seo
How to replatform without losing your search traffic
Migrations lose traffic for one of four reasons, and all four are preventable. The redirect map, the URL audit, the staging leak and the launch-week panic.

Most replatforms lose organic traffic. The ones that don’t aren’t luckier — they treated the URL layer as a deliverable with a named owner, in the same way as the design or the checkout.
Here is what actually goes wrong, roughly in order of how much damage it does.
1. The redirect map was built from the wrong end
The common approach is to launch, export the new site’s sitemap, and map old URLs to it.
By then the information you needed is gone: nobody remembers that /collections/sale-2019/
earned links from three magazines, so it maps to the homepage, and those links stop
counting for anything.
Build it from your own data, before design starts. If nobody on the project owns search, this is the moment to bring in an ecommerce SEO agency rather than after launch:
- Search Console — every page with impressions in the last 16 months. Not just clicks. A page ranking at position 14 is an asset you are about to either promote or destroy.
- Analytics — every URL with a session in the last 12 months, including paid and email landing pages.
- Your backlink data — every URL anyone else has linked to, however old.
- The raw CMS export — everything that exists, to catch what the three above missed.
De-duplicate that into one list. It is usually two to five times longer than anyone expected, and that discovery is the entire point of doing it early.
2. Redirects point somewhere “close enough”
Every redirect should land on the closest genuine equivalent, and a 301 to a page that doesn’t answer the same question is treated much like a soft 404.
The rules worth holding:
- One hop. Old URL straight to final URL. Chains of two and three redirects accumulate fast across migrations and every hop is latency on a real visitor’s first click.
- No mass redirects to the homepage. If there is no equivalent page, consider keeping a page. A discontinued product with links and traffic is worth a real page explaining what replaced it.
- Query strings and trailing slashes are different URLs. Pick one canonical form, enforce it at the edge, and redirect the rest.
- Case sensitivity changes between platforms.
/Shop/and/shop/may be one page on the old stack and two on the new one.
3. The staging site got indexed
A staging site on a public URL with a copy of your catalogue is a duplicate of your store. It gets crawled, sometimes ranks, and occasionally outranks you — and then launch day involves de-indexing a site while trying to index another.
Password-protect staging at the HTTP level. robots.txt is a request, noindex only works
if the page is crawled to find it, and neither stops a link from being shared.
The mirror image of this is worse and more common: the noindex that protected staging
ships to production. Before you celebrate launch, view source on the live homepage and look
at the robots meta with your own eyes. This is the single most expensive one-line mistake in
ecommerce, and it is invisible from the front end for weeks.
4. Nobody watched the right numbers afterwards
Traffic moves for many reasons and launch week is noisy. Decide beforehand what you will watch, so that a dip triggers investigation instead of argument:
| Signal | Where | What a problem looks like |
|---|---|---|
| Crawl errors | Search Console | 404s climbing from day one |
| Indexed pages | Search Console | A falling count that doesn’t recover in three weeks |
| Top 50 landing pages | Analytics | Individual pages at zero, rather than an even decline |
| Redirect health | A crawler over your old URL list | Any 404, any chain, any redirect to the homepage |
| Core Web Vitals | Search Console, field data | A new stack that is slower than the old one |
Run a crawl over your full old-URL list on launch day, day three, and day thirty. It takes minutes and it catches the entire category of “we forgot a rule”.
What a normal migration looks like
A 10–20% organic dip for two to six weeks is normal while search engines recrawl and reassess. It should recover, and a well-executed replatform onto a faster stack usually ends up ahead within a quarter.
What is not normal: a dip deeper than 30%, a decline that is still deepening after a month, or one concentrated in your highest-value pages. Those are faults. The most useful thing you can do is be able to tell the difference, which means having the before numbers written down somewhere other than someone’s memory.
What to ask an agency before you sign
- Who owns the redirect map, and when do I see a draft? (Answer should be a named person and a date before design sign-off.)
- How is staging protected from indexing?
- What is your launch-day checklist, and can I see it?
- What are we measuring in the first 30 days, and what number triggers a rollback?
An agency that has done this before will have a written answer to all four already. The replatforming agencies in this directory list migration as a service; if the move is onto Shopify, start with the Shopify migration specialists.


