Workshop · 1 to 2 days
Relaunch Workshop.
No relaunch loses zero traffic. The costly decisions fall in the spec, months before go-live. In one or two days your product owner and agencies settle URLs, redirects, CMS rules and the launch checklist.
For
- Product owners
- Project leads
- Dev and design agencies
- SEO leads
Is this you?
- A new CMS, domain, frontend or merger on the roadmap
- Agencies building without checking what Google will index
- Wireframes and tickets that say nothing about SEO
- A launch day nobody owns
A relaunch is decided in the spec
Google does not see your project plan. It sees old URLs that redirect and new URLs it has never ranked. What those URLs look like, what the pages contain and what the server answers is written down months earlier, in wireframes, tickets and a CMS spec.
Google on how long the move takes: “As a general rule, a small to medium-sized website can take a few weeks for most pages to move, and larger sites take longer.” (Google Search Central, How to move a site)
A URL decision costs one sentence in a spec. After go-live it costs traffic. That is why I want the product owner and both agencies in one room before the build. My line at one kick-off: “The goal is not zero loss. The goal is controlled loss.” In a deck for a CMS move I added that it is “very very very likely that some traffic loss will happen”. The workshop keeps it small and short.
What happens when the spec skips all this, I measured after go-live. That is on the Relaunch SEO page. This page is about what the workshop finds earlier.
What you leave with
Five things, built in the room from your own data:
- The URL migration matrix. Every old URL from every source I can get: Search Console, sitemaps, server logs, the CMS. Ordered by clicks. One rule and one expected answer per row: a 200, a 301 to a page that answers 200, or a deliberate 410.
- Redirect rules. Written as rules your server can run, each with the test that proves it.
- The CMS SEO spec. Page type by page type: title, H1, canonical, structured data, internal links, sitemaps, robots.txt. The CMS team builds against it.
- The launch checklist. Launch day, the first 48 hours, the first two weeks, the first three months. Every task has an owner and a priority.
- Annotated wireframes. Each wireframe marked with the HTML element every part must be, and the function behind it.
What the workshop finds in a relaunch spec
These are the gaps I found in the specs, wireframes and staging builds of my recent relaunch workshops. Each is cheap to fix in a spec and expensive after go-live:
| Where it breaks | What I found in the room | What the team specifies |
|---|---|---|
| Design that is not there yet | A third of the planned pages had no wireframe; no mobile design at all | A phone wireframe per page type before the template spec is signed |
| A first screen that is already full | About 650 px on a phone; no product name above the fold; banner, location and assistant widget unruled | One written rule per overlay, tested with a cold-load phone screenshot |
| Design, markup plan and HTML disagree | Ratings in the design, none in the markup plan; list tiles that were not links | Links are <a href> in the server HTML; one markup row per design element |
| A URL paper nobody can measure | Every page type at the root, the headline as slug | A short prefix per page type; a slug that never changes |
| A matrix that treats all URLs alike | About 30k old URLs, none unchanged; one table carries 60 percent of clicks | Rules ordered by clicks; the top 4k URLs checked first |
| Days where it takes weeks | The top 4k URLs need 2 weeks, all 30k 7 weeks to over 3 months | A launch plan with judgement points at week 2 to 3 and month 3 to 4 |
| A staging nobody can test | On the new CMS staging the banner was the largest element, the header needed JavaScript, blocking time was close to 2 s | Staging open for one URL per page type; a failed test blocks go-live |
| A launch day without owners | About 30 launch-day tasks, each needing an owner | An owner and a date on every task before the room empties |
Design that is not there yet
In a one-day spec review I went through about 35 wireframe screens, rendered three ways: phone above the fold, phone full page, desktop. I counted what was drawn against what was planned.
- A third of the planned pages had no wireframe. Most of the overview pages and two more templates were missing, the pages that carry the URL and markup decisions.
- The design file’s page for list and detail pages was empty. There was no list page design anywhere.
- No frame in the design file was a mobile width. Every one was drawn for desktop.
Google indexes the phone version: “Google uses the mobile version of a site’s content, crawled with the smartphone agent, for indexing and ranking.” (Google Search Central, Mobile-first indexing) Every question about the first screen, the H1 or the interlinking has to be answered against a phone. The only phone record in that project was the one I rendered myself.
What I specify: a phone wireframe for every page type before the template spec is signed. Where a template is missing, one sentence on record: same as page type X, yes or no.
A first screen that is already full
I render each wireframe on a phone and draw the fold. On an iPhone in Safari that is about 650 px. On the category wireframe, the top bar, logo, account icons, a selector, menu, search, breadcrumb, H1, intro text, filter chips and sort control filled it. The fold cut through the first product tile. No product name and no price sat above the fold.
Three more things were coming, and nobody had ruled on them:
- A cookie banner. It was in no wireframe at all.
- A selector. It sat in the header of every screen, with no rule against a location prompt on load.
- A floating assistant button. It was on almost every screen and covered the second product tile on the category screen.
