Ask any prospective Shopify migration agency for the field mapping sheet before you ask for the price. Not a proposal, not a Gantt chart β the actual document that says which field in your current database lands in which field in Shopify, and what happens to the ones with nowhere to go. Agencies that run a plugin and call it a migration cannot produce that document, because the plugin decided for them and nobody wrote it down. That single request sorts the market faster than any reference call.
We've picked up enough rescue jobs from plugin-only migrations to know the pattern. The catalogue looks fine on launch day. Three weeks later organic traffic is down a third, the accounts team can't refund an order, and 4,000 old category URLs are 404ing. Below are the eleven questions we'd ask if we were buying the work instead of doing it.
First, the honest bit about migration tools
Using a tool is not the problem. We use Matrixify on nearly every job, plus custom scripts against the source database and the Admin API. Bulk CSV work is faster and more auditable than hand entry, and anyone who claims to type 8,000 products in by hand is either lying or overcharging.
The problem is when the tool is the method. A one-click connector reads your products, guesses at Shopify's schema, and writes what fits. Everything that doesn't fit is silently dropped: custom attributes, tax classes, tier pricing, SEO titles, category page copy, image alt text, order line-item metadata. Nobody chose to drop them. That's the difference. A real migration is a series of decisions someone made on purpose and can explain to you a year later.
Questions 1β3: the data model
1. "Can I see the field mapping document for my data, before we sign?"
Two or three pages, source field on the left, Shopify destination on the right, and a third column for the awkward ones β metafield, tag, dropped, needs a decision from you. If they offer to build it after the deposit, fine, but ask for a sample from a past project with the client name stripped out. If nothing exists, they're a plugin shop.
2. "Which parts are scripted, which are manual, and who does the manual part?"
Every migration has residue. On a WooCommerce store it's usually the product descriptions with hardcoded shortcodes and inline table markup. On Magento it's the attribute sets. Ask for the estimated row count of the manual pile and the hourly rate attached to it. An agency that has done this work knows the pile exists and has priced it. One that says "everything is automated" is going to send you a change request in week three.
3. "How will you handle products that don't fit Shopify's option structure?"
This is where plugin migrations quietly mangle catalogues. Say you sell trousers in 6 waist sizes, 5 colours, 4 fabrics and 3 inseams. That's 6 Γ 5 Γ 4 Γ 3 = 360 sellable combinations across four option axes. Shopify gives you three option axes per product. The variant ceiling itself has moved upward on newer plans; the three-axis structure has not. So somebody has to decide: split into four products by fabric, move inseam to a line-item property, or rebuild it with a variant app. A connector will just import the first three attributes it finds and drop 270 SKUs on the floor.
Same question applies to bundles, configurable/grouped products, and anything with a per-customer price.
Questions 4β5: customers and orders, where the surprises live
4. "What happens to customer passwords?"
Nothing. They don't migrate. Password hashes can't be moved into Shopify, so every returning customer either resets or accepts an account activation invite. That's a fact of the platform, not an agency failing β but the agency should raise it before you do, and should have a plan: staged invite sends by segment, a first-login discount to make it worth the customer's trouble, and a re-order flow that works for logged-out visitors. If they haven't mentioned it, they haven't done many migrations.
5. "Will the imported orders actually be usable?"
Ask this one slowly and watch the face. Orders created through the Admin API come in without gateway transactions attached. They show in the timeline, they're searchable, they feed your LTV reports β and you cannot refund them through Shopify. Refunds for pre-migration orders go back through your old gateway's dashboard. Same for GST credit notes if your invoicing app reads Shopify transaction data.
Decide the cut-off with your finance team, not your developer. A common answer for Indian D2C brands: import 24 months of order history for reporting, keep the old admin read-only for 90 days to cover the return window, and brief customer support on which system to use for what.
Question 6: the redirect map is most of the SEO job
6. "Who writes the 301 map, and how do you handle categories that collapse?"
Numbers make this concrete. A mid-size store we scoped recently had roughly 9,400 URLs in Search Console: 5,100 products, 320 category pages, 480 blog and content pages, and the rest faceted or paginated variants of those. SKU matching resolved about 4,870 products automatically. That left roughly 230 rows needing a human β discontinued items, merged duplicates, bundles that became a single product. At 45 rows an hour that's five hours of genuinely dull work. Five hours. The plugin does none of it, and those 230 URLs are disproportionately the ones with backlinks.
