E-commerce SEO for catalogues that outgrew the template

Why e-commerce SEO is its own discipline

Most search advice quietly assumes a site where a person decided that each page should exist. A catalogue site is not that. Pages are generated — by the product feed, by the filter component, by the pagination rule, by whatever the platform does when a colourway gets its own address. The central question stops being how do we make this page rank and becomes which of these several hundred thousand URLs should exist at all, and are we prepared to defend that decision when the catalogue turns over next quarter.

The second difference is that value is not evenly distributed. On a content site, two pages earning the same traffic are worth roughly the same. On a store, one category page holding a head term can be worth more than several thousand product pages combined, while a beautifully ranking page for a discontinued line contributes nothing at all. Any programme reporting on sessions rather than revenue by landing page is measuring something that does not pay wages.

The third is that the platform has opinions. Shopify, WooCommerce and Adobe Commerce each generate URLs, canonicals and collection structures in their own way, and several of those defaults are actively unhelpful. A meaningful share of e-commerce SEO is knowing which platform behaviours can be changed, which can only be worked around, and which are not worth the fight.

Category architecture is where the demand actually sits

Buyers search for categories far more often than for products. “Waterproof walking boots” is a category query with sustained volume and clear commercial intent. A specific model name is a product query, and by the time someone types it they have usually already decided — frequently somewhere else.

So the map starts at the category layer: real demand grouped into the terms your buyers use, matched to the page type that can satisfy each one, and reconciled against what you actually stock. That last constraint is the one most keyword-led approaches skip, and it is why stores end up with empty category pages built for terms nobody can currently buy against.

When a facet deserves its own page

Some filter combinations are categories in disguise. If “waterproof” exists on your site only as a checkbox, but people search for waterproof boots consistently and you always carry them, that combination should be a permanent, indexable URL with its own copy, its own place in the navigation and its own internal links — not a filtered state a crawler may or may not stumble into.

The test we apply is deliberately awkward: could you write two honest paragraphs about this page that are not machine-generated, and will there still be stock behind it in six months? If the answer to either is no, it stays a filter. Promoting facets indiscriminately is how a store acquires forty thousand thin pages and then wonders why the good ones stopped being crawled.

Product pages are a template problem

Optimising product pages one at a time does not survive contact with a catalogue that changes weekly. Everything has to be expressed as template behaviour: a title pattern that stays useful past ten thousand SKUs rather than collapsing into brand plus product code, headings driven by structured fields, specification tables the merchandising team can populate, breadcrumbs, and internal links to the parent category and to genuinely related items rather than only to whatever the recommendation engine surfaces.

Supplier copy is the recurring problem. There is no duplicate content penalty in the sense people usually mean, but identical text across forty retailers gives a search engine no reason to prefer your listing. Since nobody is going to write ten thousand descriptions, the honest approach is to rank the catalogue by revenue and margin, commission real first-party writing for the small share that earns most of the money, and make everything else differentiate structurally — complete attributes, real photography with meaningful filenames, sizing and compatibility notes, delivery and returns detail, and first-party reviews.

Facets, filters and crawl waste

Google’s crawl budget documentation is refreshingly blunt: most sites do not need to think about it at all, and it names roughly a million pages updated weekly, or ten thousand pages updated daily, as the point at which it starts to matter. Faceted stores cross that line without anyone noticing, because the URL count is not the product count — it is every filter dimension multiplied together.

The available controls are not interchangeable, and treating them as though they were is the most common mistake we find. Disallowing in robots.txt stops crawling but does not remove URLs already in the index. Fragment-based filters are generally ignored by Google, which makes them neutral for crawl budget but also invisible if you wanted them indexed. A canonical tag is a strong hint that consolidates signals over time, not an instruction. A nofollow attribute on filter links only helps when every anchor pointing at that URL carries it, which is rarely true once a sitemap or a third-party template is involved. And empty filter combinations should return a 404 rather than a cheerful page with no products on it.

What we produce is a ruling per facet dimension, written down with its mechanism and its reason, so that the next person to open your robots.txt can tell whether a line is a deliberate decision or an artefact. Undocumented crawl rules routinely outlive everyone who understood them.

Duplicate content: variants, parameters and syndicated text

Three unrelated problems get filed under the same heading.

Variants. A size or colour with its own address. Google’s e-commerce URL guidance accepts either a path segment or a query parameter, and suggests that where a parameter is optional, the parameter-omitted version should be the canonical — which makes the relationship between variants explicit rather than leaving it to be inferred.

Parameters. Sort orders, view toggles, tracking tags and session identifiers. Use a consistent key-and-value format, never repeat the same key twice in one URL, keep session identifiers out of addresses entirely, and put self-referencing canonicals on every page you want indexed.

Syndicated text. Covered above, and the only one of the three that is a content problem rather than an architecture problem.

It is worth being precise about what happens when duplicates proliferate: not a penalty, but consolidation. Google picks one URL as canonical and the others stop earning independently. If it picks a different one from the one you intended, that is a signal problem, and the fix is redirects and internal linking rather than louder canonical tags. Our technical SEO work covers the same ground on non-retail sites; scale is what makes it hard here.

Pagination and listing pages

Each page in a paginated sequence should have its own URL and its own self-referencing canonical. Pointing every page back at page one is a widespread habit and it removes pages two onward from consideration entirely. Links between pages must be real anchors with href attributes, because Google crawls URLs found in href and does not press load-more buttons. Infinite scroll with no paginated fallback usually means a crawler sees the first screen of products and nothing else — which, on a category holding four hundred items, makes a substantial part of your catalogue invisible.

