Rebuilding Without Losing Your Rankings

Where the losses actually come from

Not from the new design, and not from the content being different.

Almost every serious drop after a rebuild traces to addresses that changed without being mapped, so a search engine returns to a page it had ranked and finds nothing there.

Repeat that across a few hundred addresses and the positions built over years are handed back over a fortnight.

The recovery is slow and partial, which is why this is the one part of a rebuild worth being pedantic about.

What a redirect map is

A spreadsheet with two columns that matter: every address on the old site, and the specific address it should send people to on the new one.

That is the entire artefact. Its difficulty is not conceptual, it is that it is tedious and has to be complete.

It is also the deliverable most likely to be quietly skipped, because it produces nothing anybody can look at and admire.

Building it

Start from the crawl of the old site, which gives you every address that exists.

Add the traffic data, so the list is sorted by what people actually visit rather than alphabetically.

Then work down it, and for each old address write the new address that best serves somebody who was trying to reach it.

Do the top of the list by hand and carefully. Patterns will emerge for the long tail, and those can be handled with rules, but the pages that matter deserve individual attention.

The rule for choosing a destination

Send each address to the closest equivalent page, and interpret closest generously.

A page about one specific service goes to the new page about that service.

A page about a service you have dropped goes to the category it sat in, or to the nearest thing you still offer.

A merged page goes to the page it merged into.

The only addresses that should point at the homepage are the ones that genuinely have no better destination, and on most sites that is a handful rather than a third of the list.

Why the homepage catch-all fails

It is the shortcut everybody reaches for, because one rule handles everything and it takes ten minutes.

It fails twice.

For search engines, a redirect to something unrelated is increasingly treated as equivalent to the page having gone, so the value that address carried is not passed along.

For people, it is worse and more immediate. Somebody clicked a link about a specific thing and arrived at your front door with no explanation and no way to find what they wanted, which is a more annoying experience than a plain not-found page.

Permanent, not temporary

Redirects come in two kinds and the distinction matters here.

A permanent one says this has moved for good, and passes along most of what the old address earned.

A temporary one says the original will be back, so the old address is kept and little is passed along.

Rebuild redirects are permanent, and the temporary kind gets used by mistake often enough that it is worth checking explicitly rather than assuming.

A worked example

A parts distributor rebuilt a site of about eleven hundred addresses.

The top ninety were mapped individually over two days. The remaining thousand fell into four patterns handled by rules.

Testing the map against staging found two faults: one rule had a typo affecting roughly two hundred addresses, and forty pages had no mapping at all because they were not in the original crawl, having been reachable only through a search box.

Both were fixed before launch.

Traffic dipped by about eight percent in week two and had recovered by week five.

The comparison worth noting is a competitor who rebuilt the same quarter with a homepage catch-all, and was still below their previous traffic nine months later.

What the crawl will miss

• Pages reachable only by search on the old site.
• Pages behind a form or a login.
• Old campaign landing pages not linked from anywhere.
• Documents and PDFs with their own addresses.
• Pages that were removed but still receive visits.

Cross-check the crawl against your search console and your server logs, which between them know about addresses no crawler will find by following links.

After the switch

The map is not finished at launch, and the first fortnight is when the remaining gaps surface.

Watch the not-found reports in your search console and in your server logs daily for two weeks.

Every address appearing there with real traffic behind it is a line the map missed, and adding it late still recovers most of the value.

Submit the new sitemap so the new addresses are found quickly, and leave the old one in place briefly rather than deleting it immediately, so the redirects are discovered rather than the old addresses simply vanishing.

The links inside your own content

The map handles people arriving from outside. It does not fix the links inside your own pages.

Migrated body copy is full of links written against the old structure, and after launch every one of those quietly relies on a redirect to work.

That is not a failure, since the links function, but it means part of your own site now depends permanently on a mapping layer that was meant to serve outsiders.

It also compounds: a future migration inherits redirects pointing at redirects, and eventually somebody clears the old rules out and the internal links break all at once.

The fix is a search-and-replace pass across migrated content once the new addresses are settled, followed by a crawl to confirm no internal link is being redirected. Navigation, footers, and any links inside documents deserve the same check.

The counter-case

There is a case for deliberately not redirecting, and it is narrower than people use it.

Addresses that were never indexed, never linked, and never visited are genuinely worth nothing, and mapping them adds maintenance for no return.

Pages removed because they should never have been published, such as duplicated location pages, are sometimes better allowed to return not-found honestly than redirected to something that does not match what was promised.

The test is evidence rather than instinct: if an address has no traffic, no links, and no index presence, letting it go is defensible. Otherwise map it.

The sequence

1. Crawl the old site and export every address.
2. Cross-check against search console and server logs.
3. Sort by traffic and map the top of the list by hand.
4. Use rules for the patterned long tail.
5. Confirm every redirect is permanent.
6. Test the whole map against staging.
7. Test again after launch, and watch not-found reports for two weeks.

Step six is the one that catches the typo affecting two hundred pages, and it takes an hour.

Keeping addresses in the first place is covered in keeping the addresses when you rebuild.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.