Moving to Shopify without losing your rankings
A 40% drop in organic traffic after a replatform is common, avoidable, and almost always caused by the same four things. Here is the order we do it in.
Written by Alex Price
The horror story is real, and it is why a lot of people put off a migration they know they need. Organic traffic falls off a cliff in week two, nobody can say why, and it never fully comes back.
The good news is that it is nearly always the same handful of causes, all of them avoidable, and all of them cheaper to prevent than to fix.
Do the redirect map first
Before a single screen gets designed. Crawl the current site, pull every URL that has traffic, rankings or links pointing at it, and decide where each one lands. Every page that earns something gets a home — a relevant home, not a blanket redirect to the homepage, which search engines treat as a soft 404.
This feels like the boring part and it is the whole ballgame. A migration with a redirect map done properly at the start recovers in weeks. One where the map was assembled the night before launch often does not recover at all.
The four things that actually break
- Redirects. Missing, chained five deep, or all pointing at the homepage.
- Structured data. The old platform had product markup; the new theme quietly does not, and your rich results disappear.
- Duplicate collection URLs. Shopify will happily serve the same product under several paths if you let it, and split your own rankings between them.
- A sitemap full of things that should not be in it — tags, filtered views, out-of-stock variants.
What normal looks like afterwards
A wobble in the first fortnight is expected while Google re-crawls and re-indexes. Rankings move around, impressions dip, and it is uncomfortable. What matters is the shape of week three onwards.
A dip that recovers is a migration. A dip that plateaus is a redirect map that was written too late.
Watch it properly for a month
Not a glance at analytics a week later. Crawl errors, index coverage, rankings on your top hundred terms, and revenue — every day for thirty days. Almost everything that goes wrong is findable in the first fortnight and fixable the same day, if somebody is actually looking.
Should you redesign at the same time?
You do not have to, and there is a real argument for not doing so: change the platform and the design at once and you lose the ability to tell which one caused what. Sometimes it is still the right call — two disruptive projects is worse than one — but make it a decision rather than an assumption.
And write down a rollback plan before you go near a DNS record. It is rarely needed. It is very calming to have.
