Most ecommerce keyword research starts at the wrong end. You open a tool, type a seed term, export a thousand keywords sorted by volume, and then spend weeks trying to figure out where any of them should live. That is backwards. On an ecommerce site the keyword is not the deliverable — the page is. The real question is never “what are people searching,” it is “which page on my store should own this term, and which revenue-carrying pages don’t exist yet.” Get that routing wrong and you publish a blog post for a query that wanted a category page, then wonder why it converts at 0.2%.
Why Volume-First Research Fails Stores Specifically
Content sites can afford to chase raw volume because every ranking page monetizes the same way — an ad impression or an affiliate click. A store cannot. A term with 25,000 searches a month and no buying intent is worth less to you than a term with 400 searches from someone comparing two SKUs with a credit card open. This work has to price a keyword by its position in the purchase journey, not by its search volume, because your pages are graded on revenue, not sessions.
The practical failure mode is subtle: you rank, traffic arrives, and nothing sells. That is almost always a routing error — an informational query pointed at a product page, or a transactional query answered with a 2,000-word guide. Fixing it after the fact means rebuilding the page or 301-redirecting it, both of which cost you the authority you already earned. It is far cheaper to route correctly the first time.
Start With Your Catalog, Not a Seed List
Before you touch a keyword tool, export your actual inventory — every product, category, and attribute you already sell. That list is your ground truth, and it exposes the single most common problem in ecommerce keyword research: vocabulary mismatch. Stores name products the way merchandisers and brands talk; shoppers search the way they talk. A catalog full of “performance outerwear” is invisible to the far larger crowd typing “waterproof jacket,” and no amount of on-page optimization closes that gap until you notice it exists.
Run each catalog term through a keyword explorer and read the “also ranks for” and “matching terms” reports as a translation dictionary. You are looking for the demand-side word for every supply-side word you use. This is unglamorous and it is where most of the money is.
The Four-Bucket Routing Framework
Every keyword you find should be sorted immediately into the page type it wants. There are only four buckets, and the intent behind the query tells you which one:
- Product terms — a specific model, SKU, or brand-plus-item (“Garmin Forerunner 265”). These want a single product page.
- Category terms — the head noun for a range (“running shoes”). These want a category or collection page, never a blog post.
- Modified category terms — a category plus a qualifier (“trail running shoes for wide feet”). These want a filtered subcategory or facet page, and this is where most of your untapped volume hides.
- Informational terms — questions and research (“how to choose trail running shoes”). These want a guide, and their job is to capture demand early and internally link down to the category and product pages.
The test is mechanical: look at what Google already ranks for the term. If page one is all product pages, it is a product term. If it is all category listings, it is a category term. Google has already run the intent experiment for you across millions of clicks — copy its answer rather than guessing.
Modified Category Terms Are Where the Volume Hides
Head terms like “running shoes” are dominated by Amazon, big-box retailers, and brands with domain authority you will not out-muscle for years. The winnable volume lives one layer down, in the modifier combinations those giants treat as an afterthought. Five modifier families cover most of it: attribute (“waterproof,” “leather”), audience (“for women,” “for beginners”), price (“under $100,” “budget”), use case (“for marathon training,” “for flat feet”), and comparison (“vs” and “alternative” queries).
The mechanical catch is that most stores already have the page — it is a faceted-navigation URL that filters the category by that attribute — but it is set to noindex, canonicalized away, or buried behind JavaScript Google never renders. So the demand exists, the inventory exists, and the page exists, yet you rank for none of it because of a crawl-and-index configuration decision made years ago. Auditing which facets deserve to be indexable landing pages is often the highest-ROI hour in the whole project.
Mine Competitors for the Categories You’re Missing
Your catalog can only tell you what you already sell. To find the pages you should build but don’t have, you need the terms your rivals rank for and you don’t. A content and keyword gap analysis across four or five direct competitors surfaces exactly that — and on an ecommerce site those gaps are merchandising signals, not just content ideas. If three competitors all rank for “petite winter coats” and you don’t carry a petite collection page, that is telling you something about both your SEO and your assortment.
This is one of the steps SEO Rocket automates: its competitor gap analysis runs against real Ahrefs index data across up to five rivals, so you see the category and modifier terms driving their traffic and can decide which are worth a new indexable page. It turns a vague “we should have more pages” into a ranked list of specific pages with real demand behind each one.
Price Every Keyword by Expected Revenue, Not Volume
Here is the mechanism that turns a keyword list into a build queue. For each candidate term, estimate a crude expected value: monthly volume × realistic click-through at your target position × page conversion rate × average order value × margin. You do not need precision — you need relative ranking. A “buy” or “for sale” modifier at 500 searches often out-earns a bare category term at 8,000 because the intent is downstream and the conversion rate can be five to ten times higher.
