How to redesign a website without losing your search traffic

The redirect map is the whole job. Where to pull every old URL from, what has to move and what must not, launch day in order, and the fortnight after.

Nitesh PadghanNitesh PadghanFounder, Storifex Media
6 min read
On this page (9)
  1. 01The short answer
  2. 02The checklist
  3. 03The redirect map is the whole job
  4. 04Your top twenty pages
  5. 05Keep the staging site out of Google
  6. 06What else has to move, and what must not
  7. 07Launch day, and the fortnight after
  8. 08A large site can launch in stages
  9. 09Where to start

A redesign is the most reliable way to lose search traffic you already have, and almost none of the risk is in the design. It is in the addresses. This is the work that protects them, in the order it has to happen, and the order matters more than any single item on the list.

The short answer

Before anyone writes code, list every address the old site has and decide where each one goes on the new site. That spreadsheet, the redirect map, is most of the job. Then carry the titles, the structured data and the content of your best pages across on purpose rather than letting the new templates regenerate them. Launch early in the week, watch Search Console's Pages report rather than the traffic graph, and expect two to four weeks of wobble while Google recrawls.

If the redesign has already happened and the traffic has already gone, start with traffic dropped after a website redesign instead. This piece is the version you do beforehand.

The checklist

When What Why
Before the build Export every old URL from four sources No single list is complete
Before the build Map each old URL to one new one This is what keeps the rankings
Before the build Note your top twenty pages by impressions These are the pages you must not break
During the build Keep the staging site out of Google So it does not compete with the live one
During the build Carry titles and descriptions across Change one thing at a time, later
During the build Rebuild the structured data Inherited schema describes the old site
During the build Decide one form of every URL One address per page, not three
Launch day Remove the staging block The most common launch mistake
Launch day Switch on the redirects and test them Every old URL, one hop, a real page
Launch day Submit the new sitemap So Google finds the new pages quickly
The fortnight after Watch the Pages report Errors show up days before traffic moves
The fortnight after Request indexing for the top pages Speeds up the pages that matter

The rest of this piece is each of those, in that order.

The redirect map is the whole job

One spreadsheet, made before any code, with a row for every URL the old site has. Not the pages you remember. Every URL, pulled from four places, because no single one of them is complete.

Source What it gives you
Your old sitemap What the site thinks it has
Search Console, Pages report What Google has actually indexed
A crawl of the live site What is reachable by following links
Server logs or analytics What real people and crawlers still request

The four lists never agree, and the gaps are the point. A URL in your logs that is in no sitemap is usually an old page somebody still links to, which makes it one of the most valuable rows in the file.

Every row gets a destination on the new site. Then two rules:

A page with no equivalent goes somewhere honest, not to the homepage. A mass redirect of two hundred dead pages to the homepage is treated as a soft 404, so you get the work of the redirect and none of the benefit. The nearest real page, or a genuine 410 if the thing is truly gone and nothing links to it.

No chains. Old to new, one hop. A redirect to a redirect to a redirect slows down how quickly Google moves the old page's standing to the new one.

Your top twenty pages

Sort the Search Console Performance report by impressions over the last three months and take the top twenty pages. These are the ones a redesign can hurt most, and they get special treatment:

  • Leave their body copy alone through the launch. Restyle them, do not rewrite them. The paragraph that looks like waffle is often the one answering the search the page ranks for.
  • Keep their addresses if you can. A page that does not move cannot lose anything in the move.
  • Improve them a month later, one at a time, when you can see what each change did.

Keep the staging site out of Google

The new site needs to be built somewhere, and that somewhere must not be found by Google, or the two sites compete with each other. Password protection is the safest way. A noindex tag or a robots.txt block also works, and is the one most often forgotten on launch day. Write "remove the staging block" on the launch list now, in capitals.

What else has to move, and what must not

Titles and meta descriptions. Carry them over rather than regenerating them from a template, then improve them on purpose afterwards. A redesign that rewrites four hundred titles at once has changed too many variables to learn anything from.

The structured data. Rebuilt, never inherited. An Organization declared twice, or a BreadcrumbList still describing the old hierarchy, is worse than none.

Canonicals and the trailing slash. Pick one form of every URL and use it in the canonical tag, the sitemap and every internal link. Three spellings of one address is how a crawler ends up with three pages.

Internal links. If the new navigation is simpler, make sure the pages that fall out of it are still linked from somewhere strong. A page nobody links to is a page Google stops taking seriously.

Launch day, and the fortnight after

Do it early in the week, never on a Friday. Then, in this order:

  1. Remove the staging block, and check robots.txt and one page's source to make sure it has gone.
  2. Switch on the redirects and test the whole map, not a sample. Every row should answer with one 301 and land on a real page.
  3. Submit the new sitemap in Search Console.
  4. Request indexing for the top handful of pages in URL Inspection.

Then watch the Pages report rather than the traffic graph, because coverage errors show up days before a ranking change does.

Expect a wobble. Two to four weeks of movement in both directions while Google recrawls is normal and is not evidence that something is broken. What is evidence: 404s climbing in Search Console, redirect chains in the crawl, or a specific high-value page losing position while the rest hold. If you see those, the traffic drop guide walks through finding the cause.

A large site can launch in stages

For a big site the safer version is often not one launch but several, a section at a time, each watched for a fortnight before the next. It costs more, and it is worth it when a bad week of traffic would cost more than the difference. How we run that is on our enterprise website development page.

Where to start

Start the spreadsheet before the designer starts the mockups. It is the least glamorous document in the project and the one that decides whether the new site inherits the old one's search traffic or starts again from zero.

If you would rather hand it over, every rebuild we do maps the old addresses before launch, and on the larger ones the redirect map is handed over as a document at the end, so your team has a record of where everything went. That is part of a website development project from the start, not an extra.

Found this useful?
LinkedInX
Nitesh Padghan

Nitesh Padghan

Founder, Storifex Media

I build websites that turn visitors into revenue, and rebuild the LinkedIn profiles that send people to them. Clients in six countries, most of them shipped in two weeks.

Connect on LinkedIn

Related reading