A store's site is a small number of templates rendered a large number of times: category, product, filtered listing, search result, and pagination of each. A crawl of a store returns findings in the thousands, and nearly all of them are one of a dozen causes in one of those templates. An audit read per page is unreadable; the same audit read per template is a short list.
Most impressions on a store land on category pages, not product pages: the searches are 'women's running shoes', not one SKU. That is where a fix pays and where a comparison is possible, because a store has many category pages of similar depth and traffic to compare against each other.
The specifically ecommerce problems are all about multiplication. Every filter combination is a URL; every out-of-stock product is a page; every colour variant is a near-duplicate. The catalogue grows, the crawl budget does not, and the pages that matter get crawled less often because the ones that do not are in the way.
The findings that recur on this kind of site
- Filtered listings indexed as separate pagesSize, colour and price facets producing thousands of duplicate-content URLs with the same products in a different order. The fix is a canonical to the unfiltered category, or a noindex on the facet, decided once per facet.
- Product descriptions copied from the supplierThe same paragraph as fifty other stores, and often the same meta description across variants of one product. Google picks one store to rank for it, and it is rarely the smallest.
- Out-of-stock pages returning 200 with nothing on themThin content at scale, still in the sitemap, still being crawled. Decide a policy — keep with a restock note, redirect to the category, or 410 — and apply it in the template.
- Pagination canonicalised to page oneEvery products page after the first canonicalised away, which tells the crawler the products on pages 2–40 do not exist as far as this category is concerned.
- Category pages that are three clicks too deepSubcategories reachable only through a menu the crawler cannot open, showing up as deep pages or orphans despite being the pages with the most to gain.
What to change first, and what to compare it against
- Fix by template, ranked by the impressions behind itSum impressions across the pages each cause touches. The facet cause touching every category outranks the product-description cause, however many products there are, because the categories are where the traffic is.
- Use categories as each other's controlsChange the template for half the categories, matched by traffic band, and leave the rest. This is the control-group method, and a store is the easiest kind of site to run it on because it has so many comparable pages.
- Measure over a window that covers a full promotional cycleSales, launches and seasons move category traffic more than any fix does. Four weeks is the minimum; compare against the control over the same dates, never against last month.
- Read a no-effect verdict as informationA facet cleanup that changed nothing in four weeks is a real result. It tells you the crawl budget was not the constraint, and the next fix should be somewhere else. A verdict that can only say yes is not one.
Questions from people running one
Should out-of-stock products be removed?
Depends on whether they come back. A seasonal product that returns should stay live with a clear restock note, keeping its history. A discontinued one should redirect to the closest replacement or the category, or return 410 if there is none. What it should not do is return 200 with an empty page indefinitely, which is the default most platforms ship.
Does Rankelle understand my platform?
It crawls the rendered HTML like Google does and does not integrate with any platform's admin, on purpose. The findings are the same on Shopify, WooCommerce or a custom stack; what differs is where the template lives, and that is your developer's side of the fix. What we do not do has the full list.
Everything above is the method, usable with any crawler and a Search Console export. Rankelle runs it on a schedule: the crawl, the grouping into causes, the watch for the fix landing, the frozen control set and the verdict four weeks on — including the verdicts that say the work did nothing. One site is free, permanently, because the first verdict takes six weeks to arrive.
Try it on your own site, free
The whole audit of every website you add, no card and no expiry. Pay only when you want the proof.