Home › Resources › Checklists › Migration
eCommerce Migration Checklist
Replatforming fails in one specific way: the redirects. Everything else — theme, apps, data import — can be fixed after launch. A broken redirect map cannot, because by the time the rankings have gone the crawl that would have found the problem has already happened. Forty-two checks, with mapping first.
CHECKS
42 checks
TIME TO RUN
Across the whole project
LAST UPDATED
September 2026
LEVEL
Project lead with developer support
A migration is a redirect project with a website attached. Treat it the other way round and you will pay for it in organic revenue.
- Before you start
Capture the baseline first
You cannot prove a migration went well, or badly, without a record of what existed beforehand. Take it before anything changes.
This is the step most often skipped and the one you will want most. Six months after a migration, arguments about whether traffic fell are unresolvable without a pre-migration snapshot — and they will happen.
- A full crawl of the existing site, exported and saved
- Every indexed URL from Search Console, not just the ones in your sitemap
- Top 500 landing pages by organic entrances and revenue
- Current rankings for your top non-brand terms
- Twelve months of organic revenue, by month
- Core Web Vitals field data for your main templates
Save all of it somewhere that is not the old platform, because you are about to switch it off.
SECTION 01
7 checks
The redirect map
The only part of a migration that cannot be fixed comfortably after launch. Build this before the new site exists, not after.
Every old URL maps to one new URL. Anything that cannot be mapped is a decision, not an oversight.
List every indexed URL, not just the ones in your sitemap
Search Console, analytics landing pages, your crawl and your backlink data. Sitemaps routinely miss pages that rank.
Map every old URL to a single most relevant new URL
One to one wherever possible. Where it genuinely does not exist any more, map to the closest category rather than the homepage.
Never bulk-redirect to the homepage
Search engines treat a homepage redirect for a deleted page as a soft 404, and you lose the signal entirely. It is the single most damaging migration shortcut.
Use 301, not 302, for everything permanent
A temporary redirect tells search engines to keep the old URL. On a migration that is exactly wrong.
Check for redirect chains and collapse them to a single hop
Old chains from previous site changes compound during a migration. Two hops is tolerable, four is not.
Test the full map on staging before go-live
Every single URL, automated. Sampling finds the rule that works and misses the exception that does not.
Keep the map, and keep the redirects, indefinitely
Removing them a year later undoes the migration. They cost nothing to retain.
SECTION 02
7 checks
Data and catalogue
What moves across, what silently does not, and what arrives subtly corrupted.
Export products, variants, images, categories, customers and orders before anything changes
Keep the raw export. Platform exports are not always repeatable once a migration tool has run.
Check product descriptions survived formatting conversion
HTML in descriptions is where migrations mangle content. Spot-check fifty products across different categories, not five.
Confirm variant structure, SKUs and barcodes imported intact
Variant mapping is the most common data corruption in a migration, and it is invisible until somebody orders the wrong size.
Verify stock levels and prices against the source system
Compare totals, not a sample. A currency or decimal handling difference will show up in the sum.
Migrate customer accounts, and plan the password reset communication
Passwords never transfer. Customers who cannot log in on day one will simply not buy, and will not tell you why.
Preserve order history where the platform allows it
Support cannot answer questions about orders it cannot see, and returns windows do not pause for your migration.
Migrate product reviews deliberately
Reviews are a conversion asset and a structured data signal. They are also the thing most frequently forgotten until after launch.
SECTION 03
7 checks
Preserving search visibility
Beyond the redirects: the signals that have to be recreated rather than transferred.
Recreate title tags and meta descriptions, do not accept platform defaults
Most migrations silently overwrite these with a template. It is a sitewide regression that nobody notices for weeks.
Rebuild structured data and test the rendered output
Product, Breadcrumb and Review markup all have to be reimplemented. Theme documentation is not evidence; test the live page.
Preserve heading structure on templates
Themes decide H1s. A theme that puts the logo in the H1 will quietly demote every page on the site.
