eCommerce Data Migration: What to Verify Before You Go Live

12 minutes
, ,
Retail operations and technology colleagues reviewing laptops in a warehouse office

Three colleagues reviewing a store migration in a warehouse office

A new store can look ready while its data is still wrong. Products appear in search, customers can sign in and a test payment succeeds. Yet a trade account sees the wrong price, an imported order enters the warehouse twice, or an old customer loses access to the purchase they need to return.

For Australian and New Zealand retailers planning a replatform, the important decision is whether the evidence supports switching live trading to the new environment. A successful import is one input to that decision. The business also needs proof that the right information arrived, relationships survived and changes made during the migration were captured.

This guide sets out a practical acceptance framework for eCommerce data migration. It gives retail, finance and technology leaders a shared way to define scope, investigate differences and approve launch. If you are still choosing a platform or delivery partner, start with our broader eCommerce replatforming guide.

Key Takeaways

  • Agree which records must migrate and how each population will be accepted.
  • Reconcile identifiers, relationships and commercial meaning alongside counts.
  • Prove final changes and recovery before approving launch.

Define what must move, remain accessible or be retired

Start with business obligations and customer journeys. Identify which products, customers, companies, addresses, orders, returns, credits, gift balances, content and permissions the new operation actually needs. Record the source, destination, accountable owner and acceptance rule for each population.

Separate active operational records from historical reference material. An unfulfilled order needs an agreed fulfilment owner after launch. A completed order may only need to remain searchable by customer service. Moving both into the same destination workflow can create unnecessary processing risk.

Choose explicitly between migrating history, keeping a controlled archive and retaining read access to the old system. Check that the chosen approach supports returns, customer enquiries and your own record retention requirements. Assign responsibility for access after the legacy hosting or software contract ends.

Also document exclusions. If obsolete products are deliberately omitted, keep an approved exclusion list with identifiers and reasons. Otherwise, a lower destination count cannot be distinguished from accidental loss. Data cleansing decisions should be approved before the final rehearsal, with later changes controlled.

The output is a migration scope register that business owners can understand. Each row should say what will happen to the records, how success will be measured and who can accept an exception. This becomes part of delivery acceptance, rather than an informal assumption between teams.

Check the destination’s actual import capabilities

Migration methods have different limits. Shopify’s migration guidance distinguishes products, customers and historical orders, and recommends importing products and customers before historical orders so their relationships can be established. It lists apps and APIs for historical orders rather than a standard order CSV import.

Shopify also explains that customer passwords from another platform cannot be transferred using its customer CSV import, and order information cannot be imported through that customer file. Its customer import documentation is worth checking before promising that an account export will recreate the entire customer experience.

Ask the delivery team to demonstrate the chosen route with difficult records early. Confirm supported fields, destination behaviour and any app or API dependencies. Include customer sign-in, company membership, price access and historical order visibility in the demonstration where those are required.

Adobe’s Data Migration Tool documentation describes a specific Magento 1 to Magento 2 migration path. Its delta migration guidance warns that default handling covers data created by Magento’s own modules, with additional work required to track extension tables. That is not a general promise that every Adobe Commerce migration automatically captures every change.

Platform, deployment model, source version and extensions matter. Require a written coverage statement for your implementation. Do not treat a tool’s name or a completed demonstration of simple records as proof of complete migration coverage.

Reconcile products by identity and meaning

Merchandising manager comparing industrial fittings with a laptop catalogue

Compare products using stable identifiers and an explicit mapping between old and new IDs. Display names are unreliable matching keys because they can change or be duplicated. Preserve source references where customer service, integrations or historical order lines will need them.

Count products and variants separately. One configurable product with several sellable variants is not equivalent to the same number of independent products. Check parent relationships, options, SKU uniqueness, category assignments and publication status in each intended channel or market.

Then compare meaning. A weight imported into the wrong unit can produce an incorrect shipping rate even when the field contains a valid number. A pack price assigned to a single item can look like a successful price import. Include units of measure, pack quantities and currency in the comparison rules.