Structured data that earns something

The list of markup worth implementing on a store is shorter than the list of markup that exists. Product markup on product pages, ProductGroup where variants are modelled, BreadcrumbList across the site, Organization on the home page, and ItemList on category pages where the listing genuinely is the main entity. Beyond that, returns fall away quickly.

Two rules govern all of it. Markup must describe what a visitor can see on the page. And availability must be truthful: schema.org defines a full set of values including in stock, out of stock, back order, pre-order and discontinued, and wiring those to live inventory rather than a hardcoded string is the difference between markup that helps and markup that eventually causes a problem.

Merchant listing markup and product snippet markup are also different things with different eligibility. Where products can be bought on the page, the merchant listing properties — price, availability, shipping, returns — are what open the shopping surfaces. Getting product data into Merchant Center as well as onto the page is worth doing, because free listings put your catalogue in front of people who never reach a classic blue link, and the feed and the page are cross-checked against each other. Inconsistency between the two is a common and entirely avoidable cause of disapprovals.

On-site search tells you what the catalogue is missing

Site search logs are the cheapest and most honest keyword research available to a store: first-party, unsampled, and typed by people already on your site with money in hand.

Zero-result queries in particular sort into three buckets with three different owners. Either you do not stock the thing, which is a merchandising decision; or you stock it under a name nobody uses, which is a content and synonym problem and almost always an organic opportunity too; or your internal search engine is simply bad at matching, which is an engineering fix. Most stores read the report once, find it interesting, and never assign the actions. Reviewing it on a schedule with a named owner outperforms a great deal of more expensive work.

Out-of-stock, discontinued and seasonal

These get treated as one problem and they are three.

Temporarily out of stock. Keep the URL, keep it returning 200, mark availability accurately, and give the visitor something to do — a notify-me form, close alternatives, an honest restock note. Deleting a page because of a temporary condition throws away everything it has accumulated.

Seasonal. Keep the URL live year-round and let the content change. Deleting and recreating a seasonal category every year is one of the most reliably self-inflicted wounds in retail search: the page starts from nothing each time, precisely when you need it working.

Permanently discontinued. If a genuine successor exists, redirect to it. If nothing replaces it, redirecting to the parent category is defensible only where that category is a reasonable destination for the person who clicked. Otherwise let the URL return 404 or 410 and accept the loss — Google warns that redirecting many old URLs to loosely relevant destinations produces soft 404s, which is exactly what a blanket redirect to the homepage achieves.

Performance on image-heavy templates

Category pages are the worst-case template on most stores: dozens of product images, a filter component, a merchandising banner, a review widget and usually a personalisation script. The Largest Contentful Paint element is normally either the category headline or the first product image, and the fixes are unremarkable — correctly sized modern image formats, explicit dimensions on every image so the grid does not shift as it loads, lazy-loading below the fold but never on the LCP element, and genuine scrutiny of third-party scripts.

The part worth arguing about is the banner. A full-width promotional image above the product grid is a merchandising decision that costs load time on the template that converts, and it deserves to be defended on its numbers rather than assumed. Where the build itself is the constraint, that is store development work rather than an SEO recommendation, and we will tell you which of the two you are looking at.

Replatforming without giving the revenue away

This is where the largest single-day losses in e-commerce search happen, and they are almost always avoidable.

The redirect map is the visible half: every existing URL inventoried from a crawl, the sitemaps, Search Console, analytics and backlink data, then mapped to a single-hop, server-side permanent redirect. Backlink data earns its place because a page with negligible traffic can hold the external links propping up the section around it. Redirects should stay in place for at least a year on Google’s own advice, and chains should be kept short.

The invisible half is content parity, and it fails more often. Migrations routinely ship a new category template that dropped the three hundred words the old one carried, a product template without the specification table, breadcrumbs rendered only in JavaScript, or a faceted system with entirely new URL patterns and no rules attached to them. None of that appears on a redirect checklist. All of it costs traffic.

We rehearse the map against staging before launch rather than validating it in production, and we will argue — in writing, on the record — against a go-live date that lands inside peak trading.

Measuring against revenue rather than rankings

Position tracking is a diagnostic, not a report. What a store needs to know is which page types earn money, and whether the mix is shifting.

That means GA4 ecommerce events checked end to end rather than assumed. The shopping events are only useful if they actually fire with the right item parameters, and broken implementations are extremely common — a purchase event missing item-level data quietly removes your ability to attribute revenue to a category at all. It also means splitting organic performance by page type and by brand versus non-brand, because brand demand created by your paid channels will otherwise flatter the organic numbers. And it means joining Search Console and Merchant Center data to landing pages, so impressions on shopping surfaces sit alongside classic results. Where that reporting needs building properly, our analytics practice does that work.

How this fits with everything else

E-commerce SEO overlaps with more disciplines than most search work. The category and facet decisions are architecture, which makes them a conversation with whoever builds the store. The measurement is analytics engineering. The demand map produced for organic is the same map that should inform paid search and shopping bidding, and the two channels compete for the same terms often enough that running them without a shared view of the catalogue wastes money in both. It also sits inside the broader organic programme, which sets the topical and link strategy this page assumes.

If you are not sure whether your constraint is search, merchandising or the platform itself, tell us what you are seeing and we will give you our honest read — including, sometimes, that SEO is not the thing holding you back.