Ecommerce migrations fail differently from other site moves. A brochure site has tens of URLs and one template; a store has thousands of product URLs, a filter layer generating many more, and a category structure carrying most of the commercial rankings. The volume is what breaks the process.
What makes a store harder to move
- URL count. Redirect mapping cannot be done by hand, so it has to be generated and then verified.
- Platform-imposed URL patterns. Most platforms dictate their own product and category path structure, so paths change whether you want them to or not.
- Discontinued products. Every migration surfaces a backlog of products that no longer exist, and each needs a decision.
- Filter and parameter URLs that may be indexed on the old site and behave differently on the new one.
- Categories carry the rankings. Losing category URLs costs far more than losing individual products.
Sequence that works
1. Export and classify the old URLs
Crawl the live site and combine that with Search Console and analytics data. Classify each URL as category, product, content, filter or parameter. Products with no traffic, no rankings and no stock are the ones you can retire; everything else needs a target.
2. Map categories first, individually
Categories are where the commercial demand sits, so map them one by one rather than by rule. If the new structure differs, decide deliberately which new category inherits each old one. See category page SEO.
3. Map products by rule, then spot-check
Product redirects can be generated from SKU or handle matching. Generate the map, then manually verify a sample across each category and every edge case you can find.
4. Decide what happens to discontinued products
- Replacement exists: redirect to the replacement product.
- No replacement but the category is right: redirect to the category.
- Genuinely gone with nothing comparable: return 410 rather than redirecting everything to the homepage.
Redirecting large numbers of dead products to the homepage is treated as a soft 404 and wastes the equity you were trying to keep.
5. Do not carry over filter URLs
Indexed filter URLs on the old site are usually a problem you inherited. A migration is the natural point to stop indexing them rather than faithfully reproducing them. See faceted navigation SEO.
6. Test on staging before launch
Crawl the staging build and confirm the redirect map resolves, canonical tags are correct and pagination works. See staging site SEO.
The mistakes that cost rankings
- Redirect chains, where old URLs point to URLs that themselves redirect. Point everything at the final destination.
- Mapping all products in a category to the category page instead of to individual products.
- Launching without the redirect map live, even for an hour.
- Forgetting to update the sitemap, so the old URLs keep being submitted.
- Blocking the new site in robots.txt from staging and not removing it.
That last one is the most common single cause of a catastrophic post-launch drop, and it is trivially avoidable with a pre-launch check.
Frequently asked questions
How long before traffic recovers?
A well-executed store migration typically shows a dip for a few weeks while URLs are reprocessed. Large catalogues take longer simply because recrawling everything takes longer. See SEO migration timeline.
Should I keep the old URL structure?
Where the platform allows it, yes. Preserving URLs removes most of the risk. Where the platform imposes its own structure, accept it and invest the effort in an accurate redirect map.
Can I migrate and redesign at the same time?
You can, and it makes diagnosis much harder if something goes wrong, because you cannot tell which change caused it. Separating them is worth the extra scheduling.
What about old URLs still showing in Google?
Normal for a period after launch, provided they redirect. If they persist for months, something in the setup is wrong. See old URLs still indexed.
Consulting CTA
If you are replatforming a store, book an SEO consultation before launch rather than after. See also website migration SEO cost.