Keep internal linking patterns, especially collection to collection
Internal links are a ranking signal that does not transfer with a redirect. Rebuild them deliberately.
Match or improve page speed on the templates that carry traffic
A migration that makes the site slower converts the SEO risk into a conversion loss as well.
Submit the new sitemap on launch day and keep the old one available briefly
It helps search engines discover the redirects faster, which shortens the recovery window.
Confirm robots.txt on the new platform does not block anything valuable
Defaults differ between platforms. This is a thirty-second check that occasionally saves a quarter.
SECTION 04
7 checks
Commercial and operational continuity
The systems around the store that assume the old platform still exists.
Reconnect every integration: ERP, 3PL, accounting, PIM, feeds
Each one is a separate project. List them early, because the one nobody remembered is always the one that stops orders shipping.
Rebuild product feeds and resubmit to Merchant Center and any marketplace
Product IDs frequently change during migration, which orphans your entire shopping history if the feed is not remapped.
Re-create every lifecycle email flow and test each trigger
Flows do not migrate. Abandoned checkout in particular has to be rebuilt and fired at least once as a test.
Confirm tax configuration matches the old platform exactly
Tax differences surface in the first week as either angry customers or an unpayable liability.
Check discount codes, gift cards and loyalty balances transferred
Gift card balances are a legal obligation. Losing them is both a customer disaster and an accounting one.
Plan for subscription customers separately if you have them
Payment tokens rarely transfer cleanly. Get this wrong and you churn your most valuable customers in one afternoon.
Tell customer support what is changing, before it changes
They will field the confusion. Give them the password-reset explanation and the order-history situation in writing.
SECTION 05
7 checks
Launch window
The forty-eight hours either side of the switch, where attention is worth more than planning.
Do not migrate during your peak trading period
Obvious, routinely ignored. If your quarter depends on November, migrate in February.
Remove the staging noindex before go-live, and confirm the live site is crawlable
The single most common migration failure. Check the rendered robots directive on the live domain, not the setting in the admin.
Confirm the staging environment is now blocked or taken down
A crawlable duplicate of your new site competing with itself is a self-inflicted penalty.
Run the full redirect test against the live domain immediately after switching
Staging tests prove the rules. Only the live test proves the server applies them.
Place a real order on the live store within minutes of launch
Payment, tax, shipping rate, confirmation email. All four, on a real card.
Watch server response codes and error rates for the first day
A spike in 404s or 500s is your earliest signal, and the cheapest moment to act on it.
Have a rollback decision agreed in advance, with a named decision-maker
Deciding whether to roll back at 2am, without a pre-agreed threshold, produces the wrong answer.
SECTION 06
7 checks
The ninety days after
Migrations dip. What matters is whether the dip recovers, and you only know that if somebody is watching the right numbers.
Expect a temporary dip in organic traffic and say so in advance
A well-executed migration usually dips for two to six weeks. Telling stakeholders beforehand turns a crisis into a forecast.
Monitor Search Console coverage weekly for excluded and error URLs
This is where redirect gaps surface. Weekly for three months, not once after a fortnight.
Recrawl the site monthly and compare against the pre-migration baseline
Missing pages, new redirect chains and orphaned products all appear here first.
Track your top 500 landing pages individually, not just sitewide totals
A sitewide figure can look stable while your best category page has disappeared entirely.
Compare conversion rate against the pre-migration baseline by device
Migrations frequently improve desktop and damage mobile. The blended number hides it.
Fix broken redirects within days, not at the next sprint
Recovery time is proportional to how long the wrong signal was live.
If traffic has not recovered in ninety days, stop waiting
At that point it is not settling, it is a problem. Audit the redirect map against the original crawl.
- Getting it wrong
Six ways migrations lose revenue
Ranked roughly by how much damage each does.
Redirecting everything to the homepage
Treated as a soft 404, so the signal is lost entirely rather than transferred. It is the fastest way to undo years of accumulated rankings in one afternoon.
