The short answer
Do not switch the old system off on the day the last page moves. First prove that nothing still depends on it, by reading its logs. Then take a backup you have tested, decide what happens to every leftover URL, restrict the old site for a while to see what breaks, and only then cancel it. Removing the proxy is a separate step, and you should treat it as a small launch.
First, decide if it should go at all
Sometimes the right ending is two systems, on purpose. A docs section or tool may belong on its own platform for good, behind the same domain. That is fine when all of this holds:
- The section has a named owner and a budget.
- The proxy has an owner, monitoring and a place in your yearly plans.
- The shared header, footer and tracking have a process for staying in step.
What you want to avoid is the accidental version, where a few forgotten pages stay on an old system nobody looks after, because nobody ever decided what to do with them.
Decision rule Every path still on the old system needs one of three labels: moving by a set date, retiring by a set date, or staying on purpose with an owner. If a path has no label, give it one this week.
Prove nothing depends on it
Your route table shows what you meant to send to the old system. The logs show what actually goes there. Pull the old system's access logs and group them by path. If you have no logs there, use your proxy's.
| What you may find | What to do |
|---|---|
| Pages with human visits | Move, redirect, or decide on purpose to remove them |
| PDFs, images and downloads | Move them with the same paths, or redirect them |
| Old scripts or styles loaded by new pages | Host them with the new site, then remove the references |
| Form or API endpoints | Find the sender, then move or replace the endpoint |
| Crawler visits to old URLs | Check them against your redirect map |
| Bots probing for login pages and exploits | Ignore these |
Then look at what the logs will not show. The old system may also send things out: scheduled jobs, emails from a plugin, feeds and webhooks to other tools. Check each of these by hand with whoever looks after the system.
Leftover URLs
Each URL still left over gets one treatment. Our redirect map tool helps with the list, and the Playbook's redirect chapter has the rules.
- A close match exists. Use a permanent redirect straight to it. Google's site move guide says to keep redirects for at least a year, and longer for users, if you can.
- No match, and no value. Let it return a real 404 or 410. Google's error guidance says URLs answering with these are removed from the index. Do not send many old URLs to the home page. Google advises against sending lots of old URLs to one irrelevant page.
- Old redirects that lived on the old system. These stop working when it does. Copy them to the new system, and point each one straight at its final URL, so you avoid chains.
The run sheet
Each step has a proof. Do not move on until you have it.
| # | Step | Proof it worked |
|---|---|---|
| 1 | Label every remaining path | Route table has no blanks |
| 2 | Stop publishing on the old system | Editors told, in writing |
| 3 | Back up the database and files, and try a restore | The restored copy opens |
| 4 | Save rendered HTML of any page being removed | Files stored with the project |
| 5 | Put redirects on the new system | Old URL list checked, each one correct |
| 6 | Change the proxy's default to the new system | Missing pages return a 404 from the new system |
| 7 | Remove the old sitemap from the index and robots.txt | Sitemap lists only live public URLs |
| 8 | Block public access to the old system, keeping it running | Old hostname asks for a login |
| 9 | Wait, and read the logs and error reports | No real traffic or breakage reported |
| 10 | Shut it down, and cancel the contract | Backup confirmed first, cost removed |
| 11 | Remove old DNS records | Mail and other services still work |
Step 8 is the safety net. If something still needs the old system, you find out while switching it back on takes a minute. Agree how long to wait at step 9 based on what you found, and allow for anything that runs monthly or quarterly.
Take care at step 11. Your domain's DNS may also hold records for email and other tools. Remove only what you know belonged to the old site.
Removing the proxy
Once a single system holds everything, the proxy does nothing useful, and it is one more thing that can fail. Plan its removal like a small launch.
- List what it does beyond routing. Redirects, security headers, rewrites, caching and logging may all live in it. Move what you still need to the host.
- Change the host's settings first where needed. On Webflow, that means making the public domain the default again, in place of the origin subdomain. Review the global canonical, the sitemap base and robots.txt as part of this, and remove any base tag and href prefix you added for a subdirectory setup.
- Lower the DNS TTL ahead of time. Do this at least a week before, then point the public domain directly at the host.
- Check SSL on the public domain right after the change.
- Run your batch check from chapter 3 across the whole site.
- Deal with the origin hostname. Redirect it to the public domain so old links and caches resolve, and keep watching for it in search.
The URLs have not changed, so you should not need Search Console's Change of Address tool. Google's guide says it is only for moves between domains or subdomains.
What to keep afterwards
- The tested backup and your saved page snapshots.
- The final route table, and the full list of redirects.
- A short note on what moved, when, and which systems were involved, in your site docs.
- A date on the calendar to review the redirects, in about a year.
- An eye on search for a while. The Playbook's first 30 days has a schedule you can reuse.
Then look at what the move was for. Can the team now publish without help? Is there a single place to edit a page? If you want help looking after the site from here, see our care plan.
Done when One system answers every URL, or every path on a second system has an owner, and nothing you pay for is serving a page nobody chose to keep.


