Rankings attach to individual URLs, not to your business and not to your domain. A redesign keeps them when every address that already earns impressions is written down before anything changes, and lands in a single hop on the page that replaced it. Everything below is the detail behind those 2 sentences, in the order it has to happen.
Order is the part people get wrong. Redesigns that lose traffic usually lose it because the mapping was attempted after the new site existed, when the only move left is bending old addresses towards whatever was built.
Record what you have before anyone opens a design file
You cannot protect an asset you have not listed. Before the new structure is agreed, export 4 things and keep them somewhere that is not the old website.
- Every live URL. Crawl the site and export every address returning a 200. The free version of Screaming Frog handles 500 URLs, more than most small business sites have. Your XML sitemap is a starting point rather than a substitute: orphaned pages that still rank are missing from it.
- The Pages report from Search Console. Performance, date range at the full 16 months, view switched to Pages, sorted by impressions, exported. Search Console keeps 16 months and no more, so this export cannot be recreated later.
- Top queries per page. For every page with clicks, record the queries it matches. Queries survive a URL change, so they are how you check afterwards whether the topic came through intact.
- Your linked pages. The Links report lists your top linked pages for free. Any address on it that vanishes without a redirect throws away a link somebody gave you.
Then add the thing no tool exports: which pages your enquiries come from. Those get protected ahead of anything justified by impressions alone. Save the title tag and H1 of every page earning impressions too, because once the old site is gone the only copy is in the Wayback Machine.
Map every old URL to exactly 1 destination
The map is a spreadsheet, and it is the whole deliverable. A row for every old address, 1 destination per row, no blanks. Build it before the new structure is locked, because the map usually reveals pages the new site was going to be missing.
| Column | What it decides |
|---|---|
| Old URL, exactly as served | Protocol, www, trailing slash. The rule has to match what the server returns. |
| Impressions, 16 months | Whether it is an asset. Zero means nobody ever saw it. |
| Clicks, 16 months | Whether it earns visits, which raises it in the queue. |
| Top query | What the page is understood to be about, which narrows the destinations that will do. |
| Current title tag | What earns the click today. Copy it before improving it. |
| Links from other sites | Whether dropping it costs something unrecoverable. |
| Destination and decision | Keep, merge, redirect or retire. Every row gets a call. |
A pair of rules keeps the map honest. If 2 old pages fold into 1, the survivor has to contain what both covered, or you have redirected a query to a page that no longer answers it. And if a page has real impressions but no sensible destination exists, the new site is missing a page.
Single-hop 301s, never chains
Every redirect should be a single hop: old address, 301, final address, 200. Chains appear when a site has been migrated before and new rules are layered over old ones, so a 2019 URL points at a 2022 URL that now points at a 2026 URL. Each extra hop is another rule that can break, another delay for a crawler, and on some hosts a redirect limit that returns an error instead. Flatten them so the oldest address reaches the final one directly.
Invisible hops catch people. If your rule sends /old-page to http://example.com.au/new and the host then upgrades to https, adds www and appends a trailing slash, you have shipped a 4-hop chain believing you wrote 1. Write every destination as the final absolute URL in the exact form the new site serves, and match the old site's trailing-slash and www behaviour rather than changing it in the same launch.
| Response | Use it when | What Google does |
|---|---|---|
| 301 | The page has a genuine replacement. | Indexes the destination, passes the signals across. |
| 302 | Almost never in a redesign. | Keeps the old URL as the indexed address. |
| 410 | Content is deliberately gone, nothing replaces it. | Drops it faster than a 404, with no ambiguity. |
| 404 | The page was never meant to exist. | Drops it eventually. Fine in small numbers. |
| 301 to something unrelated | Never. | Treated as a soft 404. The signals go nowhere. |
Never point everything at the homepage
The shortcut is a single rule sending every retired address to the homepage, and it is the most expensive 5 minutes saved in this process. Google assesses whether a destination is a reasonable replacement for what was requested, and when it is not, the redirect is classified as a soft 404. You can watch that in the Page indexing report. The old page stops counting and the homepage gains nothing.
The same logic rules out the other convenient destinations: a service page does not redirect to the contact page, and a suburb page does not redirect to a booking form. Where no direct replacement exists, use the closest hub that genuinely covers the topic, a category or parent page a visitor would accept as the right place to land. Where content is retired deliberately and nothing covers it, serve a 410.
