Most stores don’t lack an ecommerce seo strategy — they lack a sequence. They run everything at once: rewriting product copy while the crawl is broken, buying links to pages Google is filtering as duplicates, chasing head terms with a taxonomy that doesn’t match how people search. The work isn’t wrong. The order is.
Ecommerce SEO compounds when you fix things in dependency order. Crawl and index first, because nothing downstream is measurable until Google can see your pages properly. Categories next, because that’s where the commercial volume sits. Product copy after that, tiered by opportunity. Links last, because links amplify pages that deserve to rank and waste money on pages that don’t.
Phase one: make the site crawlable and countable
Start with a full crawl and a month of server logs. You’re looking for the ratio of wasted crawl to useful crawl, and on an unmanaged store it’s usually alarming — 60 to 80 percent of Googlebot requests landing on filtered, sorted, or parameterized URLs while new products wait weeks for discovery.
Three fixes account for most of the recovery. Faceted navigation gets a per-facet policy: canonicalize filters with no independent search demand, block sort orders and price sliders in robots.txt, use noindex-follow where a page must stay crawlable but shouldn’t be indexed, and give static URLs only to facets people genuinely search for. Pagination gets self-canonicals and crawlable anchor links, with rel=”next” and rel=”prev” left out entirely since Google stopped using them years ago. And redirect chains get flattened to a single hop.
Budget four to six weeks for this phase and expect your indexed URL count to fall. A shrinking index during facet cleanup is the fix working, not a problem.
Phase two: build the category set demand actually wants
Category pages usually carry more commercial search volume than product pages. “Standing desks” gets searched far more than any individual desk model. So the second phase is aligning your taxonomy with how demand is phrased, which is almost never how the warehouse organizes inventory.
Pull keyword data across every plausible description of what you sell and look for demand clusters with no page behind them:
- Attribute clusters — “waterproof hiking boots,” “standing desks under $500”
- Audience clusters — “office chairs for tall people,” “beginner road bikes”
- Compatibility clusters — “cases for iPhone 15 Pro,” “brake pads for Civic 2019”
- Use-case clusters — “small kitchen appliances,” “gifts for new dads”
Build a page where there’s real volume and enough inventory to fill it. Not every combination — a category with four products and no demand is a thin page wearing a category’s clothes. Each new category gets a short intro above the grid, a genuine buying guide below it, and internal links from its parent and siblings.
Check intent before assigning targets. If page one for a term is dominated by listing pages, it belongs to a category. If it’s individual products and reviews, it belongs to a product page. Getting that backwards produces cannibalization, and Google resolves it by picking one of your pages, often the weaker one.
Phase three: tier your product content
The reason product content projects stall is that nobody decides what not to write. You cannot hand-craft 20,000 descriptions, and pretending otherwise means the project dies at SKU 400.
Rank the catalog by opportunity — search volume on brand-plus-model terms, weighted by margin and stock depth — then run three tiers. The top few hundred products get bespoke copy carrying information the manufacturer never wrote: sizing notes from returns data, compatibility warnings from support tickets, honest comparisons against the nearest alternatives you sell. The middle gets attribute-driven templates that assemble genuinely different specs, dimensions, and compatibility data, so the pages actually differ rather than swapping a name into the same sentences. The tail gets reviews and Q&A, which accumulate unique text on pages you’ll never touch.
This is the phase where the manufacturer-description trap gets closed. Forty retailers pasting the same 90 words produce forty near-identical documents, Google keeps roughly one, and it isn’t you — site strength breaks ties. There’s no penalty involved, which is why it’s so often misdiagnosed: pages sit indexed with near-zero impressions on their own product names.
Phase four: links, and only then
Links are necessary and not sufficient. A thin or duplicated page with links still loses, which is why link building belongs after the content tiers are in place — you’re amplifying pages that can now hold a ranking.
On ecommerce, the assets that actually attract links are rarely product pages. Buying guides, comparison content, original data from your own sales patterns, and genuinely useful tools are what earn coverage; internal links then distribute that equity to the categories and products that convert. Digital PR, supplier and manufacturer listings, and niche publication reviews are the practical channels.
Be honest about cost estimates. Link cost bands are directional; relevance, quality, and velocity decide outcomes far more than volume does. Benchmark against the weakest site currently on page one rather than the category leader, because that’s the gap you actually have to close.
Set up measurement before you need it
Ecommerce SEO reporting fails when it starts from rankings. Positions bounce two or three places daily as ordinary noise, and a weekly rank report will tell you contradictory stories in consecutive weeks.
Track four things instead. Crawl distribution from logs, which moves first and confirms phase one worked. Impressions by page type in Search Console, which shows Google understanding new pages before clicks arrive. Indexed-versus-submitted ratios, which reveal bloat and thin-page filtering. And revenue by landing page type, which is the only number the business cares about.
Connect Search Console and GA4 as ground truth beside any third-party estimates. Third-party volume, difficulty, and position data are modeled — roughly twelve-month averages from periodic crawls — and useful for prioritization, not for reporting outcomes.
Housekeeping that runs forever
Two policies need to exist permanently rather than as project tasks. First, product lifecycle: temporarily out-of-stock pages stay live and indexed with OutOfStock in schema and related items shown; discontinued products with a successor get a 301 to the nearest equivalent; discontinued products with no equivalent, no links, and no traffic get a 410; seasonal URLs are never rebuilt. Silent deletion, the platform default, throws away years of rankings.
Second, recurring crawls. New categories launch without copy, plugin updates rewrite canonicals, migrations leave redirect chains. Whatever you fixed in phase one degrades without monitoring.
SEO Rocket runs this loop in one place at a flat monthly cost: keyword research with country-specific indexes to decide which categories and products deserve pages, content gap analysis across up to five competing retailers, an AI writer built on the template behind a campaign that scaled a site past 30,000 published, ranking pages, rank tracking with trend framing plus Search Console and GA4 connected as ground truth, and technical audits ranging from a 25-page quick scan to full-site crawls verified at 900+ pages with actual URLs and titles attached to every issue. It does not manage product feeds, Merchant Center, or inventory — those stay in your ecommerce platform.
Realistic timelines
Phase one shows up in logs within two to four weeks and in index counts within eight. Category work takes four to twelve weeks per cohort before positions stabilize, longer on competitive head terms. Product content moves impressions before clicks, so watch impressions at week six rather than declaring failure at week three. Link acquisition compounds over quarters.
Twelve months is a fair horizon for a store starting from a broken crawl and manufacturer copy. Anyone promising faster on a large catalog is either working with an unusually strong domain or not being straight with you.
