Contents
An established retailer wanted to move from Magento 2 to Shopify without losing a product catalog built up over years, or its Google visibility. Instead of simply copying the data from one system to another, we built a universal product database as the single source of truth, synced it bidirectionally with Shopify and put AI agents for enrichment, analysis and SEO on top. In this post we walk through the architecture step by step, the decisions behind it, the SEO side of the migration and 12 points you should settle before your own move.
Whether switching to Shopify pays off for your store, and from which system, is covered in our overview Switching to Shopify: migrating from Magento, WooCommerce, Shopware & Co..
The key idea first: a shop migration is really a data migration. If you structure the product data properly, you can switch shop systems. If you only switch shop systems, you take every data problem with you.
Starting Point: A Capable System with Grown Data
Magento 2 is a very flexible system. It allows custom attribute sets, complex product types, multiple store views and deep customization. That flexibility has a flip side that has nothing to do with the software itself: hosting, security updates, version upgrades and extensions have to be run in-house, and over the years the data model drifts a little further with every employee and every season.
That was the case here too. Attributes existed several times under slightly different names, units were recorded inconsistently, product copy varied in completeness, and categories had evolved historically. Every product change took time, and data quality depended on who had created the product. In grown catalogs, this is the rule rather than the exception.
The goal was therefore twofold from the start: move operations to a SaaS platform, and at the same time organize the product data so that it is no longer tied to a single system.
The Central Decision: Database First, Then the Shop
There are two basic routes for a Magento to Shopify migration.
| Direct migration | Migration via a product database | |
|---|---|---|
| Process | export from Magento, transform, import into Shopify | export from Magento into your own database, clean up there, sync from there to Shopify |
| Upfront effort | lower | higher, because data model and sync have to be built |
| Data cleanup | once, in the export script | permanently, in a dedicated system |
| Future channels | each channel connected separately | channels pull data from the same source |
| Next platform switch | another project | a new sync, not a new project |
We chose the second route. The reason: after the migration, the catalog should not just have moved, it should be maintained better for good, and the AI agents needed a place to work without writing directly to the live shop. A dedicated product database does both.
The second question was whether an off-the-shelf PIM or a lean custom database fits. An off-the-shelf PIM brings ready-made interfaces and workflows. A custom database can be tailored exactly to the catalog and to access by agents, but someone has to run and develop it. Every company should make this trade-off for itself. What matters is catalog complexity, the number of channels, in-house technical skills and how much automated enrichment is planned.
The Architecture Step by Step
| Building block | Role |
|---|---|
| Universal product database | system-independent single source of truth for all product data |
| Mapping & normalization | unifies attributes, units and categories |
| Bidirectional Shopify sync | keeps database and shop consistent without duplicate effort |
| AI enrichment agents | add copy, attributes, metadata and internal links |
| Analysis layer | finds gaps, duplicates and SEO potential in the catalog |
1. Extraction from Magento. We export not just products, but everything needed later: attribute sets, options, categories with descriptions, images with alt texts, metadata, store view translations and the URL rewrites, meaning the list of all old URLs including historical redirects.
2. Mapping and normalization. This is where most of the work lies. Duplicate attributes are merged, units unified, values moved into fixed lists. Magento product types are mapped to the Shopify model: configurable products become products with variants and options, additional properties become metafields or metaobjects, categories become collections. It pays to decide which collections are curated manually and which are filled automatically by rules.
3. The product database. It holds the canonical version of every product, in a model that mirrors neither Magento nor Shopify, but the catalog. Every field has a status (draft, proposal, approved) and an origin (import, human, agent, manufacturer data).
4. Bidirectional sync. The database writes to Shopify via the GraphQL Admin API. Shopify designated the REST Admin API as legacy in October 2024, so new integrations should be built on GraphQL. Bulk operations suit the initial load, webhooks the ongoing operation. Shopify guarantees neither the order nor single delivery of webhooks and explicitly recommends additional reconciliation jobs. Our sync therefore treats webhooks as the fast path and a periodic reconciliation as the safety net.
