Run Both Sites in Parallel, and Cancel Nothing Early

The instinct during a migration is to stop paying for the old thing as soon as the new thing exists. It is the wrong instinct, and it is the one that turns a recoverable problem into a permanent one.

For roughly the price of one extra month, the overlap window buys you three things you cannot buy any other way: somewhere to fetch images from, somewhere to look things up, and somewhere to retreat to. Here is how to run it properly.

Two sites is the safe state, not the dangerous one

People worry that having the old and new sites both existing will confuse search engines or split their traffic. It does not, provided only one of them is visible to the public at a time.

The old site keeps the domain and stays live. The new site sits on a temporary address, behind a password, invisible. On cutover day the domain moves and the visibility swaps: the new site becomes public, the old one becomes an archive that only you can see. At no point are two public versions competing.

The dangerous state is the other one — deleting the old site first, then building. That is a period of days or weeks with nothing where your website used to be, and it is remarkably easy to fall into if you cancel on the day you start.

Keeping the new build out of search results

Every platform has a way to do this, and every one of them works differently.

Squarespace lets you set a site-wide password while the site is in progress, and recommends it in its own migration guidance precisely so you do not end up with two live versions of yourself. It is the cleanest of the options — nothing loads at all without the password.

Shopify gives new stores a storefront password by default. Leave it on. Turning it off is the act that makes your store public, which makes it a clean, deliberate launch switch.

WordPress has a setting under Reading to discourage search engines from indexing the site, and it is worth understanding exactly what that does: it asks crawlers not to index, and well-behaved crawlers comply. It is a request, not a lock. If the content genuinely must not be seen, use a password or a staging environment your host provides, not the checkbox alone.

Whichever you use, check it. Open the temporary address in a private browsing window and confirm you are stopped.

The duplicate content question, answered

This worry comes up constantly and it is mostly misplaced.

Content that is not publicly reachable cannot be indexed, so a password-protected build creates no duplication at all. What does create duplication is leaving the old site publicly reachable on its own platform subdomain after the domain has moved — the same pages, on two addresses, both crawlable.

The fix is one action on cutover day: after the domain points at the new platform, put the old site behind a password or unpublish it. It stays in your account, you can still log in and read it, and the public and the crawlers cannot.

What the old site is still doing after launch

This is the part that surprises people, and it is the strongest argument for the overlap.

It is hosting your images. Importers commonly fetch media from the live old site rather than from the export file — the official WordPress guidance for importing is explicit that the original site needs to be publicly accessible for attachments to download. If you have cancelled, there is nothing to fetch from, and you get posts full of broken images. Note the tension with the previous section: if you are still running an import, the old site needs to stay reachable until the media has landed, so do that work first and lock it down afterwards.

It is your reference. You will find a page you did not rebuild, a phone number you mistyped, a product description you truncated. Having the original in front of you turns each of those into a two-minute fix.

It is your retreat. If something is badly wrong — a checkout that will not take payment, a catalogue that imported at the wrong prices — you can point the domain back while you sort it out. That option evaporates the moment the old account closes.

The order of cancellations

Website builders typically sell three separate things on one account: the site plan, the domain registration, and a business mailbox. They renew independently and they cancel independently, and Wix documents this explicitly for its own products — cancelling a site plan does not cancel the domain, and any other services bought through the account need turning off separately.

So the order matters.

1. Nothing, for two to four weeks. Launch, watch, fix.
2. The site plan. Turn off auto-renewal rather than deleting the account. Deleting takes the exports with it.
3. The business mailbox, but only once mail is confirmed working on the new arrangement for a full week, in both directions.
4. The domain registration. Last, or never — keeping the registration where it is costs nothing extra and removes a step from the critical path.

A note on what cancelling a site plan actually does. On Wix, for example, the site does not vanish: it reverts to the free tier, the custom domain connection is dropped, and platform advertising appears. The content stays in the account. That is useful to know, because it means turning off auto-renewal is less destructive than it sounds — but do not rely on it as a backup strategy, because other platforms behave differently and none of them promise it forever.

What the overlap actually costs

One or two billing cycles on the old platform. For most small sites that is somewhere between the price of a takeaway and the price of a dinner.

Set that against what it insures: your images, your ability to check anything, and your ability to reverse a bad launch. Nobody who has kept the overlap has regretted it. Plenty of people have written about regretting cancelling early.

If money is genuinely tight, the compromise is to shorten the window rather than skip it. Two weeks is enough for most brochure sites. A month is right for anything with a store or a large archive, because seasonal pages and rarely-visited product URLs surface their problems slowly rather than on day one.

There is a second cost people forget to count, and it runs the other way. Rushing the cancellation to save one subscription usually means rushing the verification too, and an unverified launch generates support work — customers emailing about broken links, orders that did not confirm, a form nobody is receiving. That labour is not free either, and unlike the subscription, you cannot predict how much of it there will be.

When to finally close the account

Work down the verification list in the moving-day checklist before you close anything: sample redirects landing correctly, forms submitting and notifying, email flowing both ways, products matching on price and stock, analytics recording, images served from your own domain.

When every one of those is true and has been true for a fortnight, turn off the renewals. Keep your export folder and your screenshots somewhere permanent. And make a note in your calendar a month out to check the 404 report one last time, because the long tail of forgotten URLs surfaces slowly.

The full order of operations that this window sits inside is on the plan for switching website builders.