Shopify Migration: What Actually Happens to Your Traffic

Every migration proposal you’ll read has a section on SEO. It is always short, always reassuring, and always says roughly the same thing: we’ll map your redirects, preserve your URL structure, and there may be a brief dip.

“A brief dip” is doing an enormous amount of work in that sentence.

Here’s what the evidence actually shows, and why the answer you get depends entirely on who you ask.

What Google says

Google’s own site-move documentation is genuinely optimistic. For a small or medium site done correctly, it says most pages move in the index within a few weeks. Larger sites take longer. Some ranking fluctuation during the transition is described as normal and expected.

John Mueller has been more specific over the years: processing a migration takes several months, redirects should stay in place for at least a year, and every URL’s accumulated signals have to be forwarded somehow — otherwise they’re simply lost.

Read that carefully and you’ll notice it’s conditional. Google is describing what happens when a migration is executed correctly. It is not a prediction about your migration. It’s a description of the ceiling.

What the largest dataset says

In June 2026, SALT.agency published an analysis of 1,052 domain migrations. It is, as far as I can find, the biggest study of its kind that exists.

The findings:

• Only about 23% recovered organic traffic within 90 days.
• Median time to recovery: 304 days.
• 16% took two years or more.
• 13.9% had still not recovered after three years.

One honest caveat, which SALT states themselves: the sample skews toward migrations that dropped 60% or more, because those are the ones that get analyzed. This is not a random sample of all migrations. It’s closer to a sample of migrations that went badly enough to study.

That caveat matters, and I’d rather give it to you than let you find it later. But it doesn’t dissolve the finding. It sharpens it into a different one: when a migration goes wrong, it goes wrong for about ten months. Not a quarter. Not “a brief dip.” Three quarters of paid spend covering a hole where your organic traffic used to be.

Hold the two sources side by side. Google says weeks, if you do it right. The largest body of real-world data says the median failure runs 304 days. Nobody is lying. The gap between those two numbers is entirely made of execution — and execution is the thing nobody can promise you in a pitch meeting.

Two brands that lived it

Marks & Spencer is the most useful public case, because M&S said it themselves. After a roughly £150M, two-year site rebuild, online sales fell 8.1% in the following quarter. The company attributed the drop to the new site — customers were forced to re-register, and the checkout had problems. This was a company with resources, a two-year runway, and every advantage. It still cost them a quarter of trading.

Topshop lost roughly 80% of its search visibility during consolidation into ASOS, according to SISTRIX. The identified cause was mundane and entirely preventable: wildcard redirects pointing large groups of URLs at category pages, instead of page-level 301s pointing each URL at its actual equivalent.

That second one is worth sitting with, because it’s not a strategy failure. It’s a spreadsheet failure. Somebody decided that mapping thousands of URLs individually wasn’t worth the hours.

The one artifact that prevents most of this

If you take a single operational thing from this piece, take this: the redirect map is the highest-leverage document in the entire project, and it is almost always treated as a task rather than a deliverable.

A real redirect map is a row for every indexed URL on your current site, with the exact destination URL on the new one. Not a rule. Not a pattern. A row. Wildcards are where visibility goes to die, because a wildcard says “close enough” about a page that a search engine spent years scoring individually.

What to insist on, before the build starts:

• A crawl of every indexed URL you currently have, including the long tail — old blog posts, discontinued products, PDFs, paginated category pages. The long tail is where the surprise lives, and discontinued product URLs frequently still carry links and traffic.
• A named owner for the map who is not the person doing the build. Different job, different attention.
• Redirects live for at least a year, per Mueller. Budget for that; don’t let them get cleaned up in a tidy-up sprint six months in.
• A pre-launch diff: every old URL, its status code on the new site, and its destination. Anything returning a 404 or landing on a generic category page is a decision someone should make consciously rather than by omission.
• A baseline you can measure against. Snapshot organic sessions, revenue, and top landing pages before launch. You’d be surprised how many teams can’t tell whether they’re recovering because they never recorded where they started.

None of this is exotic. It’s a week of unglamorous work that prevents the thing that costs ten months.

How to tell a normal dip from a botched one

This is the question I get asked most often, always in week three, always by someone who can’t sleep.

Normal looks like: a dip that starts within days of launch, is broadly proportional across page types, and shows steady week-over-week improvement by week four to six. Your top landing pages are still your top landing pages, ranking lower.

Botched looks like: a dip concentrated in specific page types or directories, a flat line rather than a recovery curve after six weeks, and a top-landing-pages report that doesn’t resemble the old one. Crawl errors climbing rather than falling. Pages that used to rank not appearing at all rather than appearing lower.

The distinction matters because the responses are opposite. Normal wants patience. Botched wants somebody in the redirect map today, and every week you spend waiting to see if it recovers on its own is a week added to the 304.

What this should change about the decision

Not necessarily whether you migrate. Plenty of migrations are correct.

It should change what goes in the business case. If your model shows the rebuild paying back in eight months, and the median recovery for a migration that dips is ten, your model has an entire quarter of missing traffic in it that nobody priced. Add a line for paid spend covering the organic gap. Add a line for the roadmap that freezes while the team is heads-down. Then run the payback again.

The number will be worse. That’s not an argument against doing it. It’s the difference between a decision and a hope.

And if you haven’t fully committed yet, back up one step: the complaint test is the five-minute version of this question, and the platform comparison you’ve been reading covers about 30% of the decision.

Sources

• SALT.agency. (2026, June). Analysis of 1,052 domain migrations. Sample skews toward migrations with drops of 60% or more.
• Google Search Central. Site moves with URL changes (official documentation).
• Mueller, J., via Search Engine Roundtable (2022, January) — migration processing timelines and redirect retention guidance.
• Marks & Spencer plc. Public trading statement, 2014; Econsultancy analysis of the post-relaunch quarter.
• SISTRIX. (2021). Topshop search visibility loss following ASOS consolidation.