Rehearsed, then switched
WordPress Migration: Host Transfer With No Downtime or Lost Mail
Move a WordPress site to new hosting without the outage everyone expects. The new install is built and tested while the old one keeps serving, DNS switches only once the copy answers correctly, and the original stays up until it is provably unnecessary. Send the address for a written transfer plan naming every moving part before anything is touched.
Get a free written site audit
Send the address and we return a written report inside one working day: patch backlog, exposure, load timings, restore status.
- The old install stays live until the new one is proven
- Files, database, mailboxes and certificates, not just the files
- A redirect map built before the move, not discovered after it
- 1,000+ sites built and moved on this platform over 10+ years
What this covers
- Full inventory before anything moves: PHP version, plugins, cron jobs, mailboxes
- The database, including options and serialised values the search-replace must respect
- Uploads, at full resolution, with the generated sizes rebuilt rather than copied
- Certificates reissued and verified on the new host before the switch
- Email left working, or moved deliberately, rather than discovered broken on Monday
- A staging copy on the new host, tested against what the site earns money from
- DNS switched with the TTL lowered in advance so the window is minutes
- The old install kept running until the new one has served real traffic
What actually goes wrong in a host move
Almost never the files. Copying a directory is the easy part, and it is the part every migration guide spends its time on.
What goes wrong is everything attached to the site that is not the site. Email that lived on the old host and stops the moment the MX record follows the A record. A certificate that was never reissued, so the new server answers with the wrong name. Cron jobs that existed at the server level and silently do not exist any more, which nobody notices until the backups stop. Serialised values in the database that break when a naive search-replace changes a string length.
Every one of those is found by an inventory taken before anything moves. That is why the first step of this process is not a copy.
Why the old install stays up
DNS does not switch. It propagates, and during propagation some visitors reach the old server and some reach the new one.
Two things follow. The old install has to keep working, or a proportion of your traffic sees an outage that looks like a total failure to them. And anything written during that window – an order, an enquiry, a comment – lands in whichever database that visitor happened to reach, which is why the window is kept to minutes by lowering the TTL a day in advance and why both sides are reconciled afterwards rather than assumed identical.
Keeping the old host running for a fortnight after the switch costs very little. It is the only rollback that works instantly, and the alternative is restoring a backup onto infrastructure you have just left.
Moving hosts, and moving platforms, are different jobs
A host move keeps the URLs. Nothing about the addresses changes, so search engines have nothing to re-learn and the risk is operational rather than editorial.
A platform move changes them, and that is a different problem entirely. Squarespace puts blog posts under one prefix and Wix historically used fragment URLs; neither maps onto WordPress permalinks by accident. Getting that wrong is how a site emerges from a redesign with its rankings gone, and the reason each source platform below has its own page rather than a paragraph on this one.
The common part is the redirect map: every old address, its new address, one hop, built before the move and tested after it. The part that differs is what the old platform will let you take with you.
After the switch
The move is finished when the new install has served real traffic without incident, not when DNS resolves.
For the first fortnight the site is monitored more closely than usual, the old host stays paid for, and the redirect map is checked against real 404s rather than against the spreadsheet it was planned on. Search Console will report the change; that is expected and it settles.
Most sites arrive here because nobody was looking after them, which is usually what caused the move. Ongoing cover is set out on maintenance plans and pricing, the step-by-step version of this process is in the WordPress migration checklist, and hardening the new install before it is exposed is security and hardening. The whole recurring job is WordPress maintenance.
How the work runs
- InventoryVersions, plugins, themes, cron, mailboxes, certificates, and where the DNS actually lives - which is frequently not where the client thinks. This is the step that finds the surprise, and finding it now is the whole point.
- Build on the new hostA working copy, same PHP, same data, same plugins. Not a restore into the unknown: the target is stood up and exercised while the live site carries on untouched.
- Test what earns moneyCheckout, booking, contact forms, login, search. A migration that moves every file and breaks the enquiry form has failed at the only thing it was for.
- Lower the TTLTwenty-four hours before the switch, so propagation is measured in minutes rather than a day of split traffic writing to two databases.
- Switch and watchDNS moves, and both installs are monitored while it propagates. Anything written to the old one during the window is reconciled rather than lost.
- Keep the old oneFor a fortnight, running and paid for. It is the rollback, and it costs a few pounds against the alternative.
Plans that include this
- Starter$39/monthor $390 a year — two months freeBlogs, portfolios and brochure sites
- UpdatesMonthly
- BackupsWeekly, 30-day retention
- Uptime5-minute checks
- Malware removalNot included
- Most chosenProfessional$79/monthor $790 a year — two months freeSmall businesses and lead-generation sites
- UpdatesWeekly + PHP version management
- BackupsDaily, 60-day retention
- Uptime1-minute checks
- Malware removalIncluded
- Business$149/monthor $1490 a year — two months freeStores, membership sites, anywhere downtime costs money
- UpdatesWeekly, tested on staging first
- BackupsDaily, 90-day, multi-location
- Uptime1-minute + 2-hour restore
- Malware removalIncluded
What clients say
They found malware two of our previous agencies had missed, cleaned it in an afternoon, and told us exactly how it got in.
Operations Director, professional services, 4 sitesThe staging step is the whole point. Three years and not one update has taken our booking form down.
Practice Manager, healthcareWe resell it to 31 clients under our own brand. The reports go out with our logo and our clients have never heard their name.
Agency Founder, digital agency
Related pages
Questions people ask before calling
How much downtime is there?
None that a visitor sees. The new install is serving before DNS points at it, and the old one keeps answering until propagation finishes, so both are live during the window rather than neither.
What about email?
It is the most commonly forgotten part of a host move and the most disruptive to get wrong. Mailboxes are inventoried at the start and either left where they are or moved deliberately, with MX records handled as their own step rather than dragged along with the website.
Will my rankings drop?
A host move does not change URLs, so there is nothing for search engines to re-learn. What does affect rankings is the site being slow or unreachable during a botched switch, which is what the rehearsal prevents.
Can you move a site that is already broken?
Yes, and it is common - a migration is often the moment someone finally addresses a site nobody has maintained. We say plainly what needs fixing and quote it separately rather than absorbing it into the move.
What does it cost?
Quoted after the inventory, because a five-page brochure site and a WooCommerce store with 4,000 orders are not the same job. Migration onto one of our maintenance plans is handled as part of onboarding.
Do you move from other platforms too?
Yes - Squarespace, Wix, Weebly, Shopify, Magento, Drupal, Webflow, Umbraco and Sitecore each have their own page, because each has a different failure mode and a different redirect problem.