None of the three was among the agency’s roughly 35 open questions for the day.
What I specify: one written rule for each. The banner is no full-page blocker, covers neither the H1 nor the largest element and adds no layout shift. No location permission prompt on load. The assistant loads on the first tap. The test is a cold-load phone screenshot that shows content and no overlay. The wireframe renders I took before anyone added a banner are the baseline to compare staging against.
Design, markup plan and HTML disagree
An audit of the wireframes’ HTML:
- On the list wireframe, the product tiles were not links. The whole screen had no link to a product page. Most likely an export artefact, but it is the failure that sinks a migration. Google: “Generally, Google can only crawl your link if it’s an
<a>HTML element (also known as anchor element) with an href attribute.” (Google Search Central, Link best practices) - The product page showed a star rating and a review count above the fold. The agency’s markup plan had no rating markup. Either the ratings are real data and belong in the markup, or they come off the design.
- No screen had a main landmark. Two page types each had more than one H1.
What I specify: a commitment on record that tiles render as <a href> in the server HTML, tested with JavaScript off. One markup row for every data element in the design. One H1 per page, the biggest headline on the phone’s first screen.
A URL paper nobody can measure
A URL paper from the project team came to one workshop. It got the idea right: one URL per resource, independent of categories. But it put every page type at the root and used the headline as the slug.
- No page type in the URL means no filter in Search Console by page type. After launch you cannot tell which part of the site lost traffic without going through URL by URL.
- A headline as slug means the URL changes when an editor fixes the headline, or it stops matching the page.
What I specify: a short prefix per page type. It costs a few characters and makes every page type its own filter. A slug that is set once at publication and never changes, whatever happens to the headline. The rule behind it is in my book.
A matrix that treats all URLs alike
The old-to-new matrix of one relaunch started with about 30k old URLs from Search Console (16 months), the sitemap and a crawl. None survives unchanged. Ordered by clicks, it looks like this:
- One lookup table, for one page type, carries about 60 percent of the clicks: about 31k of 51k clicks in 28 days.
- Three more page types lift it to over 90 percent: about 47k of 51k clicks in 28 days.
- About 250 URLs carry half of all clicks in 16 months, about 4k carry 90 percent, out of about 25k URLs with clicks.
Before the workshop I checked all of those 4k on the old site. Almost 96 percent answered 200. About 30 answered 404, with about 8k clicks in 16 months. About 140 were redirects, some straight to the home page, with about 20k clicks. The old site already had holes, and a matrix built from the sitemap alone copies them.
What I specify: rows ordered by clicks, one rule per page type, each with its expected answer. The top 4k URLs checked first. Dead URLs get a 410 or a rebuilt page, never a redirect to the home page. Embedded files get rows too: “Make sure that you include URLs of embedded content in your site move plans: videos, images, JavaScript, and CSS files.” (Google Search Central, How to move a site)
Days where it takes weeks
Every manager asks how long until Google has digested the move. I answer with arithmetic from the Search Console crawl stats of a site with about 30k URLs:
- Googlebot made about 6k requests a day. About 2k were HTML. If the share of 200 answers holds for HTML too, about 1.3k of those answered 200. If half of those are repeats of popular pages, that is about 600 new pages a day. Both steps are my assumptions.
- Every migrated URL is fetched twice: the old one to see the redirect, the new one to index.
- The top 4k URLs need about 8k fetches: about 2 weeks at 600 a day.
- All 30k need about 60k fetches: 7 weeks if every successful HTML request counts as new, over 3 months at 600 a day.
Google’s “a few weeks for most pages” above is the same order of magnitude. About two fifths of the requests went to one file type that was neither a page nor an image. Clearing that before launch frees budget the move needs.
What I specify: a launch plan with the old URLs in a redirect sitemap. Judgement points at week 2 to 3 for the top URLs and month 3 to 4 for the long tail. Launch week is not the verdict, and everyone who reports on the launch agrees on that before it.
A staging nobody can test
My three SEO tests and Google’s tools need a URL they can open. Staging usually sits behind a login or a blanket Disallow, so nobody runs the tests. Google: “Don’t forget to remove any noindex or robots.txt blocks that were only needed for the migration.” (Google Search Central, How to move a site)
In one CMS-move workshop the new CMS stood on a staging host that was closed to everything except one URL per page type. I ran the tests there, next to the live site:
- The cookie banner was the largest element on the page, so it set a late LCP.
- The header needed JavaScript to render.
- The hero image was lazy loaded.
- PageSpeed Insights reported a total blocking time of close to 2 seconds.
My slide for that day: “DO NOT MIGRATE IF ANY OF THE 3 SEO TESTS ARE NOT MET.” The three tests are PageSpeed Insights, the page with JavaScript off and Google’s live test, see SEO basics.
