How to Prepare Your Product Data for a Platform Migration

Product data prep is the highest-leverage work in any platform migration: clean SKUs, consistent variant logic, a defined metafield schema, and a categorization plan — all settled before anything gets imported. In our experience, dirty product data is the single most common cause of migration delays and post-launch bugs, and it’s the one part you can fix yourself, for free, starting today.

Why does product data make or break the migration?

Import tools are literal. They move what exists — duplicates, typos, inconsistent option names — with perfect fidelity. The bugs then surface after launch: wrong images attached to variants, search results full of noise, filters that miss half the catalog, shopping feed disapprovals.

When we built NVMOS, an off-road parts brand with 75,000+ SKUs, from zero to live in under 90 days, the schedule held because catalog structure was settled before the first import — not discovered during it. Scale wasn’t the risk. Ambiguity was.

How do you clean SKUs and variants?

Start in a spreadsheet, one row per variant:

• One SKU convention. Unique SKU per variant, no spaces or special characters, one format across the catalog.
• Kill duplicates. The same product listed under three names is a returns and inventory problem waiting to happen.
• Normalize option logic. Pick canonical option names (Size, Color, Material) and canonical values — “XL” or “X-Large,” not both.
• Know the platform model. Shopify’s current APIs support up to 2,048 variants per product, but plenty of themes and apps still assume the older 100-variant limit. Verify what your chosen theme and apps handle before you flatten products together or split them apart.
• Resolve parent/child structures. Magento configurables and WooCommerce variable products map to Shopify’s product-plus-variants model. Standalone “simple” products that should really be variants of one product — merge them now, while it’s a spreadsheet edit.

What belongs in metafields?

Anything customers filter, compare, or search by: dimensions, materials, weight, compatibility and fitment, care instructions, size charts, ingredients.

The rule is simple — if it’s a fact, make it a field. Specs buried in description prose can’t drive filters, feeds, or comparison. Define the schema first in a sheet: metafield name, type, product-level or variant-level, allowed values. Then populate it as columns before import.

This work pays forward beyond launch. Shopping feeds and AI shopping agents read structured attributes, not paragraphs — an agent can’t recommend a tent whose capacity and weight live only in lifestyle copy. More on that in our agent-readiness guide.

How should you handle images?

Images are the most commonly botched part of catalog migrations. Four rules:

• One naming convention tied to SKU or handle, so mapping is mechanical.
• Sufficient resolution — roughly 2,000px on the long edge for zoom — with a consistent aspect ratio per category so grids don’t look ragged.
• Dedupe and prune. Don’t migrate images for products you retired in 2022.
• Treat alt text as a data column, not an afterthought. It’s accessibility, SEO, and agent-readability in one field.

And map variant-image assignments explicitly in your sheet. “Blue shirt showing the red photo” is the classic post-launch bug, and it’s always a mapping gap.

How do you plan categorization?

Map your old category tree to Shopify collections before import. Where a category page holds rankings, rebuild it one-for-one and 301 the old URL — carrying rankings through URL changes is its own discipline, covered in our 301 redirects guide.

At scale, prefer rule-driven collections over hand-curated ones. Automated collections built from clean tags and attributes are the only way a 75,000-SKU catalog stays maintainable — nobody hand-places parts into “Jeep JK bumpers.” Decide your tag vocabulary deliberately: tags for merchandising rules, metafields for specs. Then assign Shopify’s standard product taxonomy categories, which feed tax settings and shopping channels.

The full prep checklist:

1. Export the complete catalog to spreadsheets — one row per variant.
2. Dedupe products and standardize SKUs.
3. Normalize option names and values.
4. Rewrite titles into one consistent order (brand, model, key attribute).
5. Define the metafield schema and populate it as columns.
6. Audit, rename, and prune images; write alt text.
7. Map variant-image assignments.
8. Map old categories to collections, with rules where possible.
9. Choose the tag vocabulary and taxonomy categories.
10. Import a 50-product pilot, QA it hard, then run the full import.

FAQ

How long does product data prep take? A small catalog takes a focused week. Tens of thousands of SKUs take three to six weeks, usually running parallel to design work. Start before you sign with an agency — it lowers your quote either way.

Should I migrate discontinued products? Migrate anything with traffic or backlinks so you can redirect or mark it unavailable; drop dead SKUs entirely. Historical orders will still reference them fine.

Can’t the migration app just handle this? Migration apps move data faithfully — including the mess. They don’t decide your variant logic, metafield schema, or taxonomy. That’s the prep, and it’s human work.

Want a second set of eyes on your catalog?

A one-hour catalog review before you migrate is the cheapest insurance in the whole project. We audit data structure as part of every platform migration — send us a sample export and we’ll tell you what will break and what won’t.