5. Field ownership instead of “last write wins”. Exactly one system is the authority for each field. Product copy, attributes and SEO fields belong to the database; stock levels belong to the system where they originate. A conflict is therefore not resolved but prevented. On top of that comes a loop guard: a change triggered by the sync itself must not come back as a new change.
6. Agents and approval. The agents read and write exclusively in the database, never directly in the shop. Their results are proposals with an origin.
Agent-First: Enrichment, Analysis, SEO
On top of the database, specialized AI agents do the work that otherwise costs countless clicks in a grown catalog:
// Every product runs through the pipeline, approval rules decide on the sync.
async function processProduct(product: RawProduct) {
const normalized = normalize(product); // unify attributes & units
const enriched = await enrichWithAI(normalized); // copy, missing attributes, metadata
const seo = await optimizeSEO(enriched); // title, description, internal links
const issues = analyze(seo); // gaps, duplicates, inconsistencies
const needsReview = issues.length > 0 || hasNewCopy(seo); // approval rules
return needsReview ? flagForReview(seo, issues) : markForSync(seo); // into the database, never straight into the shop
}
The enrichment agent writes product copy following a style guide, proposes missing attributes and creates metadata and alt texts. The analysis agent checks the entire catalog for gaps, duplicate content and contradictions, such as a weight that does not match the variant. The SEO agent handles titles, descriptions, collection copy and internal linking.
Three rules keep the system reliable:
- Write the copy, never invent the facts. An agent may write copy, but must not generate technical values. Dimensions, materials, standards and compatibilities come from manufacturer data or from a human, with a source.
- Approval rules instead of gut feeling. It is defined upfront which changes may go into the sync without findings and which a human always sees. New product copy and anything flagged by the analysis belong in the second group.
- Every change is traceable. Who or what changed a field and when is visible in the database and can be undone.
It is the same division of labor as in any AI agent system in the enterprise: volume automated, judgment human.
SEO During the Migration: Taking Rankings Along
The most delicate part of any shop migration is not the catalog but visibility. Google has learned every old URL. If one disappears without a redirect, its ranking disappears with it.
Shopify's URL structure is fixed. Products live under /products/handle, collections under /collections/handle, content pages under /pages/ and blog posts under /blogs/blogname/. Only the handle can be chosen. Magento URLs with a .html suffix or a category path in the product link therefore cannot be rebuilt one to one. A custom structure would only be possible headless.
The redirect mapping is mandatory. Every old URL with traffic, rankings or backlinks gets a 301 redirect to the matching destination: product to product, category to collection, CMS page to page. The basis is a list from several sources: a crawl of the old store, Search Console, analytics landing pages, backlink data and the URL rewrites from Magento. Redirect chains from the Magento era are resolved to the final destination.
Know how Shopify redirects behave. Redirects are managed in the admin and can be imported via CSV. They only take effect when no page exists at the old URL anymore. URLs with parameters, such as filtered category pages, cannot be redirected reliably. Here you need to decide which filter pages really have search value and whether they should be recreated as their own collection.
Canonicals and internal links. Shopify can also serve products under a collection path, in which case the canonical points to the /products/ URL. Internal links in the theme should point directly to that canonical address where possible, so Google does not get contradictory signals.
Content parity. Titles, meta descriptions, category copy and alt texts are carried over, not reinvented. A collection without the copy that made the old category rank is a weaker page.
Structured data. Many themes output only basic product data. Brand, GTIN, ratings and availability belong fully in the JSON-LD, fed from the metafields. Duplicate markup, for example from a review app in addition to the theme, should be avoided. Validation is done with the Rich Results Test.
Go-live and after. Before switching over, every redirect is tested with a crawler against the mapping list. After that, submit the sitemap Shopify generates automatically in Search Console and closely watch 404 errors and indexing over the first weeks. Short-term fluctuations are part of every migration, but mapping errors should be fixed within days, not months.
Clean, complete product data also pays off where Google AI Overviews and AI assistants answer product questions: they build their answers from structured, unambiguous information. More on this in Generative Engine Optimization (GEO).
