An seo friendly ecommerce website is decided in decisions made before launch — URL patterns, category depth, how filters generate links, whether products live under one path or five. Tag-level fixes afterward are patching. Structural mistakes made on day one cost years.
This is the setup that holds up as a catalog grows from 200 products to 20,000, written in the order the decisions actually come up.
URL structure you won’t regret
Keep URLs short, readable, lowercase, and hyphenated. That much is uncontroversial. The decisions that matter more are the ones nobody revisits later.
- Serve products from a flat path.
/products/acme-trail-400/rather than/shoes/running/trail/mens/acme-trail-400/. A product reachable through five categories should not have five URLs, and a flat path removes the duplication problem instead of canonicalizing around it. - Categories get real paths reflecting the hierarchy:
/hiking-boots/waterproof/. Two levels deep is plenty; three is the ceiling. - Never put dates, IDs, or SKU codes in URLs that customers see.
/p/48839201/tells Google and shoppers nothing. - Decide trailing slashes once and enforce it with a 301. Same for www versus non-www and http versus https. Mixed variants are free duplication.
Whatever you choose, treat URLs as permanent. Renaming a category because marketing changed the label, without a 301, is a self-inflicted traffic loss that surfaces three weeks later when nobody connects it to the change.
Category architecture driven by search demand
Merchandising teams organize catalogs by supplier or season. Shoppers search by attribute and use case. The gap between those two structures is where most missed ecommerce traffic sits.
Category pages usually carry more commercial search volume than product pages — “standing desks” gets searched far more than any specific desk model — so your category set should mirror how demand is actually phrased. Pull keyword data across every plausible description of your inventory and look for clusters that don’t have a page: attribute categories (“waterproof hiking boots”), audience categories (“office chairs for tall people”), compatibility categories (“cases for iPhone 15 Pro”).
Build a page where there’s real volume and enough products to fill it. Don’t build one for every attribute combination — that’s how stores end up with 40,000 near-empty listings. A category with four products and no search demand is a thin page in a category’s clothing.
Keep click depth shallow. Every product should be reachable within three clicks of the homepage. Anything at five clicks gets crawled rarely and treated as unimportant, and no XML sitemap compensates for that — sitemaps aid discovery, they don’t signal importance.
Filters: the decision that scales worst
Faceted navigation is the biggest technical risk in ecommerce and it’s easiest to control before launch. Five filters with six options each generates roughly 46,000 URL combinations per category, before sort order and pagination. Left crawlable, that’s an infinite URL space eating your crawl budget.
Set the policy per facet from the start:
- Canonicalize filtered views to the clean category URL when the facet has no independent search demand — color and size usually qualify.
- Block parameter paths in robots.txt for anything that should never be crawled: sort order, view toggles, price sliders, session IDs.
- Use noindex, follow where a page must stay crawlable for link discovery but shouldn’t be indexed. Note that robots.txt disallow and noindex cancel each other out — a blocked URL is never fetched, so its noindex is never read.
- Give static URLs only to facets with real search demand. “Waterproof hiking boots” earns its own page with a title, H1, and intro copy. “Blue, size 9, price ascending” does not.
Cheapest control of all: render filters you don’t want crawled as buttons with JavaScript state rather than anchor tags. The URL space never gets discovered because it never gets linked.
Pagination, current rules only
Google stopped using rel=”next” and rel=”prev” as indexing signals years ago and said so publicly. Most ecommerce checklists still list them. They’re inert.
What works: keep paginated pages crawlable, self-canonicalize each page to itself, and link between pages with real anchor tags. Never canonicalize page 5 back to page 1 — that orphans every product past the first page. Keep the descriptive category copy on page one only so pages 2 onward aren’t duplicating it. If a category runs to 40 pages, the fix is more subcategories, not better markup.
Content that isn’t the manufacturer’s
The single most common reason ecommerce sites don’t rank has nothing to do with technology. Forty retailers paste the same 90 words of manufacturer copy, Google keeps one, and it’s the marketplace or the brand’s own store because site strength breaks the tie.
Escaping that doesn’t require thousands of words. It requires information the source text lacks: sizing notes from returns data, compatibility warnings your support team repeats weekly, an honest comparison to the nearest two products in your own catalog, who the product suits poorly. Two paragraphs of that outperform 600 words of paraphrased spec sheet.
On categories, add a short intro above the grid and a genuine buying guide below it. On products, add a spec table, an FAQ built from real support questions, and reviews — customer reviews are the cheapest source of unique text at scale and they keep accumulating without you.
Product schema and stock lifecycle
Mark up products with name, image, description, brand, SKU or GTIN, and an offers block carrying price, priceCurrency, and availability. It won’t lift rankings directly, but price and availability snippets lift click-through, which on a crowded SERP is worth about as much. Keep availability accurate; schema claiming InStock on a sold-out page invites a manual action. Review markup must not be self-serving — genuine customer ratings are fine, ratings you wrote about your own products are not.
Decide the stock lifecycle policy before your first product goes out of stock, because platform defaults will otherwise delete pages that took years to rank. Temporarily out of stock: keep the page live and indexed, set OutOfStock, show restock timing and alternatives. Discontinued with a successor: 301 to the nearest equivalent, not to the homepage. Discontinued with no equivalent, no links, no traffic: 410 Gone. Seasonal: keep the URL forever rather than rebuilding it each year.
Speed, and keeping it healthy as you grow
Core Web Vitals move both rankings and conversion on ecommerce, which makes them worth real effort — but they’re a tiebreaker, not a rescue. A fast page with duplicate content still loses to a slow page with original content. Compress and correctly size images, lazy-load below the fold, cut dead third-party tags, and reserve layout space for late-loading elements. Judge on real-user field data, not a desktop lab score.
The maintenance problem is that everything above degrades. New categories launch with missing copy, a plugin update rewrites canonicals, a migration leaves four-hop redirect chains. Recurring crawls are what catch this. SEO Rocket runs instant quick scans of around 25 pages and deep full-site crawls verified at 900+ pages, returning concrete evidence per issue — actual duplicate titles, actual URLs, actual H1 text — alongside PageSpeed and Core Web Vitals with real-user field data. Its keyword research and content gap tools decide which categories deserve pages, and its writer produces category and product copy that isn’t recycled boilerplate. It does not manage product feeds, Merchant Center, or inventory; that stays in your platform.
Set expectations honestly on timing. Structural changes need four to twelve weeks of recrawling before results appear, daily movement of two or three positions is noise, and third-party volume and position numbers are modeled estimates. Benchmark against the weakest site currently on page one, not the category leader — that’s the gap you actually have to close first.
