Table of contents
The riskiest moment in a website's life
A site that has been online for years has accumulated something that never appears in the design mockups: indexed pages, links earned from other sites, searches it shows up for, trust built slowly. All of it is attached to one thing, mundane and fragile: the address of every single page.
A redesign puts exactly that up for grabs. The menu structure changes, pages get merged, others are dropped, the new system generates different addresses. Meanwhile Google keeps sending people to the old addresses — and if there is nothing there any more, those people see an error page and leave.
The typical result is a new site that works beautifully for people who already know you and disappears for everyone else. The drop does not arrive on launch day: it arrives over the following fortnight, as Google recrawls and finds the gaps. Which is why it is so often noticed too late.
The good news is that this is not a matter of luck, or of how good the designer is: it is a matter of preparation, and nearly all of it happens before a single line of the new site is written.
Before touching anything: the baseline
One rule governs everything else: you cannot protect what you have not measured. Before the old site is switched off you need a record of how it is doing, and it has to be made now, not afterwards.
Four things are needed. The full list of live addresses, taken from the sitemap and from a crawl of the site, not from the menu: the pages that bring traffic are often the ones that dropped out of the menu years ago.
The pages that bring visits and enquiries, from Search Console and Analytics over the last twelve months. Not all pages are equal: usually a dozen of them account for most of the traffic, and that is where the attention goes.
The searches the site appears for, with impressions, clicks and average position. This is the yardstick you will judge the aftermath by: without the before, any post-launch number is an opinion.
The links received from other sites, because an address that has earned links is an address you cannot afford to break: that value only transfers if the redirect is in the right place.
Keep the data outside the tools. Search Console holds sixteen months of history, and Analytics can be reconfigured during the project. Export everything to a spreadsheet before you start: the day you need to prove what the traffic really was in September, that file will be the only evidence you have.
The redirect map
This is where a migration is won or lost, and it is also the step most often skipped when time is short. The redirect map is a two-column list: every old address, and the new address that replaces it. No exceptions.
The redirect to use is the 301, permanent: it tells search engines the move is final and that the accumulated value should transfer to the destination. A 302, temporary, leaves everything in limbo — a common mistake that produces exactly the symptoms of a botched migration. Google's own documentation on site moves with URL changes is explicit about this.
Two rules matter more than all the others. A redirect must point to the page that genuinely corresponds: sending thirty product pages to the homepage is the fastest way to lose all thirty, because to Google it reads as those products no longer existing. And it must be direct: if A redirects to B which redirects to C, you add hops, lose loading time and occasionally break the chain halfway.
Then there are the pages that simply disappear, because the service no longer exists. If there is no sensible replacement, the correct answer is not some arbitrary redirect but a 410: it says the content was removed deliberately, and closes the matter cleanly.
The map is prepared beforehand, tested on a staging copy, and goes live at the same moment as the new site. Not the day after.
What to carry over, what you can change
A redesign is an opportunity to change things, and it would be foolish not to use it. But it helps to know which changes carry a ranking cost and which do not.
Copy can be rewritten, not halved. This is the most common temptation: the new site is clean and minimal, so pages of fifteen hundred words become three paragraphs. Except those words were the reason the page appeared at all. Rewrite them better, reorganise them, but keep the substance.
Page titles are an asset. The title shown in search results and the main heading of each page are the summary of what that page is. If the new site replaces them all with slogans, you have changed the subject without meaning to.
Images should be regenerated, not just re-uploaded. This is the right moment to move to modern formats and sensible dimensions: speed is one of the few areas where a new site starts ahead, and it is worth not wasting that. If the old site was slow, here are the seven most common causes.
Design, colours, menus, layout: all free. None of these carry ranking value with them. That is the paradox of a redesign: the part the client looks at is the part that matters least to Google, and the part nobody looks at — the list of addresses — is the part that decides everything.
Launch day: the checklist
The order matters as much as the content. In sequence, before calling it live:
- Remove the block on search engines. During development the staging site is almost always closed to indexing. Forgetting that one line is by some distance the most frequent and most expensive mistake: the site is perfect and invisible, and it gets noticed weeks later.
- Deploy the redirects with the site, not after it. Every hour of gap is traffic hitting an error page.
- Test the twenty most important addresses by hand. The ones your baseline says carry most of the traffic. One by one, checking where they actually land.
- Check the certificate and the single canonical domain. Everything must resolve to one form of the address, over https. The versions with and without www must converge, not coexist.
- Generate and submit the new sitemap in Search Console, so crawling starts immediately instead of waiting.
- Verify that tracking works: Analytics installed, goals configured, contact forms tested by actually sending an enquiry. A broken form throws no error — the enquiries simply stop arriving, and you find out when the phone goes quiet.
And one thing not to do: do not switch off the old hosting straight away. Keep it for a few weeks. If a page turns out to be missing, having the original to hand turns an emergency into ten minutes of work.
The first four weeks
After launch comes the part almost nobody does, and it is the part that separates a migration you managed from one that happened to you.
Week one, check the not-found report in Search Console every day. Every error that appears is an old page someone is still looking for: add it to the map and fix it. It is normal for some to surface that were not on the original list — those are the ones reached from external links, old emails, printed documents.
Weeks two and three, watch indexing: how many new pages Google has taken, how many are still missing. The number has to climb. If it stalls, something is blocking crawling and needs finding straight away.
At week four, compare traffic against the baseline — and for the first time the comparison means something. Before that the numbers swing too much to say anything.
Throughout that period one rule applies: change nothing else. If pages are added, menus moved and copy rewritten while the site is settling, you will no longer be able to tell what caused what. A redesign is judged with the site standing still.
When a dip is normal and when it is a problem
Some dip is almost inevitable, even when everything was done right: Google has to recrawl every page and recalculate, and until it has finished the numbers wobble. The difference is in how much and for how long.
A modest dip that starts recovering within a few weeks, with impressions moving before clicks, is normal. A sharp fall that shows no sign of recovering after a month, or a drop concentrated on a handful of pages that used to perform, is not.
When it happens, the causes are almost always the same four, in this order of frequency: the indexing block left switched on, missing redirects or redirects all pointing at the homepage, content cut in the move to the new site, and duplicate addresses making two versions of the same page compete. All four are diagnosable in half a day — if you have the baseline.
The point, in the end, is that a redesign is not a design project with a little technical care bolted on: it is a house move, and like every house move it is judged by how much arrived intact. The nice part — the new site — is the part you see. The part that decides whether it works is a two-column spreadsheet nobody will ever look at again.