Chapter 10: International and E-commerce SEO

Store architecture retrofit SOP (Shopify or WooCommerce)

1. Stop the duplicates before adding anything. Shopify in particular exposes the same product at several paths; remove collection-scoped links from templates so every internal link points at the canonical product URL. Then block internal search and non-valuable sort parameters from indexing and crawling.
2. Put policy data into the markup. Organization, BreadcrumbList, and Product markup that includes shipping details and return policy — the fields that decide whether a result can show the extras that lift click-through.
3. Split the collection tree by intent, not by taxonomy. Under a head category, build the sub-collections customers actually search for — the attribute combinations that appear in queries, not the ones that appear in your warehouse system. Twenty useful sub-collections beat two hundred generated ones.
4. Give the top collections a real page. A short intro that states what the collection is for, internal links to the sibling refinements, and a buying guide plus FAQ lower down. A grid of products with no text is a filter, not a landing page.
5. Fix the feed, then connect it. Merchant Centre titles built on a pattern — brand, audience, product, material, colour, size — outperform the manufacturer’s name. Enable free listings, and treat the feed as a separate ranking surface with its own rules.
6. Watch the long tail weekly. Which parameter combinations are getting impressions? That report tells you which whitelist entries to promote to real pages next.

Interview questions and answer frameworks

Q1: A brand is launching in the US, Germany, France and Japan. Which domain structure do you choose, and what are the traps?

Framework: For almost every growing brand, subdirectories on one root domain — /, /de/, /fr/, /jp/ — because authority accumulates in one place instead of being split four ways, and each new market inherits the root’s strength rather than starting at zero. Separate country domains only make sense when local legal or brand reality demands it, and the cost is years of link building repeated per market. Then the three traps, named concretely: forced IP redirects, which can leave the German directory permanently uncrawled because the crawler resolves to a US address — use a dismissible banner instead; hreflang that is one-way, missing x-default, or conflicting with a non-self-referential canonical — the German page must canonically point at itself, not at English; machine translation with no transcreation — the query vocabulary differs, and so do the trust signals, from an Impressum in Germany to local payment methods and tax-inclusive pricing.

Q2: A store has 50,000 SKUs and faceted navigation generates millions of parameter URLs. How do you handle it?

Framework: Two tiers, and be explicit that the ratio is the point. Tier one, deny by default — combinations with no search demand: sorting, price sliders, stock status, ratings, and any stack of three or more attributes. Do not emit real links for these in the HTML; use asynchronous interaction instead. Block or canonicalise the rest at the URL rule level so the crawler cannot loop. Tier two, promote the few that earn it — the combinations with real query volume: head category plus one or two core attributes. Give those clean static URLs, their own titles and headings, an intro written for that combination, a self-referential canonical, and a sitemap entry. The discipline is that a facet becomes a page only when there is evidence someone searches for it; otherwise you have built an infinite crawl trap with good intentions.

Q3: How do you decide whether a market needs its own content or a translated version?

Framework: Start from the SERP, not from the language. If the top results in that market are local brands, local retailers and locally hosted content, translated pages will not compete — the market expects a local voice, local proof and local logistics information. If the top results are global brands with localised pages, translation plus proper hreflang is enough to start. Then check the query vocabulary: if the market searches for a different concept rather than a translation of yours, you need transcreation, not translation. And check the trust layer — payment methods, tax presentation, legal pages — because a German buyer will leave a store that looks like it was built for California.

Q4: Which e-commerce metrics do you actually optimise for?

Framework: Revenue per session, not sessions. That framing forces you to work on the surfaces that change it: which categories get indexed, whether collection pages answer the query, whether product pages carry the information that removes doubt — shipping cost, delivery time, returns, sizing — and whether the feed surface is competing at all. Then the second-order metric: indexed pages that earn nothing. On a large catalogue, the largest single win is usually not more traffic but fewer useless URLs competing with the useful ones.