This single reframe kills the most expensive habit in ecommerce keyword research: prioritizing the biggest number on the screen. Volume is an input, not the answer. The keyword worth building for is the one with the best expected revenue per unit of effort, which is a very different sort.
A Worked Micro-Example
Say you sell desk chairs. Your explorer returns three candidates. “Office chairs” shows 90,000 searches but a difficulty score in the 80s, page one owned by Amazon and Wayfair, and it is a broad category term — realistically a two-year fight. “Ergonomic office chair for lower back pain” shows 2,400 searches, difficulty in the 30s, and page one is a mix of mid-sized retailer category pages and guides — a winnable modified category term you can build a filtered collection around. “Herman Miller Aeron vs Steelcase Leap” shows 1,100 searches, is a pure comparison query, and page one is all editorial.
The volume-first instinct picks “office chairs.” The revenue-first read is different: build the ergonomic-back-pain collection page first (winnable, high intent, direct to product), publish the Aeron-vs-Leap comparison second (captures late-stage research and links straight to two high-margin SKUs), and put “office chairs” on a long-term roadmap you revisit once your category authority grows. Same three keywords, opposite priority order — and the second order actually converts.
Turn the List Into a Prioritized Build Queue
Score each surviving page idea on three axes: commercial value (the expected-revenue estimate above), attainability (difficulty judged against the weakest page-one result, not the average — you only have to displace position eight to ten), and effort (a brand-new page and product photography costs far more than optimizing a facet you already have). Sort by value-over-effort and you have a queue your writers and merchandisers can work top-down.
Keeping that queue in one place matters more than it sounds. SEO Rocket pools your keyword research, competitor gaps, and target pages into a single project, so the term you validated, the page you assigned it to, and the rank tracker watching it are the same object rather than three disconnected spreadsheets. Its validation-gated AI writer can then draft the guide and comparison pages against that brief, and its real-crawler site audit flags the facet and canonical issues that quietly strand your modified-category volume.
Honest Caveats the Tools Won’t Tell You
Keyword research is directional, not gospel, and pretending otherwise burns budget. A few truths worth internalizing:
- Volume figures are annual averages. A “space heater” at 40,000 searches is really 5,000 in July and 150,000 in December. For seasonal catalogs, look at the monthly trend, not the headline number, or you will build the wrong page at the wrong time.
- Difficulty scores are a single tool’s model. They are useful for sorting, not for absolute go/no-go decisions. Always sanity-check by actually looking at who ranks.
- Watch for keyword cannibalization. Routing two close terms to two separate pages can split your authority so neither ranks. When intent overlaps, consolidate into one stronger page rather than two weak ones.
- Zero-volume long-tail terms still convert. Tools under-report ultra-specific queries, and those are often your highest-intent buyers. Don’t delete a “long size 15 steel-toe work boots” term just because it shows “0.”
Measure the Right Thing After Launch
Give a new page eight to twelve weeks before you judge it — Google needs time to crawl, index, and trust a fresh URL, and rankings jitter daily regardless. Track non-brand impressions in Search Console before you obsess over clicks; impressions rising means you are getting indexed and considered, and clicks follow once you climb into the visible positions. Treat index-based rank estimates as directional and reconcile them against Search Console and GA4, which are your ground truth for what actually earned money. Rank tracking that shows a trend line beats any single-day snapshot.
Frequently Asked Questions
How many keywords should ecommerce keyword research produce?
Quantity is the wrong target. A mid-sized store might end up with a few hundred prioritized terms mapped to pages, but the useful number is how many new or improved revenue pages the research justifies. Ten well-routed keywords that each earn a converting page beat a thousand-row export nobody acts on.
Should I target product keywords or informational keywords first?
Build the highest-intent winnable pages first — usually modified category and comparison terms that route straight to product. Informational guides matter for capturing early-stage demand and internal linking, but they convert slowly, so they belong in the second wave once your money pages are live.
How is ecommerce keyword research different from a normal blog?
A blog routes almost everything to articles. A store has four destination types — product, category, faceted subcategory, and guide — and the entire skill is assigning each term to the right one. Miss the routing and you rank without selling.
Can I automate ecommerce keyword research?
The data-gathering and gap analysis, yes — that is exactly what tools like SEO Rocket do against real Ahrefs data, using a playbook proven across 1,000,000+ ranking pages. The routing judgment — which page owns which term — still needs a human who understands the catalog and the margins. Automate the collection; own the decisions.
The Bottom Line
Good ecommerce keyword research is not a list-building exercise, it is a routing and prioritization exercise. Start from your catalog to catch vocabulary gaps, sort every term into one of four page types by copying the intent Google already ranks for, mine competitors for the categories you are missing, price each keyword by expected revenue instead of raw volume, and build in value-over-effort order. Do that and you stop publishing pages that rank but don’t sell — and start building the ones that were always going to pay for themselves.
