Back to News & Blog

10 things that can go wrong in a CMS migration, and how to catch each one

Illustration of an old web page with a broken image and a gallery shortcode, a new page listing redirect status codes, and a magnifying glass over a broken link

A CMS migration can look finished on launch day and still lose traffic over the following months. In previous migrations we’ve inherited, the damage rarely comes from the new platform itself, but rather the overlooked details nobody checked, such as a redirect missing for a URL with backlinks, copy pulled from the wrong field, or a form that quietly stopped reporting. Below are the ten we see most often, what to look for and how to fix each one before launch. For the wider picture, we’ve also written up how we run a content migration from audit to launch.

1. The redirect map only covers what’s in the sitemap

A sitemap only lists the indexed pages a site has today, and the URLs that hurt most when they break are often missing from it. These can be pages removed years ago that still carry backlinks, old campaign landing pages, PDFs, legacy paths ending in .html, and rules that lived in the server configuration instead of the CMS. Server-level rules are the easiest to overlook, because they don’t come across with a copy of the CMS at all.

What to look for

Build the redirect map from every source available, including a full crawl, the sitemaps, at least a year of Search Console page data and the backlink profile. Label each URL as live, already redirecting or already broken, and give each URL a one-to-one target.

The fix

Test every row of the map on staging and again on launch day, checking both the status code and the final destination. Broken URLs that still carry backlinks deserve a relevant target page, so reserve the homepage for URLs with genuinely nowhere better to go. On the Coller EQT domain migration, building the map from all four sources brought 93 legacy paths held in server configuration to the surface, and each one was checked against the new site before launch.

2. Old redirects turn into chains, and old domains get forgotten

Redirects left behind by previous migrations cause trouble when a rule from five years ago points at a URL that now redirects again. This can result in a redirect chain, where each extra hop slows crawling and adds another point of failure. Domains cause a similar problem: the bare domain without www, old brand names and retired subdomains all need to keep working. Registrar-level domain forwarding often can’t serve HTTPS, so anyone arriving on the https:// version of an old domain can hit a certificate error instead of a redirect.

What to look for

List every hostname the organisation has used, then request the http:// and https:// versions of each, plus a sample of old URLs, and count the hops to the final page.

The fix

Rewrite old rules so every legacy URL reaches its final destination in one hop, and host old domains somewhere that can serve a valid certificate. The Coller EQT move took the main site and two subdomains onto the new domain on a fixed date, with more than 1,200 URLs redirected one-to-one.

3. Content lives in more than one field, and the wrong one gets migrated

Most CMSs store a page’s content in more places than the editor shows: the main body field, custom fields added by a page builder or ACF, excerpts, and leftovers from a previous rebuild. A migration script that reads the default body field can move a version of the page nobody has seen for years. It’s especially common on WordPress sites that have been rebuilt at least once, where the theme displays copy from a custom field while the main content box still holds an older version that the REST API and RSS feed keep serving.

What to look for

Write a field-level mapping for every template, listing each source field, whether it appears on the page and where it goes. Then compare rendered pages on the old and new sites for a sample of every template, because comparing database fields alone won’t show which one the old theme displayed.

The fix

Map from what renders on the page, and retire the unused fields during the migration so the confusion doesn’t carry over.

4. Embeds and shortcodes don’t survive the move

Shortcodes from old plugins turn into raw text such as [contact-form-7 id="12"] when the plugin isn’t on the new site. Iframes for videos, maps, forms or data dashboards keep pointing at the old domain, and inline styles from an old editor fight the new design. These rarely show up when a few key pages are checked visually, because they tend to sit in older content that nobody reviews before launch.

What to look for

Before import, search the exported database for the old domain, for [ followed by known shortcode names, and for <iframe. After import, crawl the new site and list every resource still loading from the old domain.

The fix

Convert shortcodes to blocks or components during the migration, and run search and replace with a tool that handles serialised data, such as WP-CLI’s search-replace command, so stored settings aren’t corrupted.

5. Media and documents are still linked to the old site

Images and PDFs hard-linked to the old domain or CDN keep working until the old hosting is switched off, then all break at once. PDFs also tend to carry backlinks of their own; on the Coller EQT migration, media URLs kept their redirects for exactly that reason. Alt text is often lost along the way when files are re-uploaded instead of migrated with their metadata.

