The short answer
A full cutover is wrong when part of the site cannot move by the date the rest must, or when the site is so large that the old one keeps changing while you rebuild it. If neither is true, switch everything at once. It is simpler, and our Site Move Playbook covers that path.
Google's site move guide points the same way. It recommends that small and medium sites move all URLs at the same time, and says large sites can move a section at a time. That guide is about moves where URLs change, but the reasoning holds for a platform change too.
Two kinds of staged move
People say "in stages" and mean two different things. Work out which one you need, because the work is very different.
| Routing stages | Platform phases | |
|---|---|---|
| What is live | Two systems at once, on one domain | One system at a time |
| What moves | Batches of pages | The whole site, at planned points |
| Needs a reverse proxy | Yes | Not by default |
| Main risk | Two systems drifting apart | One phase never ending |
| Covered in | Chapters 2, 3, 4 and 6 | Chapter 5 |
If you only want the idea of a reverse proxy and a quick view of the SEO points, read the short post on moving a site to Webflow in stages. This guide is for people who have decided to look closer and need the working detail.
Signs a full cutover is wrong
You need at least one of these, and you should be able to name the pages it applies to.
| Sign | How to test it |
|---|---|
| A section depends on a system you cannot replace yet | Name the system, the pages and who would have to rebuild it |
| The rebuild outlasts a realistic content freeze | Ask the content team how long they could stop publishing, then compare that with the build plan |
| The old contract or the new design has a date the full build cannot meet | Put both dates side by side, with the page count for each |
| The site is large enough that you want to see each section settle | Count the URLs in search, and ask if you could check them all in one launch day |
| A team owns one section and works to its own schedule | Ask that team when it could join a single launch |
Google also says that, where it makes sense, you can start with a section that changes little, to see how traffic and indexing behave. That is a good first batch whichever method you choose.
Reasons that sound good but are not enough
- "It feels less risky." Stages split one big risk into many small ones, and add a new one: a live proxy. If your only worry is launch day, a better run sheet fixes that for less money. See the launch-day order.
- "We can pay for it in parts." You can usually phase the budget without phasing the routing. Launch the pages that matter on one date, then add the rest as new pages.
- "We want to redesign page by page." Visitors will cross from the new look to the old one on every visit. That is fine for a short time. It gets hard to defend if it lasts.
- "We might keep the old system." Undecided is not a plan. Decide, then read chapter 6 on staying with two systems on purpose.
The options in between
A proxy is the heaviest option here. Check the lighter ones first.
| Option | How it works | Good fit | Catch |
|---|---|---|---|
| Smaller cutover | Remove or merge weak pages, then switch in one go | The site is long but only part of it earns anything | Every removed URL still needs a decision and a redirect |
| Short freeze | Stop publishing for an agreed window, then switch | The team can pause | Needs agreement across marketing |
| Separate subdomain | The part that cannot move goes to something like docs.example.com | A clear, self-contained section | Those URLs change, so they need redirects and their own move |
| Reverse proxy | Both systems sit behind one domain, and URLs stay the same | URLs must stay, and parts must wait | Someone has to run it, and everything depends on it |
What two live systems cost
Add these to your plan before you commit. They run every week until the move ends.
- Two headers, footers and menus to keep matching.
- Two places to publish, and a list of which pages live where.
- Two sets of plans and contracts, including the paid Webflow Site plan the proxy setup requires.
- A proxy to host, watch and update, with someone on call for it.
- Caches at more than one level, and a way to clear them.
- Tracking, consent and forms that must behave the same on both sides.
- Support on your own. Webflow's help article on self-managed proxies says it does not support third-party proxies. If you are on Enterprise, it sends you to Sales to talk this through.
An outage matters more here too. If the proxy fails, every page behind it fails, on both systems.
Five questions before you commit
- Which exact pages cannot move with the rest, and why?
- Who owns the proxy, by name, and who covers when they are away?
- What is the order of the batches, and which comes first?
- What date do you expect the last batch, and what happens if it slips?
- What will you do with the old system at the end?
Decision rule Choose a proxy only if you can answer all five in writing. If you cannot name the pages, switch all at once. If you cannot name the owner, use a lighter option from the table above.