Mapping only the URLs in the sitemap
Sitemaps miss pages that rank — old blog posts, discontinued products with backlinks, legacy category paths. Use Search Console and your backlink data as well.
Leaving the staging noindex on the live site
Every platform has one. Weeks of confusion follow, and the recovery takes longer than the outage did.
Migrating before peak season
The dip is normal. Taking it in your highest-revenue quarter turns a manageable cost into a bad year.
Forgetting an integration
The feed, the 3PL, the accounting sync. Usually discovered when orders stop shipping rather than when the site stops working.
Not capturing a baseline
Without a pre-migration snapshot, nobody can prove whether it went well, and the argument about it will run for a year.
Kitchenary and its case study cover a related situation: the analysis that should happen before a platform decision, not after.
- Sequencing
Where the risk actually sits
Not everything in a migration carries the same consequence. Spend your attention in this order.
Risk
What it covers
If you get it wrong
When to do it
Critical
Redirect map, staging noindex, live crawlability
Months of lost organic revenue, and a recovery that takes longer than the mistake did
Before build completes
High
Data integrity, integrations, subscriptions, gift cards
Orders that cannot ship, customers who cannot log in, balances you legally owe
Before go-live
Medium
Titles, structured data, headings, internal links
A gradual decline that is hard to attribute and easy to blame on the dip
First fortnight
Recoverable
Theme polish, app parity, design refinements
Nothing that cannot wait. These are the things teams spend launch week on instead
After launch
The pattern every difficult migration shares: attention went to the recoverable row and the critical row was sampled rather than tested.
- What next
Where to go from here
01
STILL DECIDING
Migration is expensive enough that the choice deserves more than a feature table. The comparisons set out where each platform actually differs.
02
MOVING TO SHOPIFY
Shopify has its own structural quirks — duplicate product paths, handles, tag pages. Run that checklist alongside this one.
03
ALREADY MIGRATED, TRAFFIC DOWN
Work through the technical checks against your pre-migration crawl. Most post-migration losses are redirect gaps rather than mysteries.
04
WANT IT CHECKED PROPERLY
Five working days. Cheapest before you switch, still worth it in the ninety days after.
- Questions
Migration FAQs
What people ask before and after replatforming.
Will I lose rankings when I migrate?
Some temporary movement is normal even when everything is done correctly — usually two to six weeks. Permanent loss is not normal, and when it happens the cause is almost always the redirect map rather than the platform.
How long does a migration take?
For a mid-sized catalogue, eight to sixteen weeks from decision to launch, and the redirect mapping should start in the first fortnight rather than the last. Compressing the timeline nearly always compresses the mapping, which is the wrong thing to compress.
Should I change URLs during a migration?
Only if you must. Keeping the existing URL structure removes the largest single risk. If the new platform forces a change, that is what the redirect map is for — but do not take the opportunity to redesign URLs at the same time.
When is the best time to migrate?
Immediately after your peak, so you have the longest possible runway before the next one. For most retailers that means January to March.
What about our product reviews?
Plan the migration explicitly. They are a conversion asset and a structured data signal, they rarely transfer automatically, and they are the item most often remembered after go-live.
How long should we keep the redirects?
Indefinitely. They cost nothing to retain and removing them a year later undoes the work. Treat the redirect map as permanent infrastructure.
Traffic dropped and has not recovered — what now?
After ninety days it is a problem rather than a settling period. Recrawl, compare against the pre-migration baseline, and check every top landing page individually. The gap is usually a redirect rule that matched a pattern but missed its exceptions.
- Keep going
Related reading
COMPARISON
Where the two genuinely differ on cost, control, catalogue and operations — before you commit to a migration.
CHECKLIST
Forty-two Shopify-specific checks, worth running alongside this one if that is where you are heading.
SERVICE
How we run migrations, including the parts we insist on doing before a launch date is agreed.
Planning a replatform, or recovering from one?
The redirect map is worth a second pair of eyes before you switch, and worth auditing properly if traffic has not come back. Both are covered by the free audit.