What I specify: staging closed to everything except one URL per page type, a list that grows. The three tests on each. A failed test is a ticket that blocks go-live. And a written list of the blocks to lift on launch day: “make sure you prepare what the robots.txt file should look like once the site move starts.” (Google Search Central, How to move a site)
A launch day without owners
My launch-day list for a CMS move had about 30 tasks, almost all marked must. Each has an owner (developer, SEO, editorial or project manager) and a priority. A few of them:
- Switch DNS, then activate the redirects, in that order.
- Spot-check the top 100 URLs by traffic: right target, answer 200.
- No accidental noindex, no Disallow over a section.
- Write down the exact launch time in UTC, so the Search Console data can be read later.
After it come about 15 tasks for the first 48 hours and about 17 for the first two weeks.
What I specify: an owner and a date on every task before the room empties. Scripts, not eyeballs, for the redirect check: “You can use the URL Inspection Tool for testing individual URLs, or command line tools or scripts to test large numbers of URLs.” (Google Search Central, How to move a site)
How the days run
- Before the day. You send full Search Console access, read access to analytics, the spec and wireframes in writing and your team’s questions. I read the data before I look at the site. In one spec review the agency’s question list had about 35 questions. About 10 of my answers differed from its plan.
- Day one: strategy. Traffic today by page type, brand and non-brand. The three SEO tests on the old site and on staging. A risk assessment with a best, a likely and a worst case, each with a loss at launch and a recovery time. URL rules, hreflang across markets, retiring a mobile subdomain, Google News and Discover where they carry traffic.
- Between the days. Homework: the decisions the team must take before day two. In one CMS move the two days were a few days apart.
- Day two: live specification. The pre-launch task list, the CMS SEO spec page by page, sitemaps and robots.txt, the launch checklist, monitoring, a risk register with escalation paths. Every item is written in the room and marked must, should or nice. The developers who build must attend.
- Handover. The slides as PDF, the task lists with owners, the specs, the migration matrix and redirect rules, a reference page per page type that the agencies build against, the monitoring plan.
- After. The staging audit and the launch check are part of Relaunch SEO. Search Console two weeks and two months after launch. If traffic drops anyway, the Post Relaunch Workshop.
One day or two? One day is a spec review, when about two thirds of the technical spec is set and can still change. Two days fit a CMS, domain or merger move with many URLs: strategy first, specification second.
This site, launched with the same checklist
When I moved this site on 21 October 2026, the launch checklist was a script. The migration matrix has about 550 rows. Every row carries an expected status and, for redirects, an expected target. On launch day the script fetched every row without following redirects and compared. About 650 checks on the live host, 0 failed.
Try one: an old WordPress thumbnail URL answers with a 301 to /i/autocomplete.png, in one hop. How the 550 rows were found is on the Relaunch SEO page.
Questions I get before a relaunch workshop
Who should be in the room?
The product owner, the project lead, your SEO lead, the design agency and the dev agency, with the developers who will build. Without both agencies, every decision is made twice. In one spec review the agency’s own open questions were waiting for exactly these SEO decisions.
When is the right time?
Before the first wireframe for the strategy day. Mid-build, when about two thirds of the spec is set, for the review. Not after go-live: then you need the Post Relaunch Workshop.
What do I need to bring?
Full Search Console access, read access to analytics, the current spec and wireframes in writing, and your team’s questions.
Will we still lose traffic?
Yes. How much, and why I budget for it, is under How much traffic will we lose? on the Relaunch SEO page. The workshop’s job is a loss that is small and short.
Is this the staging audit?
No. The workshop sets what staging is tested against. The audit runs the tests on the build, and the launch check runs them again on go-live day.
Can it run remote?
Yes, on site or remote.
Sources for the Google quotes on this page, read on 2026-11-12: How to move a site (last updated 2026-08-20), Mobile site and mobile-first indexing best practices and Link best practices for Google (both last updated 2025-12-10), Google Search Central. The numbers come from my own relaunch workshops and handovers in 2024 to 2026.
Some of our customers
Examples, not the full list
From real workshops
- 2026: a one-day spec review of about 35 wireframe screens, a migration matrix of about 30k old URLs, the three SEO tests as a launch gate.
- 2026: two days for a CMS move, about 100k old URLs mapped, a 200-point checklist.
- A CMS migration workshop: the three SEO tests on the new CMS staging, a redirect plan, post-launch tests.
- 2016: relaunch workshop with an old-to-new migration strategy.
Consulting for the next step
How consulting worksOther workshops
All workshop formats- SEO & GEO Workshop
- AI First Workshop
- Technical SEO Workshop
- AI Computer Use Workshop
- AI Content System Workshop
- Content SEO Workshop
- AI SEO Workshop
- Web Performance Workshop
- Post Relaunch Workshop
- SEO Concept Workshop
- Newsroom SEO Workshop
- Google Discover & News Feed Workshop
- AI Newsroom Workshop
- International SEO Workshop
- Growth Workshop
- Search Console Workshop
- AI Agents Workshop
- AI Thinking Workshop
- Dev Thinking Workshop
- SEO Training for In-house SEOs
- SEO Setup Workshop