The short version
- The situation: a 2,011-page site where 90% of SEO requests needed a developer, moving to a new platform across several markets.
- The move: SEO written into the requirements from the first scoping week, 290 SEO tickets triaged daily by what would block the launch, and management told what to expect before launch day.
- The result: the month after launch, organic sessions rose 9%, lower-funnel sessions 73% and lower-funnel conversions 21%.
- The lesson: most of the 290 tickets could wait. Deciding which ones could not was the work.
The situation
I led SEO at Mollie, the Amsterdam payments company, when it rebuilt mollie.com in 2022. The business had four reasons for a new site: a new brand, marketing that could publish without waiting on developers, better localisation for its markets, and a site built with SEO in mind from the start.
The old site had been called SEO compliant by people who were not SEOs. In practice, 90% of SEO requests needed a developer. A new landing page template took two or three developers plus UX and design two to three weeks. Every redirect, which sends visitors and Google from an old page address to the new one, was a one-off developer ticket. A page listing every page on the site took seven months to ship. The one page type built for search grew 645% in a year; most other page types stood still or declined.
A relaunch fixes that, or it wipes out years of rankings on launch day. The goal I set: how often the site showed up in search down no more than 10%, and no dip at launch or in the two to three months after.
The SEO side was me, one colleague and an external SEO agency, working with Mollie's web team (a product manager and three developers) and an outside development agency. SEO was involved from January 2022, a month before scoping started. The plan said six months from scoping to launch. Launch came on 31 October 2022.
What I did
| The move | Why it mattered |
|---|---|
| SEO requirements in the first scoping round | URLs, header and footer, homepage, hreflang (the setting that shows each country its own language) and canonicals (a setting that tells Google which version of a page counts) were written down as requirements for developers, not raised after the templates were built. I tagged the owner of each one, held one-to-ones every two weeks and followed up until it was done. |
| A content audit before anything moved | 20% of pages did not move to the new site. The cost, measured before the decision: 1% of organic traffic and one conversion. |
| The test version scanned like the live site | I ran a scan of every page on the test version of the new site before launch and took the findings to the developers with data attached, so every fix request came with a reason. |
| Four priority levels, reviewed daily | 290 SEO tickets, 31 of them critical. "Highest" meant I would block the launch (wrong canonicals). "High" meant I might, but the web team could choose to go ahead (pages missing from the new site). "Medium" had to be fixed within two days of launch (faulty redirects). "Low" meant someday (URL localisation, imperfect copy in secondary markets). Close to launch I re-ranked the list several times a day. |
| Management told what to expect before launch | Four weeks out, a call with the CMO, CCO, CTO, CPO and CEO on what a relaunch does to organic traffic. The day before, where things stood and how confident I was. Daily emails the first week, weekly after that, and a results deck a month in. |
What happened
Measured the month after launch against the month before:
- +21% on the lower-funnel conversion point
- +73% organic sessions on lower-funnel pages
- +9% organic sessions overall, +7% in the Netherlands, the largest market
- +2% in how often the site showed up in search each month, against a plan that allowed a 10% drop
- 0.5% of non-brand queries dropped out of positions 1 to 3
The site underneath changed too. Most pages now sit three to four clicks from the homepage, where a large share used to sit ten or more deep. The technical settings, from the page lists Google reads to hreflang, canonicals, breadcrumbs and URL mapping, are handled by the platform, not by a developer ticket. Redirects moved from the tech team to the people who own the content.
The work won the Friends of Search award for best SEO case in 2023.
What I'd tell you honestly
Plenty went wrong, and the launch held anyway. I had no access to server logs. The keyword research planned for the product pages never happened. My team worked out URL localisation, and development dropped it late in the project. Content was frozen for six weeks before launch. The launch date was set on a Friday, for the following Monday. None of that showed in the numbers, because the tickets that could hurt performance were fixed first.
"We handle SEO" is not a plan. The development agency said it handled SEO because it had installed Yoast. If you are rebuilding with an agency, ask them what they check and when, and have someone on your side who knows the answer. A lot of time would have been saved if the agency had had its own SEO consultant.
Most of the 290 tickets could wait. The hardest part was letting go of SEO perfectionism: questioning every ticket on whether it would hurt performance or only bother an SEO. Developers stay on side when they help make those calls instead of taking orders, and when they care enough about SEO to worry about breaking it.
This site has since been replaced. Mollie moved to a new CMS and design again in February 2024. This story covers the 2022 relaunch only.
The new website SEO checklist is the same method, written down for a site you are about to move.
I’m a revenue-first SEO consultant for B2B and SaaS teams. 14+ years of hands-on work, from in-house at Mollie to new websites for Backbase and Rentman. I also run the Amsterdam SEO Salon.
Follow me on LinkedIn (opens in a new tab)