The Results
- Enormous time savings in product maintenance: enrichment and SEO run automatically, and the team approves instead of typing it all out.
- Consistent data quality across the entire catalog: uniform attributes, complete copy, clean metadata.
- A faster frontend and lower maintenance effort after the move to Shopify.
- Future-proof: marketplaces and new channels pull their data from the same source.
What We Learned
The following points are typical pitfalls in migrations of this kind. We now plan for them from day one.
Data cleanup is the project. The sync and the agents are visible and exciting. But most of the work is attributes, units and categories that have drifted apart over the years.
Check variants first. Shopify's options and variants model differs from Magento's. Since October 2025, Shopify allows up to 2,048 variants per product, while the number of options is still limited. Whether your catalog fits belongs at the start, not the end.
Redirects are not a go-live detail. Whoever builds the mapping in the last week builds it incompletely. The URL list belongs in the first project phase.
Without field ownership, bidirectional turns chaotic. A two-way sync without clear rules about who owns which field will sooner or later overwrite good data with old data.
Agents are only as good as their sources. Without manufacturer data or internal knowledge, an enrichment agent produces plausible copy with unverified details. The “phrase yes, invent no” rule has to be enforced technically.
Customer accounts need communication. Passwords cannot be migrated; after the move customers sign in with a one-time code sent by email. Without advance notice, that looks like an error.
A content freeze before switchover. Anyone who keeps changing products in the old store until the last day loses those changes or has to maintain them twice.
The team has to learn a new way of working. From now on, maintenance happens in the product database, not in the shop admin. That is a change that needs training and clear responsibilities.
Magento to Shopify Migration: 12 Points Before You Start
- Clarify the goal: just a platform switch, or also better product data and new channels?
- Catalog inventory: product types, attribute sets, variants, languages, data quality.
- Check the variant model: does every product fit Shopify's options and variants logic?
- Define field ownership: which system is the authority for which field?
- List special features: which Magento extensions and custom developments are still needed, and are there apps or clean custom solutions for them?
- Capture integrations: ERP, inventory management, marketplaces, accounting, shipping.
- Payments, shipping, VAT: check early which payment methods your customers in Switzerland expect, plus shipping rules and taxes.
- Build the URL inventory: crawl, Search Console, analytics, backlinks, Magento URL rewrites.
- Create the redirect mapping: every relevant old URL to a matching destination, without chains.
- Secure SEO parity: titles, descriptions, category copy, alt texts, structured data.
- Customers and history: accounts, order history, gift cards, reviews, plus an announcement of the new one-time-code sign-in.
- Go-live plan: content freeze, redirect test, sitemap, monitoring of the first weeks, clear owners.
What Drives the Effort
Concrete prices cannot be stated seriously without knowing the catalog. The drivers, however, are easy to anticipate:
- Catalog complexity: number of attribute sets, product types and variants.
- Data quality: the further the data has drifted, the more work the normalization takes.
- Languages and markets: every store view with its own translation and URLs adds to mapping and testing.
- Integrations: every connected system brings its own interfaces and its own field ownership.
- Special features: whatever Magento did custom has to be solved anew in Shopify.
- URL volume and history: many old URLs and redirect chains extend the mapping.
- Depth of enrichment: whether only gaps are filled or the whole catalog is rewritten.
- Theme and design: a standard theme with adjustments or a custom frontend.
In Brief
The migration was never the goal, but the by-product of a better data structure. The product database holds the truth, Shopify is the frontend, agents take over the routine, and people decide what goes into the shop. SEO care during the move decides whether the visibility built up over years comes along.
How a catalogue can be built without retyping from the start, even without a migration, is covered in creating Shopify products automatically instead of by hand.
How we set up data flows like this in other areas is shown in Process automation for Swiss SMEs. How we build and migrate Shopify stores is on the page of our Shopify agency, the technology behind it on the Systems & AI Automation service page, and how we keep growing visibility afterwards on our SEO agency page.
Are you planning a shop migration or want to automate your product maintenance? Book an introductory call.
