Two very different projects get sold under the same word. One changes how your site looks. The other rebuilds what sits underneath it and has to inherit every URL, ranking and backlink you already have. The second is where traffic disappears, and where the agency you pick actually matters. What follows is how the work runs when it is done properly, and how to tell whether the people quoting for it have done it before.
Redesign, rebuild or replatform: work out which one you are buying
The word redesign covers three projects with very different risk profiles. A redesign changes the visual layer: typography, hierarchy, imagery, the conversion path. The platform, the URLs and the content model stay put, so the only search risk you carry is what the new front end does to page weight and rendering. A rebuild replaces what sits underneath: new templates, new routing, a new content model, and every URL has to be accounted for. A replatform moves you off one system onto another, which adds data migration on top of all that.
You need a rebuild when the constraint is structural. The usual signs: time to first byte dominates your LCP and no amount of caching fixes it, the plugin stack cannot be updated without breaking the layout, pages are page builder shortcodes rather than structured content, mobile has been patched with overrides for years, and nobody can publish a new page type without a developer.
You do not need a rebuild when the information architecture is sound, the CMS handles your publishing volume fine, and the real problem is that the site looks ten years old and buries the thing you actually sell. That project is smaller, shorter and far less likely to cost you traffic. If a reskin solves it, buy the reskin. We will say so in the estimate rather than sell you the larger project.
Everything that happens before a single screen is designed
The first deliverable on a rebuild is not a mockup. It is an inventory. We crawl the live site, pull the Search Console pages report across the full sixteen month window the interface allows, export analytics landing pages, export referring domains from a backlink tool, and read the server logs for the URLs a crawl never reaches: orphan pages, old PDFs, campaign landing pages, parameter variants. Every URL is scored on clicks, impressions, average position and referring domains. That list, not anyone's taste, decides what is kept, what is merged and what is genuinely dead.
Out of it comes the redirect map: one row per old URL, exactly one target, a 301, no chains, no catch all to the homepage. The catch all is the most common way to turn a redesign into a traffic loss, because a redirect to an irrelevant page gets treated as a soft 404 and the old URL's signals are dropped. The map also has to settle the things that break silently: trailing slash convention, case sensitivity, query strings that must survive, www versus apex, and http to https, so that no request ends up in a two hop chain.
The map is signed off before design starts, and regenerated from the actual build output before launch, not from the spreadsheet someone last edited two months earlier.
The ways a redesign quietly destroys traffic
Almost every catastrophic redesign is one of a short list of mistakes, and all of them are preventable.
Staging robots directives shipped to production. A meta robots noindex, an X-Robots-Tag header, or a Disallow: / in robots.txt that was correct on staging and fatal on the live domain. The header version is the nastier one, because it never shows up in the page source. Check the response headers, not the HTML.
Canonical tags left pointing at the staging or preview domain, which quietly tells Google that the page it just crawled is not the real one.
Pages deleted because they looked dated. The ugly old post nobody is proud of is often the one holding half the referring domains. Design opinion is not a deletion criterion. The inventory is.
Slugs changed after sign off. Someone renames a page in the CMS the week before launch and the map goes stale without anyone noticing. The only real defence is a release gate: diff the URL list the build produces against the map, and fail the deploy on any unmapped route.
Structured data dropped. Product, Article, BreadcrumbList, FAQPage and LocalBusiness markup rarely survives a template rewrite. Rich results vanish, click through rate falls, and it reads like a ranking loss when it is a SERP feature loss.
Internal links left pointing at old URLs, so every navigation click runs through a redirect. And analytics rebuilt without goal parity, so you cannot prove whether anything improved.
Launch week, and what gets measured for the two months after
On cutover day the old site is crawled one last time and every URL in the map is requested against production, to confirm it returns a single 301 landing on a 200. Robots.txt, meta robots and the X-Robots-Tag header are verified on the live domain, not on staging. The new sitemap is submitted, and the old one is kept reachable for a few weeks: it is the fastest way to get the old URLs recrawled, so the redirects are seen rather than discovered slowly.
Then the monitoring, which matters more than the launch itself. The Search Console page indexing report, watching the Page with redirect and Excluded by noindex counts. Server logs for the 404s the map missed, which is where the real user paths surface. Position tracking on the URLs that carried the traffic, not on vanity keywords. Core Web Vitals from field data, which arrives on a rolling 28 day window and therefore says nothing trustworthy for about a month, with lab measurements covering the gap from day one.
Expect movement. Individual URLs oscillate for a few weeks while signals are recrawled and reconsolidated, and that is normal. What is not normal is a whole directory dropping out of the index, or impressions falling in a straight line. Those have causes, they are visible in the logs and the indexing report, and they are fixable if somebody is actually watching.
Two people in Dijon, everything in writing, whatever your time zone
Go To Agency is two people. Robin Monteiro builds (Next.js, React), Florian Loppion handles marketing and SEO. One office, at 9 rue Jean-Jacques Rousseau in Dijon, France, and no other. That is the honest version. What you get in exchange for the size is direct access to whoever is doing the work, with nobody turning your request into a ticket first. The references we can name are French SMEs: Chouchou Ribeyre, Au Petit Detail, LB Athletic, Mediavocats and Vectosolve.
We work in writing only. No calls, no video calls, no scheduled slots, and a reply within 24 business hours. For a client in California or Bangalore that removes the worst part of hiring a European agency: nobody takes a 3am call, and nothing waits for Paris to wake up before it can be decided. It also means the redirect map, the kill list and the reasoning behind dropping a given page stay searchable a year later, when someone new asks why an old URL went away.
What changes on your side: send requests with the specifics attached (URL, screenshot, browser, what you expected), batch your feedback instead of firing off fifteen messages, and nominate one person who can decide.
Content and access set the real schedule. A migration script moves existing pages in an afternoon. Waiting on rewritten copy, product photography, legal text or registrar credentials is what stretches a project, and it is the one part we cannot do for you.
When a redesign is the wrong purchase, and when not to hire us
If your traffic is healthy and the problem is that visitors do not convert, a redesign is a blunt instrument. It resets everything you have learned about the current site in one move. Fix the form, the page speed and the clarity of the offer first, measure, then redesign if it is still warranted.
If you run a small brochure site with no rankings and no backlinks to protect, hire nobody. Squarespace, Wix or Webflow will do the job on a monthly subscription, roughly in the tens of dollars or euros depending on the plan at the time of writing, plus the domain. For straightforward retail, Shopify covers hosting, checkout, PCI scope and payments for a monthly subscription plus payment processing on every order, and rebuilding that from scratch is usually money badly spent.
If your marketing team has to assemble new page layouts every week without a developer, a custom Next.js build can be worse than a well configured WordPress or Webflow, unless you also pay for the editing experience that makes it usable.
If your content sits in a closed platform with no usable export, the work is dominated by extraction rather than design, and staying put may be the rational call.
And if you want on site workshops, a pitch presentation, a dedicated account manager or someone to call at six in the evening, we are the wrong agency. Two people, asynchronous, in writing, and that is not going to change for one project.