The faceted URLs are the second trap. Old platform filter URLs frequently rank. In Shopify they'll usually collapse into a single collection with filters applied client-side, so you need a deliberate decision per pattern: redirect to the parent collection, or build a real collection with its own handle and copy. Ask which patterns they'll rebuild and which they'll fold. If the answer is "we'll redirect everything to the homepage", walk.
We cover the full sequence in our Shopify migration process, including the two-week window after cutover when you watch server logs rather than dashboards.
Question 7: the dry run and the freeze window
7. "How many dry runs, and what's the delta plan for orders placed during cutover?"
One full dry run into a development store, reviewed by your merchandiser and your accountant, then a second run to validate fixes. Then the final pass on cutover day with a delta sync for anything that came in while the freeze was on. Ask for the freeze duration in hours and what happens to a COD order placed 20 minutes before DNS flips. Ask about DNS TTL β dropping it to 300 seconds a day ahead is basic hygiene, and it tells you whether they've done a cutover under pressure.
Also ask for the rollback condition. Not "we'll fix it", but the specific trigger and the specific steps.
Question 8: rebuilding what Shopify doesn't ship
8. "What replaces our layered navigation, and how do filters get their data?"
Magento merchants feel this hardest. Layered navigation off attribute values is native there. In Shopify, filters read from product options, tags and metafields β which means the migration has to write clean, consistent metafield values, not free-text tags with three spellings of "stainless steel". Get the metafield definitions agreed before the import runs, because retrofitting filter data across 5,000 products is a second project.
For large catalogues, native filtering runs out of road quickly on synonyms, multi-select behaviour and search relevance. That's the point where a filter and search app earns its keep, and it's why we built FilterPro. Whatever you use, decide during the migration, not after.
Question 9: subscriptions and saved cards
9. "What's the plan for active subscribers?"
Card tokens belong to the gateway that created them. Vault transfers between gateways are sometimes possible with PCI-compliant handling and both parties' approval, but in India, RBI card-on-file tokenisation rules make card portability a non-starter in practice. Recurring plans usually mean re-authorising every subscriber through a new UPI or e-mandate flow.
That's a churn event, and it needs a communications plan with a deadline, incentives and a fallback for the ones who ignore three emails. An agency that shrugs at this question has never migrated a subscription business. Get the expected drop-off in writing before you commit to a date.
Questions 10β11: the two you ask about people
10. "What do you already know will be broken on day one?"
Every migration launches with a known-issues list. Ours usually includes a handful of legacy blog posts with broken embeds, a report that needs rebuilding, and one integration that only fails under real load. An agency that answers "nothing" is either inexperienced or managing you. The useful answer names three or four things and says when each gets fixed.
11. "Who is actually on this, and who's reachable in the 72 hours after cutover?"
Names, roles, time zones. Whether the developer on the call is the developer on the keyboard. Whether support after go-live is a shared inbox or a person on WhatsApp at 11pm on a Saturday during festive season. If you're comparing an in-house team against a contracted build, our notes on how to hire Shopify developers in India cover what the trade-off actually costs.
What an Indian-market agency should raise before you do
Unprompted, without you asking: HSN codes per product as a metafield so GST invoicing apps have something to read. Tax overrides for the categories where the rate isn't standard. COD as a real payment method with partial prepaid options and an RTO plan, not an afterthought at checkout. Razorpay or Cashfree configured with the right webhooks so failed-then-succeeded payments don't create duplicate orders. Shiprocket or Delhivery order sync tested with a real pickup, not a test order. FSSAI licence numbers on product pages if you sell food.
If a quote for an Indian D2C migration doesn't mention GST or COD anywhere, it was written for a US store. That also affects the number at the bottom β we break down what drives it in our guide to Shopify development costs in India.
How to read the answers
You're not scoring for polish. You're scoring for specificity. The strong answers contain numbers, named constraints and at least one thing you didn't want to hear. The weak ones contain the word "seamless".
One more filter: ask what they'd charge to do half the job β data only, you handle the theme. A shop that does real migration work can quote that cleanly, because data is a discrete workstream with its own deliverables. A shop that sells a bundled rebuild-plus-migration and can't separate them is telling you the migration was always the thin part.
If you want a second opinion on a quote you're holding, send it over with your platform and rough SKU count and we'll tell you what it's missing. That's what our free audit is for. No obligation to hire anyone, including us.