9 min read
In brief
- A migration moves three things at once: the domain, the approval status inside Google Merchant Center, and the conversion signal your ad platforms have learned. The domain is the only one you can see.
- The predictable failure is sequencing — the new store is finished, the domain is switched, and Google, Meta and your feed all discover a wholesale change on their own schedule.
- Two identifiers must survive untouched: the product IDs your feed and ads match on, and the purchase event names your tracking sends. Change either and history resets silently.
Three assets move at once, and only one is visible
Ask a store owner what a migration involves and you get a list of pages, a theme and an app stack. That is the visible part. Underneath it, three separate things are moving, and two of them are not files.
The first is the domain and everything attached to it: canonical URLs, backlinks, redirects, the addresses inside your ads, and the domain your customers recognise in their inbox.
The second is the approval inside Google Merchant Center. Approval is not a property of your products; it is a judgement about the site behind them. A new site is a new judgement, even when every product is identical.
The third is the conversion signal. Meta and Google have spent months learning which clicks produce purchases at your store. A migration that changes URLs, event names or firing order resets part of that learning, and neither platform tells you it happened.
- Domain equity: mostly recoverable with correct redirects, never fully recoverable.
- Merchant Center approval: recoverable — but only if you decide when the re-review happens.
- Conversion signal: partly recoverable; what resets appears in no dashboard you have.
First, decide which migration you are actually doing
Four different jobs get called a migration, and they carry very different risk. Naming the job before you start is what stops the scope growing halfway through.
| What changes | Merchant Center risk | Tracking risk | Typical window |
|---|---|---|---|
| Theme or design only — same platform, same URLs | Low: no re-review expected | Low, if the same pixel calls survive the theme | 2–4 days |
| Platform change, same domain | High: a new site behind an approved account | High: cart and checkout events are usually rewritten | 2–3 weeks |
| Domain change, same platform | High: every landing URL in Merchant Center and in ads changes | Medium: cookies and click IDs reset on the new domain | 1–2 weeks |
| Catalogue restructure — new product IDs, new URLs | Very high: feed matching breaks unless IDs are mapped | High: item-level optimisation history resets | 3–6 weeks |
The one invariant
Whatever the job, the product IDs stay the same. They are the join key between your catalogue, your Merchant Center feed, your Shopping ads and every performance number you have. A migration that regenerates them is not a migration; it is a new store wearing the old store's name.
The order of operations
- 1
Audit while the old store is still the reference
Export the URL inventory, every live redirect, the feed's product IDs and titles, the event table from checkout, and a list of every domain that appears as a destination in your ads. This is the only moment the old store is still the truth.
- 2
Freeze the catalogue
No new products, no price changes and no title edits for the week either side of the switch. Two variables changing at once is how a feed problem becomes unattributable.
- 3
Build and test on a staging host
Staging must be noindexed and disallowed in robots. A test store that gets indexed is a duplicate-content problem you will be cleaning up for months.
- 4
Prove the tracking before the domain moves
Place a test order on staging and confirm the purchase event, its value, its currency and its event id reach the platforms. If that has not happened by day seven, the migration is not ready.
- 5
Move the domain and the redirects in one deploy
One-to-one 301s from every old URL to its closest equivalent, live in the same deploy as the domain change — not a week later, when Google has already crawled the 404s.
- 6
Update the feed deliberately
Change destination URLs and image links in a single scheduled upload, then request a Merchant Center review rather than waiting to be crawled at a moment you do not control.
- 7
Verify for 72 hours, then tell the campaigns
Orders, events, redirects and feed status checked across three days before anyone is allowed to increase spend against the new store.
Merchant Center: how an approved account loses its approval in a migration
Most post-migration suspensions we see are not policy decisions about the products. They are the account being re-reviewed against a site that was not finished when the crawl happened: placeholder policy pages, a checkout missing the contact block it used to have, an about page that now 404s.
The mechanism is simple. Google periodically re-crawls the destinations of your feed items. A wholesale change to those destinations — new templates, new policy URLs, new business information — is exactly the kind of change that invites a review. You cannot prevent the review; you can choose the week it happens, and make sure the site is finished when it does.
There is a quieter version too: the feed keeps pointing at URLs that now redirect. Some redirects are followed silently and some are not, and the difference surfaces as a subset of items disapproved for a reason that reads like policy and is really a URL.