Keep the title tags that already earn clicks
Rewriting every title on launch day is standard practice and a bad idea for 2 reasons. The first is direct: a title is what matched the query and earned the click, and a version that reads better to you often holds fewer of the words people search. The second is diagnostic. Change the addresses, the templates and the titles in one week and you cannot tell which of them caused whatever happens next.
The working rule: any page with clicks in the last 3 months keeps its title exactly as it is through launch. Improve it a fortnight later, on its own, so the effect is readable. The exception is a title that is pure branding, the business name repeated on a page earning impressions for a service phrase. That one is not earning anything you can lose.
What this looked like on our own website
We ran this process on this site in August 2026 and the findings were not flattering. It carried 90 URLs, 63 of which came out of template functions rather than being written by anybody. The 16-month Pages export said 49 of those addresses had never produced an impression. Measured on rendered HTML rather than source code, the worst-matched pair of pages shared 80.4% of its 7-word strings.
What survived is 36 hand-written pages, and overlap across the site now measures 20.7%. Every retired address was mapped by hand to a single relevant destination in one hop: the individual trade pages to the trades umbrella that covers them, the smaller city pages to the locations hub, and pages with a subject of their own to whichever page inherited it. None went to the homepage.
Here is the honest part. Total indexed pages fell, which is arithmetic rather than punishment: we deleted more than half a website. The removed pages had contributed nothing, so what mattered afterwards was how the survivors performed, not how many URLs remained. Expect that shape. Anyone promising an immediate rise from a structural change is selling something, because positions move while Google recrawls.
Watch Search Console for 8 weeks
Launch day starts the monitoring, it does not end the job. Keep the old XML sitemap reachable for a few weeks alongside the new one: it hands Google the exact list of addresses you want recrawled, so the redirects are found quickly.
- Week 1. URL Inspection on your 10 highest-impression old addresses. Google should report a redirect and name the correct destination. Anything reported as a 404 is a missing map row.
- Weeks 1 to 8, Page indexing. Pages with redirect rising is the migration working. Not found (404) rising means addresses you never mapped. Soft 404 rising means destinations that do not match what was requested.
- Weeks 1 to 8, Performance. Compare by query, not by page. Addresses changed, so the page view splits your history in 2, while the queries you matched before are directly comparable to now.
- Ongoing, your 404 log. Server logs show real requests for addresses a crawl never finds: old PDF files, images linked from other sites, pages orphaned years ago. That is the last 5% of the map.
After 8 weeks, stop attributing everything to the migration. A page that has not recovered by then is more likely held back by its content than by the redirect pointing at it.
The pre-launch checklist
- Full crawl exported, every URL returning a 200 listed.
- Pages report exported at 16 months, sorted by impressions.
- Top queries recorded for every page with clicks.
- Title tag and H1 saved for every page with impressions.
- Links report exported so no linked address is dropped.
- Enquiry sources recorded, so the converting pages are known.
- Mapping sheet finished with no blank destinations.
- Merge destinations checked: each covers what it absorbed.
- Redirect rules written as 301s, tested on the preview URL.
- Every rule confirmed single hop to the exact final address.
- Trailing slash, www and https matched to the old site.
- Titles carried across verbatim on pages with recent clicks.
- Business name, address and phone matching your Google Business Profile down to the punctuation.
- Analytics and Search Console verified before the DNS change.
- Old sitemap still reachable, new sitemap ready to submit.
- A 404 report you can read from day one.
Nothing there is expensive. It is a day of unglamorous work, and it decides whether a redesign carries its history forward or starts at zero while everybody admires the new photography.
How we handle it, and what it costs
We do the mapping on every rebuild before the design is agreed, because the export usually changes what the new site should contain. You keep your domain, and every address with search history gets its own destination. The build costs $0 upfront and takes 5 working days from intake. Once live it is $297 a month including GST on a 12-month subscription, $3,564 across the year, invoiced only after launch and after tracking is verified. If your website has not produced a lead from SEO within the first 90 days, you stop paying us until it does. There is no refund, and we do not promise rankings. The before-and-after gallery shows the rebuilds, and how the build runs sets out the process.
Still deciding whether a redesign is the right move? The honest test is here. And if you are unsure who holds your domain and your analytics, settle that first, because a migration you cannot execute is not a plan.