Most e-commerce SEO audits open with a keyword report. That is the wrong end of the problem. A catalogue site with 3,000 products can easily generate 300,000 crawlable URLs through filter combinations, sort parameters and session identifiers — and search engines will spend their crawl allowance on them before reaching the pages you care about.
So the first artefact is a crawl, compared against server logs where available. You are looking for the ratio between URLs that exist and URLs that deserve to exist. When that ratio is worse than roughly ten to one, no amount of content work will move the needle until it is fixed.
Every filter combination needs one of three decisions, made deliberately. Allow it to be indexed when it has genuine search demand — 'running shoes size 10' often does. Canonicalise it to the parent category when it is a legitimate view with no independent demand. Block it in robots.txt when it is a machine artefact like a sort order or a session parameter.
The failure mode we see most often is a site that has made none of these decisions and instead applied a blanket noindex to everything with a query string, which quietly removes pages that were producing revenue.
A product in eight colours can exist as one URL with variant selection, or eight URLs each with thin, near-identical content. The second option feels better for long-tail coverage and usually performs worse, because the pages compete with each other and none accumulates enough signal to rank.
The workable middle path: one canonical product URL, variants handled through parameters that are canonicalised, and separate URLs only where a variant has independent search demand. Structured data should reflect the actual offer, including price ranges and availability per variant.
A category page with a heading, a product grid and nothing else has no more to say than the sitemap. The pages that hold position include buying guidance — how to choose, what the differences mean, which option suits which use — placed where it does not push the products below the fold.
Two hundred words of genuinely useful selection guidance below the first product row outperforms both a thousand words of keyword filler and nothing at all.
Product grids load dozens of images. Lazy loading below the fold, correct dimensions to prevent layout shift, modern formats and a sensible thumbnail size are worth more than most JavaScript optimisation on a catalogue site.
Test on a mid-range Android device on a throttled connection rather than a desktop with fibre. The gap between those two experiences explains most of the disagreement between marketing teams and developers about whether a site is fast.
Crawl and indexation control first, because everything downstream depends on it. Then duplication and canonical strategy. Then internal linking, which is how authority reaches deep product pages. Then page-level speed. Then content depth on the categories that carry the most revenue.
Doing these in reverse — which is the common instinct, because content is the visible part — produces months of work whose effect is masked by the structural problems underneath it.
Every recommendation here comes from accounts we run. If something in your setup contradicts it, the account is the better authority — test it.
If this describes a problem you currently have, a short review of your accounts will identify whether it is the binding constraint or a symptom of something else. That conversation is free and does not require a commitment.
Keep reading
Performance
Performance Max rewards clean inputs. A practical account structure that keeps brand, high-intent search and discovery from bidding against each other.
E-commerce
Assortment, pack size and availability decide performance long before advertising does. A working framework for launching on rapid-delivery platforms.
Tell us your goal and we will map the fastest realistic route to it — channels, budget split, timelines and the numbers we will hold ourselves to.