The sequence that holds
Finish the new store completely — policies, contact details, business information, shipping and returns text — before the domain moves. Then move the domain and update the feed in the same window, and request the review yourself. The version that fails is switching the domain on Friday and updating the feed the following Tuesday.
Keeping conversion tracking continuous
A migration rewrites the path a purchase takes to reach your ad platforms, and it does so at the exact moment you have no baseline to notice. The signal rarely breaks loudly; it thins out.
Two structural risks do the damage. Cookies are scoped to a domain, so a new domain starts with no click history and no fbc value from the old one. And if the new checkout sends Purchase with a different value, a different currency or a different event name, the platform treats it as a different conversion rather than a continuation of the old one.
- Keep the same pixel ID, the same GA4 property and the same conversion actions. New IDs mean new history.
- Keep the same event names and the same currency. A purchase reported in another currency is a new conversion, not the same one.
- Send the purchase server-side from the order record as well, so a migration that breaks the browser path does not erase the month.
- Write the numbers down before you move: purchases per day, revenue per day, events per session. Without a before, there is no after to compare.
Why this matters more than it used to
In our own browser audit of 47 actively-advertising storefronts, 51% fired no Meta pixel within three seconds of load, and only 6% showed any detectable server-side path. A store measuring only in the browser has tracking a migration can break silently, with no second source that would reveal it. A migration is the moment that gap becomes expensive.
Redirects, and the traffic you only think you have
A redirect map is not a task; it is a test suite. “Old product URLs 301 to new product URLs” is the easy half. The half that decides whether your traffic survives is query strings, paginated category pages, and the URLs that exist only because an app created them.
| Old URL pattern | Should go to | Common mistake |
|---|---|---|
| /products/<handle> | The same product on the new store, one to one | Redirecting everything to the homepage, which Google treats as a soft 404 |
| /collections/<handle>?page=2 | The equivalent page of the new category | Dropping the query string, so paginated pages collapse into one |
| /pages/<app-page> | The closest new page, or 410 if it genuinely no longer exists | Chaining redirect to redirect, which wastes crawl budget |
| Cart and checkout URLs | Not redirected — they are session URLs | 301ing /cart, which breaks returning carts |
Test the map rather than building it
Crawl the old site first and keep the list. After the switch, fetch every old URL with a script and assert a 200 at the destination in one hop. A redirect to a redirect, or a 200 on a page whose title does not match the old one, is a bug — and it will be invisible in analytics for weeks.
What resets no matter how careful you are
- Ad platform learning: a purchase stream that pauses or changes shape costs days of optimisation, and no setting prevents it.
- Platform-native reviews: reviews attached to an app or a marketplace do not travel. Decide how you replace them before you migrate, not after.
- Referral traffic that depends on a specific URL staying put — affiliate links, influencer posts, old email campaigns. Each one is a manual update.
- If the domain changes: age, backlink profile and the association between the name and the category all stay attached to the old name.
Knowing this list in advance is what separates a migration with a dip from a migration with a mystery. Every item on it should have an owner and a date before the switch.
The fourteen-day plan on one page
| Day | Work | Done when |
|---|---|---|
| 1–2 | Inventory: URLs, redirects, feed IDs, event table, ad destination domains | You hold a list you can test against |
| 3–5 | Build on a noindexed staging host | Every template renders and every policy page exists |
| 6–7 | Tracking on staging: test order, purchase event, value, currency, event id | The event is visible in the platform, not only in the console |
| 8 | Freeze the catalogue and the ads; snapshot the baseline | Purchases, revenue and events per session written down |
| 9 | Domain move and redirect map in one deploy | Every old URL resolves in a single hop |
| 10 | Feed URLs updated, review requested | Feed accepted with no URL errors |
| 11–14 | Verify orders, events, redirects and feed status | 72 hours with no unexplained gap |
What to check before you tell anyone it is live
- Place an order and find its purchase event in Meta and in GA4, with the right value and currency.
- Fetch 200 old URLs and confirm a one-hop destination that matches the old page.
- Confirm the feed has no URL-related disapprovals and that item IDs are unchanged.
- Confirm the sitemap, canonicals and robots point at the new domain, never at the staging host.
- Search for the staging host before Google finds it — a stray indexed copy is a duplicate-content problem.
- Compare day one with the baseline. A 20% fall in event volume with flat orders is a tracking bug, not a traffic problem.
If a migration is coming, we audit the current store first — URLs, feed IDs and tracking events — and hand over the redirect map and the switch order before anything moves. An audit costs a fraction of doing the migration twice.