A useful reconciliation report records the source value, expected destination value, actual destination value and reason for any transformation. Where the new catalogue intentionally combines or splits records, compare against the approved transformation rather than demanding identical counts.

Check completeness as well as numerical values: images, documents, product attributes and relationships all matter to the buying journey. Our guide to eCommerce product data explains the catalogue information that supports product discovery and purchase decisions. Migration acceptance should verify that the information reaches the correct products.

Validate customers, permissions and commercial rules

Matching an email address does not establish the right trade relationship. A buyer may belong to a particular company or branch, have restricted purchasing permissions and receive a negotiated catalogue. Test those relationships using named scenarios approved by the sales or account management team.

Include shared inboxes, duplicate profiles, inactive accounts and customers with multiple delivery addresses. Agree which record wins when the source contains conflicting information. Preserve the explanation so customer service can resolve questions after launch without reconstructing a developer’s spreadsheet.

For Australian and New Zealand operations, test each relevant storefront and currency independently. Use approved expected tax and price results for your actual trading scenarios. Include addresses with leading-zero postcodes and local characters so formatting changes do not quietly corrupt delivery or customer details.

Check marketing preferences separately from account existence. An imported customer record should not be treated as evidence that every marketing channel is permitted. Preserve the source preference and available provenance through the mapping, then have the responsible business owner verify the resulting settings.

A price test should identify the buyer, company, delivery location, product, quantity, currency and applicable date. Compare the expected price with the catalogue, cart and order outcome. Where pricing comes from an ERP, prove that the new store is requesting and applying the correct rule, rather than only copying a historic price field.

Prove order history without restarting old transactions

Two retail office colleagues checking printed order records together

Order migration has two distinct questions: can people use the history, and can the business safely continue outstanding work? Answer both. Check customer association, original reference, line items, quantities, discounts, tax, shipping, currency, payment status, fulfilment status and refunds where included in scope.

Group financial comparisons by currency and the same agreed reporting period. Adding AUD and NZD totals together creates a meaningless control. Document whether a comparison uses original order value, remaining balance or net value after refunds, so teams do not argue over totals calculated on different bases.

For example, consider an illustrative order with two lines, one already dispatched and one awaiting stock. The acceptance evidence should show where the remaining line will be fulfilled, how customer service sees the earlier dispatch and how the order avoids being sent to the warehouse as entirely new work.

Review import side effects in a controlled environment. Confirm whether the selected method sends emails, changes inventory, triggers an ERP export or starts downstream automation. Historical records need a deliberate processing policy. A technically correct order record can still cause an operational mistake if a live integration treats it as a new sale.

Keep payment references and business references where required, but do not assume a migrated paid status proves a new payment occurred. Finance should verify the relationship to the original transaction and agree the treatment of open payments, refunds and credits with the relevant providers and systems teams.

Capture the changes made while migration is underway

A rehearsal export is a snapshot. Trading may continue afterwards: customers register, products change, orders arrive, returns are processed and stock moves. The final migration plan must explain how each relevant change reaches the new environment without being missed or applied twice.

Use a change register showing the data population, capture mechanism, final cutoff and reconciliation owner. Some changes may be transferred incrementally. Others may require a short editing freeze or a fresh export. Choose the method according to what the source and destination actually support.

Record timestamps with their timezone and define the exact boundary between batches. This matters when Australian and New Zealand teams are working in different local times. An instruction such as ‘export everything since midnight’ is insufficient without a timezone and a clear rule for records changed at that boundary.

Test an update to an existing record as well as the creation of a new one. Include cancellations, deletions or deactivations where they affect the destination. A process that captures only newly created customers will not necessarily capture an address correction or revoked access.

At final cutover, record the last accepted source change, the last applied destination change and the remaining exceptions. Our article on ERP integration failures and recovery explains why replay and reconciliation must account for actions already completed in the destination.

