you which phrase looks stronger in this run, not that one gets twice the searches. For budgets or forecasts you need a volume source or your own Search Console data. Use the index to rank ideas against each other and to spot long-tail phrases you had not considered.
What should I search for product page SEO versus category pages?
Use broad product-type seeds for category pages and specific products or model families for product pages. Deep depth adds the digits 0-9, which surfaces model numbers and version-specific queries that are useful in product titles. Questions in the results often suit a buying guide or an FAQ block on a product page. The tool does not see your catalog, so match results to the SKUs you actually stock; chasing a phrase for something you do not sell wastes the page.
Will this help with duplicate product descriptions?
Indirectly. The keyword list gives you the vocabulary shoppers use, so you can write original descriptions instead of pasting manufacturer copy that every other retailer also uses. It does not compare your descriptions with anything or detect duplicates; a similarity or thin-content check does that. Start with your ten best-selling products in your top category, rewrite the first two paragraphs using a few phrases from the list, and keep the specification table as it is.
Reference: Google's guidance on managing faceted navigation URLs explains that filters can generate infinite URL spaces, leaving crawlers less time for new, useful pages.