Redesign a website without losing search traffic
Traffic is lost in a redesign for four reasons, and only one of them is design. URLs change and nothing redirects; text is cut because the new layout has less room for it; the page's internal links are replaced by a cleaner menu that links to fewer pages; and the new build renders its content in the browser instead of the server. Each of these is a decision made during the project, usually without anyone framing it as a search decision. Preventing the loss is therefore not a task you add at the end — it is a set of constraints applied to the design brief before layouts are approved: the URL map is fixed, the content inventory is preserved or deliberately replaced, the internal linking is carried across, and the rendered HTML is verified. A redesign that keeps all four is usually flat or up. A redesign that breaks one of them can drop hard, and the drop appears weeks after the launch everyone already celebrated.
Why does traffic actually drop after a redesign?
In roughly this order of frequency:
- URLs changed without redirects. New CMS, new slug conventions,
/services/seo/becomes/what-we-do/seo/. Every old address that ranked now returns 404. - Content was shortened to fit the design. A page that answered a question in 900 words now has a hero image and 200 words. It no longer answers the question, and it stops being the answer.
- Internal links disappeared. The old site linked deep pages from the footer, from sidebars, from in-body text. The new one has a tidy five-item menu, and a third of the site is now two clicks from nowhere.
- Rendering changed. The old site sent HTML. The new one sends an empty shell and builds the page with JavaScript. What crawlers see is not what a visitor sees.
- Everything shipped at once. When design, CMS, hosting and URL structure all change in one release, a drop cannot be attributed, and therefore cannot be fixed quickly.
Note what is not on the list: visual style. Redesigns do not lose traffic because the new look is different. They lose it because of the four mechanical changes above; the fifth line is not a cause of the drop but the reason a drop cannot be attributed.
What has to be measured before the design starts?
Four inventories, taken while the old site is still live. After the switch they cannot be reconstructed, which is why the sequence matters more than the effort.
- URLs with their performance. Every indexed address, with impressions, clicks and inbound links.
This is the list that later proves whether anything was lost.
- The content of the top pages, saved. Not a screenshot — the actual text, so the new page can
be compared to it word for word.
- Titles, descriptions, canonicals and structured data, exported as data.
- The internal link graph. Which pages link to which, and how many links each page receives.
This is the inventory teams skip, and it is the one that explains slow post-launch decay.
Should URLs change at all?
Default answer: no. Keep them. URL structure is rarely worth the risk on its own, and "cleaner URLs" is not a business outcome.
When URLs must change — because the information architecture genuinely changed, or the old structure encoded something that no longer exists — then each change is a row in a map: old address, new address, reason. The map ships with the release. Chained redirects are collapsed to one hop. Redirects go to the equivalent page, not to the homepage: a bulk redirect of 400 deep pages to / is functionally identical to deleting them.
How do you keep the content when the new design has less room?
By deciding, page by page, whether the content is being reduced or moved. Those are different things, and only one of them is safe.
Moving is fine: an FAQ that becomes an accordion is still on the page, still in the HTML, still readable. Reducing is a content decision with a search consequence, and it should be made explicitly by whoever owns the traffic — not implicitly by whoever owns the layout.
The practical constraint for the design brief: for each of the top pages, the new template must accommodate the existing content. If it cannot, the template changes or the content loss is approved in writing before the page is built. In practice this single rule prevents most of the damage, because it turns an invisible side effect into a visible choice.
How do you check the new site before it goes live?
On staging, with the same four inventories used as the reference. Five checks, all of which either pass or name what failed:
- Crawl staging and diff against the URL inventory. Every old URL resolves to a 200 at the same
address or a single 301 to its mapped target.
- Fetch the rendered HTML with JavaScript disabled. The main text has to be present. If the page
is blank without JS, the rendering approach is a search decision and needs to be made deliberately.
- Compare titles and descriptions against the export. Templated titles are the usual casualty.
- Recount internal links to the pages that mattered. Pages that lost most of their inbound
internal links get them back, from body text if the navigation genuinely has no room.
- Check
robots.txtandnoindex. Staging sites are blocked from crawling by design, and the
block travels to production more often than anyone admits. This is the single most expensive one-line mistake in a launch.
What happens in the first weeks after launch?
Some ranking movement is normal while pages are re-crawled and re-evaluated; a change that settles within a few weeks is not the same as a loss. What distinguishes them is whether the four inventories still reconcile.
So the post-launch routine is short and mechanical: recrawl and diff the URL list, watch server logs for 404s from crawler user agents (they find broken addresses faster than any report), and compare page-level traffic against the pre-launch baseline for the pages that mattered. If something dropped, the inventories say which of the four causes it was — which is exactly why they were taken.
Can a redesign be launched in stages?
Yes, and it is the lower-risk path when the site is large. Templates ship in groups — one section at a time — so a drop is attributable to a specific change instead of to the whole project. The old and new templates coexist for a period, which is a cosmetic cost paid in exchange for a diagnosable one.
Where a single switch is unavoidable, the equivalent protection is a rollback path: the old site kept intact and reachable until the post-launch checks come back clean. A launch without a named rollback is not a launch plan, it is an expectation.
What do you need to start?
The URL inventory with performance data, access to the current CMS, and a decision about whether URLs may change. Those three determine whether this is a low-risk redesign or a migration with a redesign attached — and the answer changes both the schedule and the price.
Who does this: Toimi is a digital agency that redesigns websites without losing their search traffic. We work in Russian and English and have been in business since 2017. Send the three inputs above and we will come back with scope, schedule and price.
Contact: info@toimi.pro