Diese Seite ist auf Englisch. Eine Übersetzung gibt es noch nicht; kanonisch ist die englische URL. Englische Seite öffnen

Legacy modernization without switching the system off

I modernize live software systems that cannot be taken offline for a rewrite. The usual case is a content or commerce platform on an ageing CMS — Sitecore, Umbraco, Magento — that has to move to a headless front end on Next.js and a current content layer, while editors keep publishing and the public URL keeps returning 200. I have done this as a senior engineer on enterprise migrations: Sitecore 9.2 to Sitecore XM Cloud with a headless Next.js front; Umbraco and Magento to Contentful and Vercel. The work is data, URLs, editor workflow, and incremental cut-over — not a greenfield demo that never receives production traffic.

Who this is for

Companies whose public site or content platform is trapped on a CMS version that is out of support.
Teams that already bought a headless CMS and still have the old site answering production.
Engineering leaders who have been sold a rewrite and want a staged alternative.

Problems staged modernization is for

The CMS cannot be upgraded in place, but SEO, editor habits, and integrations cannot reset to zero.
A previous migration stalled at 'we will cut over next quarter'.
Data mappings were underestimated; the new CMS is empty or wrong.
The front end was rewritten while the content model was copied blindly.

What I take on

Content model design for the destination CMS, not a 1:1 clone of the old tree.
Migration pipelines with integrity checks, not a one-off script.
Headless Next.js fronts that editors can still reason about.
URL, redirect, and preview strategy so search and campaigns survive.
Strangler paths: new routes in production beside old ones until the old ones are dark.

How a live migration runs

Inventory what must be true on day one: URLs, metadata, editor entry points, integrations. Design the new model around how content is actually reused, not around the old folder tree. Automate the move, measure mismatch, and put the new front on a subset of traffic or paths. Do not schedule a Friday night cut-over of an untested mapping. Keep a rollback that is a DNS or routing change, not a restore-from-backup prayer.

Rewrite, strangler, or stay

Modernization is a risk decision. The repository age is not the criterion.

Stay and contain if the constraint is not the stack

If delivery is slow because of process or missing tests, a new CMS will not help. Fix the constraint.

Strangle if the system earns money while you migrate

New work goes to the new path. Old paths freeze except for legal and incident fixes. This is slower to announce and faster to survive.

Rewrite only when the old model is the bug

A commerce information architecture that cannot express the business is a candidate. A React version bump is not.

Never migrate content without an owner for each type

Unowned fields become 'dump into rich text'. That is how you pay twice.

How these programmes fail

Big-bang cut-over with no shadow running.
Pixel-perfect recreation of a 12-year template set in a modern stack.
Migration as a developer-only project with editors seeing the CMS on go-live morning.
Ignoring redirects until ranking drops.

Deliverables

A staged plan: what moves first, what stays, what is deleted.
Automated migration with reports on lossy fields.
Production front on the new stack for a real subset of the estate.
Redirect map and editor runbook.

Ausgewählte Arbeit

Vita Student — Studentenwohnen all-inclusive. Studios, Apartments und geteilte Zimmer — alles inklusive.
Nordic Learning Platform — Smart Learning Platform: Werkzeuge für Leitung, Lehrkräfte und Lernende auf ihrem Lernweg.

Fragen, die gestellt werden

Can you modernize a system under NDA?

Yes. Some past assignments cannot be described in public beyond the stack and the shape of the work. The method does not require naming the brand on this site.