Case study

Shipping the replatform across thirteen markets

5 minute read

Client

Eurocamp (European Camping Group)

Year

2023

Role

UX Designer, acting as product owner

Contribution

Coordinated user acceptance testing, ran defect triage with the CMO, built the CMS content model, and trained the market content teams

Team

UK, NL, DE and PL market teams tested and used it · CamdenTech built the front end · Profound Digital wrote and uploaded SEO content · a digital marketing agency and overseas teams set up the additional domains · CMO and CRO manager set priorities

Platform

Web · responsive · 13 markets

Designing the replatform was half the work. The other half was getting it live across thirteen markets without a launch date to hide behind — because the old platform’s licence was expiring and the new site had to earn its traffic rather than be given it. I coordinated acceptance testing across four market teams, ran weekly defect triage with the CMO, built the CMS content model, and trained the people who would live in it.

The people who tested the new site were the people who ran the old one.

Acceptance testing ran with individuals from the UK, NL, DE and PL teams — colleagues whose day-to-day work was the Tridion CMS we were replacing. They knew the old system’s every workaround, which made them the sharpest testers available and the ones with the most to lose.

My contribution was upstream of the tests themselves. I created the user flows, and those flows were what the test scripts were written from and signed off against. I part-owned the defect log — and it was the defect log, rather than any status report, that I took into the weekly review with the CMO.

We didn’t move traffic on a date. We moved it when conversion said we could.

A licence expiry is exactly the pressure that produces a big-bang launch. Instead the new site ran alongside Tridion and had to prove itself on live traffic.

The CMO set the allocation. We monitored conversion rate with the CMO and the data team, comparing the new site against the old one running in parallel on the same days, in the same markets, with the same demand. Only once the new site cleared a satisfactory threshold did allocation increase — 20% of traffic first, then 80%, then the remaining markets.

I didn’t own that decision and I want to be precise about it: the CMO decided, the data team measured, and I sat in the room where it was read. What I did own was making sure the defects that would have depressed that number were found and fixed before each step up.

Weekly, against risk, impact and effort — and I lost the argument about the map.

Every week I met the CMO and the CRO manager to triage what we’d found. We scored against three things: risk, impact and effort. Having a stated framework mattered more than the specific weightings, because it meant an issue could be deprioritised without it being a defeat for whoever raised it.

One I lost. The map carried regional polygons inherited from the old site. I argued they should go: they were business-defined sales regions rather than real geography, so they drew boundaries that didn’t correspond to anywhere a customer could point to. Presenting an inaccurate region seemed worse than presenting none. The polygons stayed, because they had always been there and deviating from the old site made stakeholders nervous. I still think that was the wrong call, and it’s a fair example of the caution the whole project was working against.

One I won. I argued for cutting the content above the fold so product cards came up sooner. Customers arriving on a search-led site don’t want to scroll past marketing copy to reach the thing they came for — the prices, and the sort and filter controls that let them act on them. That one landed, and it’s the same conviction as the search design itself: take everything out from between the customer and a price.

I configured the CMS myself, which is how I found out what the design system actually did.

Once design and development settled, I and one or two colleagues were given the CMS build: content modelling, component configuration, page building, and migrating existing content. It ran over several months. Profound Digital wrote and uploaded the SEO content onto the pages we created.

Configuring it taught me things designing it hadn’t. Which configuration options affected which markets. How shared components behaved against market-specific and static pages. Where the website’s structure and the development setup actually met. A component model looks sound in a design file; you find out whether it is when you have to instantiate it thirteen times — including three German-language markets and two Belgian ones that shared a language but not a configuration.

The thirteen markets were not my doing alone — it would be unfair to claim that. Overseas teams and a digital marketing agency set up the additional domains for launch. What I owned was the model they were configured from.

Training ten people is how we found out what the component library couldn’t do.

We ran weekly training sessions with around ten people across the UK, NL, DE and PL content teams. The point was to hand over a system. What happened instead was that the system changed.

Two things surfaced almost immediately. Pages needed changing en masse and the CMS didn’t support it — a gap invisible from the design side, obvious within minutes to someone facing two hundred pages. And the component library needed adjusting to present real content well, as opposed to the content we had imagined while designing it. Components kept being optimised as more content came online, which shaped the site and improved conversion as it went.

Training was supposed to be the last step. It turned out to be a research method.

What we rejected: turning Tridion off in one go.

A single cutover was the obvious answer to a licence deadline. It was cheaper, it was faster, and it needed no parallel running.

We rejected it because it offered no way back. A phased allocation cost weeks we didn’t obviously have, and bought a rollback path at every step — on a project where the fallback was a platform we were contractually obliged to leave.

The result

Thirteen markets live from one content model, with the rollout gated on conversion at every step rather than on the calendar.

The component library that came out the other side had been revised by the people using it, not just the people who designed it.

Reflection

The component library kept changing as real content arrived, which is a polite way of saying the model wasn’t right when we froze it — the content teams should have been in the room during content modelling rather than meeting it in training.

The bulk-editing gap we found in those sessions was a limitation we designed around rather than solved. And I’d fight the polygon argument again, better: I had the reasoning right and made it a taste argument when I should have made it a customer-comprehension one, with evidence.

Next The 3.0 preference test Five markets voted for the new design. The study still stopped two of them shipping · Eurocamp · 2026