Category and facet decisions for a live product catalogue

A Shopify collection with 40 tags is not one page. Tag combinations are addressable at /collections/handle/tag-a+tag-b, and Google reaches them through your own filter interface whether you intended that or not. In the seven catalogues we inherited during 2025, between 58 and 91 percent of the URLs Google had indexed were tag or filter permutations that had never received a single click. This engagement decides which of those combinations become real pages with their own copy and internal links, which are closed off at the filter layer, and what rule applies to the next tag your merchandising team invents on a Friday afternoon. The decisions come from your Search Console export and your server logs. They are not copied from a rule someone published about a store that is not yours. Who does the work: One of the two principals runs the engagement end to end, and you talk to that person, not an account manager. Between us we have made this call for an Australian homewares retailer at 24,000 SKUs, a UK footwear brand running three storefronts off one BigCommerce catalogue, and a food business whose entire category structure was rebuilt around allergen tags. On the homewares account we removed 11,000 colour-tag URLs from the index three weeks before Black Friday, over the objection of the merchandising lead. Organic revenue that November closed 18 percent ahead of the prior year and the crawl recovery was visible in logs within nine days. What you are buying is a decision of that kind, made in front of you, with the reasoning written down so your successor can read it. The threshold we work to: A facet combination earns an indexable page when three things hold at once. It has held at least 50 monthly searches across two consecutive quarters. It has eight or more products that stay in stock through a normal season, not a peak. And it can be described in a paragraph that would be false if written about its parent collection. Fail any one of the three and the combination is handled at the filter layer instead, through Shopify Search & Discovery, Boost AI or Searchanise depending on what you already run, so the page keeps working for shoppers and stops existing for crawlers. What you receive: A facet map covering every tag and filter parameter in the catalogue, marked keep, consolidate or close, each with the evidence behind it. A URL, canonical and robots specification written against your actual theme or Stencil templates, in the form your developer can implement without a translation meeting. Category copy for the collections that survive the cut, placed in theme sections or metafields so it survives your next theme update. Product template decisions on variant handling, colourway URLs, specification tables and review markup. A rule document and a working session with merchandising, so tags can be added next quarter without reopening this. Not included: Link acquisition. Platform migration execution. Paid search or Merchant Center feed work, except that we flag where your feed and your canonicals contradict each other, because that one is usually costing money. Product copy at catalogue scale, we write copy only where copy is the binding constraint, and we tell you which pages those are rather than quoting you for all of them. Who should not buy this: Catalogues under about 500 SKUs. A spreadsheet and one focused afternoon fixes those, and we will say so on the call rather than take the fee. Anyone who needs agreement with a structure already signed off internally, because we will not provide it. Businesses whose stock turns fast enough that no product is reliably available for a season, where facet pages are the wrong instrument entirely. When the answer is unwelcome: Sometimes the finding is that your category architecture is sound and the problem sits elsewhere, in availability, in a checkout that fails on mobile Safari, in email cannibalising the branded terms you are measuring. We write that up and stop. Two engagements ended at that stage in 2025 and the balance was not invoiced. If a replatform date is locked, the route specification is the artefact that has to be right before your developers freeze URLs, and we will tell you plainly if the date is unrealistic.