Xumulus Logo
Agent-ready search & discovery on Next.js + Elastic/OpenSearch

Magento migration services

Three different projects get called a Magento migration, and they fail for different reasons. Here is how we scope each one, and what we have learned running them since before Magento 1 reached end of life.

Elastic
Relevance + speed
Next.js
Modern UX + performance
RAG
Agentic discovery ready
AI Discovery Signals
Retrieval-first architecture makes agentic discovery possible.
Scroll

Which migration are you actually doing?

The word covers three jobs with almost nothing in common. Getting the name right early is most of the estimate, because each one puts its risk somewhere different.

Magento 1 to Magento 2

Magento 1 reached end of life in June 2020 and stores are still on it. This is not an upgrade — it is a rebuild with a data migration attached. Nothing about the theme, the extensions or the customisations carries across, and the honest question is whether Magento 2 is still the right destination in 2026 or whether the budget belongs somewhere else.

Magento 2 version upgrades

Adobe Commerce moves, extensions lag behind it, and a store that skipped three releases is a bigger job than a store that skipped one. The work is mostly dependency archaeology and regression testing, and it is the one case where in-place is almost always right.

Migrating off Magento

Some catalogs no longer need a monolith. We move stores to BigCommerce, to Shopify Plus, and to an index-first stack on Next.js with Elastic or OpenSearch behind it. We will say plainly when staying on Adobe Commerce is the cheaper answer — a replatform to escape a maintenance bill often just relocates it.

Discovery bottlenecks

What actually goes wrong

None of these are surprises. They are the five failures we plan around at the start, because each one is far cheaper to prevent than to discover during a cutover weekend.

01
The URL map is written after launch
Every migration moves URLs, and the redirect map is the difference between keeping the rankings and starting over. It has to be built from a real crawl and real Search Console coverage data before cutover — not reconstructed from what breaks afterwards. Permanent redirects, not temporary ones: a 307 tells Google to keep the old URL indexed and withhold ranking signals from the target, which is precisely backwards for a move that is permanent.
02
Product data is migrated, product meaning is not
Attributes survive the export. The reason an attribute exists — that it drives a facet, a rule, a bundle, or a fitment lookup — lives in the old system’s configuration and in somebody’s head. Catalogs arrive on the new platform technically complete and commercially useless, which is the most common form of a migration that "worked".
03
Extension debt is carried across intact
A Magento 1 store typically runs dozens of extensions, several abandoned, some patched in place. Porting all of them reproduces the original problem on a newer platform. Deciding which behaviour is genuinely load-bearing is a merchandising conversation, not a technical one.
04
Search and merchandising restart from zero
Synonyms, boosts, pinned results and redirects are years of accumulated tuning that no migration tool exports. Teams rediscover this in week one on the new platform, when the results pages look obviously worse than the ones they replaced.
05
Performance regressions surface under real traffic
A staging environment with a warm cache and no concurrency hides the failures that matter. Full page cache that silently is not caching, session locking that serialises AJAX requests — we have written up both, because we have had to fix both.
Key takeaway
Every one of these is a planning problem before it is an engineering problem.
Contrast

Upgrade in place, or move?

The two options are not close on cost, and the deciding factor is usually the catalog rather than the code.

01
Comparison
Stay on Adobe Commerce
Cheaper when the customisation is in checkout, pricing or B2B logic that the platform does natively
Replatform
Cheaper when the customisation exists to work around the platform
02
Comparison
Stay on Adobe Commerce
Keeps the merchandising configuration, the integrations and the team’s working knowledge
Replatform
Rebuilds all three, which is a real cost that estimates routinely omit
03
Comparison
Stay on Adobe Commerce
Right for large B2B catalogs with complex pricing and account structures
Replatform
Right when the storefront is the product and the catalog is comparatively simple
04
Comparison
Stay on Adobe Commerce
Ongoing upgrade cadence stays your responsibility
Replatform
SaaS absorbs the upgrade cadence, at the cost of what you can change
Bottom line
We have recommended both. The answer follows from the catalog and the roadmap, not from a preference.
The stack

How we run one

Three phases. The first two are where migrations are won, and they are the two that get compressed when a date is set before a scope.

01
1) Audit

Establish what is actually there

Before anything moves, we inventory the thing being moved.

  • Full crawl plus Search Console coverage, to build the URL map from evidence
  • Extension inventory, scored by whether the behaviour is load-bearing
  • Catalog and attribute audit, including what each attribute is for
  • Integration map — ERP, PIM, OMS, payments, tax, shipping
02
2) Data

Move meaning, not just rows

The migration itself is scripted and repeatable, and run many times before it is run for real.

  • Products, categories, customers, order history with referential integrity checked
  • Attribute and facet configuration rebuilt deliberately, not inferred
  • Akeneo or another PIM as the source of record where catalog complexity warrants it
  • Reconciliation counts on every entity, every run
03
3) Cutover

Make the switch boring, then watch it

A rehearsed cutover is a short one. Index recovery takes weeks after it, and the first fortnight is what tells you where the URL map was wrong.

  • Redirects deployed with the launch, not after it
  • Sitemaps and canonicals correct on day one
  • Load tested against real concurrency, not a warm staging cache
  • Rollback plan written down and tested
  • Coverage and impressions tracked against the pre-migration baseline
  • Redirect gaps found from live 404s and crawl data
  • Search relevance retuned against real queries on the new platform

Why us

Two decades of enterprise commerce delivery, and a Magento practice that goes back far enough that our own archive is mostly Magento 1 end-of-life warnings written while it was still news. We work across Adobe Commerce, BigCommerce, Shopify Plus, Akeneo and Oracle WebCenter, and we build index-first storefronts on Next.js with Elastic or OpenSearch.

That range is the point: we have no reason to argue you onto a particular platform, because we deliver on all of them.

Tell us what you are on and where you are going

Send the platform, the catalog size and roughly what is customised. We will come back with which of the three migrations this is, what it usually costs, and whether we think you should do it at all.

Search is the foundation of AI commerce