Rehearse the whole business journey

Warehouse supervisor and colleague checking a picked order at a packing station

Use full-population comparisons for fields that can be checked reliably in bulk, then add representative journey tests. Sampling alone can miss a whole excluded customer group. Counts alone can pass while the wrong prices or relationships have been loaded.

Choose scenarios by consequence. Include a high-volume product, an unusual variant, a negotiated trade price, a multi-address company, a partially fulfilled order and a customer who needs historical information. Add your known edge cases rather than selecting only records that are easy to demonstrate.

Run the rehearsal through the actual systems that will support launch, with safe controls around payments, customer communication and fulfilment. Have merchandising, finance, customer service and warehouse owners review the results relevant to them. Record the evidence, defect and decision against each scenario.

Measure the complete rehearsal duration, including reconciliation and corrections. Import speed alone does not establish a viable launch window. If the business cannot complete validation before trading resumes, reduce uncertainty earlier, change the cutover approach or reconsider the window.

Make the launch decision explicit

Agree launch blockers before the final migration. Unexplained missing active orders, incorrect customer price access and duplicated downstream processing are examples of issues that can justify stopping. The exact thresholds must reflect your operation and be accepted by the responsible owners.

Keep a single exception register with identifier, impact, owner, workaround, expiry and retest evidence. A cosmetic product description issue may be acceptable with a dated repair plan. An unresolved permissions problem affecting contract prices needs a different decision. Avoid hiding both inside a single percentage-complete figure.

The go/no-go meeting should review the scope register, reconciliation results, critical journeys, final changes and recovery plan. Give one named business decision-maker authority to approve launch, supported by explicit acceptance from the relevant operational owners.

Recovery must account for trading after the switch. Restoring an old site does not automatically move new orders, customer updates or payments back into it. Decide how those records will be preserved, reconciled and processed before treating rollback as a workable option.

Buy evidence and accountability from your delivery partner

Ask prospective partners to price the migration obligation with its scope, mappings, rehearsals, reconciliation reports, correction process and support responsibilities. Comparing only the cost of importing records leaves the most consequential work undefined.

Retain ownership of the mapping documents, exception register and final acceptance evidence. These records help your internal team explain discrepancies and support the platform after the project team changes. Agree access to any migration tooling or logs needed during stabilisation.

OSE’s commerce strategy, replatforming and systems integration services are relevant where data crosses the storefront, ERP and operational workflows. A useful engagement should turn those requirements into an agreed migration plan and verifiable acceptance evidence. The value depends on the delivered controls and accountable decisions, rather than a promise that changing platforms will automatically improve results.

If your migration is approaching launch, talk to OSE about validating the migration scope and cutover plan. Bring a source-system inventory, known data issues, proposed launch window and examples of the customer journeys the new store must preserve.

Frequently asked questions

How do we know an eCommerce data migration is complete?

Compare the agreed source population with the destination using approved mappings and transformations. Resolve unexplained omissions, duplicates and field differences, then verify critical customer and operational journeys. Completion also requires evidence that relevant changes made after the first export were captured.

Must every historical order move into the new platform?

No. Decide which records must remain operational and which only need reliable reference access. A controlled archive or retained system may serve some history, provided it meets your business and record access requirements. Agree ownership, access and cost before retiring the old environment.

Can customer passwords be migrated to Shopify through CSV?

Shopify states that passwords encrypted outside Shopify cannot be migrated through its customer CSV import. Plan and test the sign-in experience supported by the chosen customer account setup. Communicate any action customers must take, and verify their access to the appropriate company and order information.

Can we keep taking orders during migration?

Possibly, if the migration design captures new records and relevant updates reliably. Confirm the supported incremental process and reconcile the final boundary. Some activities may need a temporary freeze. Test the approach in rehearsal rather than assuming an import tool captures every change automatically.

Who should approve migration data before launch?

The business should name a go/no-go decision-maker, supported by owners of catalogue, customer, finance and fulfilment data. The delivery team supplies evidence and resolves defects. Acceptance should identify unresolved exceptions and their commercial consequences, with responsibility for monitoring and correction after launch.