What to look for

Crawl the new site after import and filter every image, document and script by hostname. Anything still served from the old domain is a future broken link.

The fix

Import media into the new library with its alt text, rewrite references to the new paths, and add redirects for old media URLs that have backlinks or appear in search results.

6. SEO metadata goes missing, or keeps pointing at the old domain

Titles and meta descriptions are usually stored by an SEO plugin rather than the CMS itself, so they only come across if someone maps them. When the domain is changing as well, and the switch is made in the database, Yoast’s stored data can keep canonical and Open Graph URLs on the old domain until it’s rebuilt, which tells search engines the old site is still the original.

What to look for

Crawl the old site and the new staging site, then compare title, meta description, canonical and Open Graph URL for every matching page.

The fix

Map SEO metadata explicitly in the migration plan, rebuild Yoast’s indexables after any domain change, and spot-check canonicals on live in the first hour. Our SEO migrations work runs these checks as standard.

7. Staging gets indexed, or “noindex” goes live

A staging site with no password can be found through a shared link and indexed as a duplicate of the real site. The reverse can happen at launch, when the “Discourage search engines” setting and noindex tags used on staging get carried over to the live site. Both are easy to miss because the site looks perfectly normal in a browser.

What to look for

Check the live site within the first hour for the robots meta tag, the X-Robots-Tag header and robots.txt, and search Google for the staging domain.

The fix

Protect staging with basic authentication, because a robots.txt block alone doesn’t stop a linked staging URL being indexed, and make the robots check a named step on the launch checklist.

8. Tracking breaks quietly

Google Analytics and Tag Manager are often built around the old site’s markup: click triggers keyed on CSS classes or generated IDs, custom variables that call functions from the old theme, and form conversions measured by thank-you pages that no longer exist. Traffic reports carry on looking healthy while conversions fall to zero. The risk is highest when a rebuild changes the front-end component library, because the IDs and classes those triggers rely on can change on every interactive element at once.

What to look for

List every conversion and key event, test each one in GTM Preview on staging, then repeat the test on live on launch day. Compare conversion counts for the week before and the week after.

The fix

Key triggers on stable attributes, such as data attributes, link URLs or page paths, and give every form a confirmation event or thank-you URL that survives the redesign.

9. International sites lose their hreflang

Multi-language and multi-region sites rely on hreflang tags that reference each other, include an x-default, use the right language and region codes, and point at final URLs. Change the URL structure of the language folders or subdomains and every one of those references has to change with it. When we moved M7 Real Estate from Umbraco to WordPress multisite, that meant a 16-language site with more than 300 pages, migrated without a single error and with hreflang applied across every regional site.

What to look for

Crawl the new site with hreflang validation switched on and look for missing return links, non-200 targets and incorrect codes.

The fix

Generate hreflang from the CMS’s own language relationships, through multisite or a translation plugin, so it updates as pages change, which hand-entered tags rarely do for long. Our multilingual websites builds are set up this way.

10. A traffic and rankings baseline wasn’t established, so before/after comparisons are inconclusive

Without traffic and ranking figures from before launch, there’s no way to tell whether a dip in week two is a real problem or the usual fluctuation, and no way to show the board that the migration worked. It’s often the first step dropped when a launch date is tight.

What to look for

Before launch, capture at least 28 days of Search Console clicks and impressions by page and query, rankings for priority terms, Google Analytics sessions and conversions, and the current indexing picture.

The fix

Check Search Console daily for the first few weeks after launch and use URL Inspection on priority pages. After the Coller EQT move, organic traffic recovered within four weeks, and a sweep of 667 URLs through the URL Inspection API two weeks after launch showed 90% of the main site’s pages indexed on the new domain, up from 70%.

Where to start

If time before launch is short, start with items 1, 3 and 8. Missing redirects lose rankings, the wrong content field loses the copy itself, and broken tracking hides both problems until someone asks why enquiries have dropped.

If you’re planning a move to a new platform and want to keep the search visibility you’ve built, talk to us about CMS migration.

Let’s scope your next project.

Whether you’re in financial services, healthcare, energy or another sector, we start every project with a conversation about your requirements. No commitment, just a discussion about your needs.

Schedule a consultation