800~ Researchers. 20+ Websites. Dev team of 3. 1 CMS project.

Not your typical CMS project

At Eurac Research, complexity isn’t a challenge, it’s the starting point. The private research centre in Bolzano brings together around 800 researchers working across fields as diverse as earth observation, renewable energy, linguistics, biomedicine, and Alpine environments. Oh. And even mummies.

When Thomas Iacopino joined the web team, none of today's infrastructure existed yet. There was a SharePoint site and a communication department that was the single source for everything published. There was a plan already in motion with a Viennese company to migrate to and ship a monolithic CMS that would handle the whole thing.

For years, that complexity was mirrored by a familiar digital setup: a centralised system, a single publishing pipeline and a communications team acting as gatekeeper for everything that went online. It worked - until it didn’t.

When Thomas Iacopino joined the web team, it quickly became clear that the organisation had outgrown its infrastructure.

“The organisation is so complex, with so many stakeholders, that it was clear you could never have a plan for all of them at any given moment. We needed a lot more flexibility than a monolithic CMS could offer.”

What followed wasn’t just a technical migration. It was a fundamental rethink of how content is created, managed and distributed across a highly decentralised organisation. Eurac Research didn’t replace one rigid system with another. Instead, it chose a modular approach built on DatoCMS, Next.js and Vercel - starting small, with a magazine prototype, and scaling from there.

However, what made the difference at the end wasn’t just the stack, but ownership. Instead of routing every update through a central bottleneck, teams now manage their own content within a shared framework. The central web team has shifted from gatekeeper to enabler.

It’s less about having the perfect setup from day one, and more about creating something that can evolve with the organisation.

The transition definitely required new workflows, clear guidelines and trust, but starting with a magazine prototype made it manageable. From there, the system evolved step by step: adding content types, refining components, and integrating new sources. The modular setup now makes it easy to adapt without starting from scratch.

The stack

The core of the Eurac setup is a single Next.js codebase deployed to Vercel, with DatoCMS handling all content. But the architecture underneath is more layered than it sounds. Their entire project is a monorepo-multisite setup that really pushes the boundaries of "Content Hub" as a Headless CMS use case.

For their conference sites, for example, the team had been dealing with a parade of WordPress installations, each with its own templates and its own maintenance overhead. Their new solution was elegant with one shared Next.js codebase that points to separate Dato models depending on which conference it's serving. Update the codebase once, all conferences get the update. Each conference still gets its own Dato content for editorial isolation, but the frontend is unified.

From a technical point of view, conferences share the same structure. So we use one Next.js codebase deployed to Vercel, and we just point it to the right DatoCMS project. You change something once and everything stays in sync.

Martin Palma, Developer

For data syncing across their various external systems, they run three separate Next.js backend projects, each responsible for connecting to a different external system via simple HTTP APIs. GitHub Actions triggers daily syncs, keeping everything up to date without manual intervention.

The team has also built several custom plugins in-house including a Brevo newsletter integration, a Next.js preview plugin, and most recently, a text-to-speech plugin that uses OpenAI to generate an MP3 from a long article, uploads it to the DatoCMS media gallery, and attaches it to the record so readers can listen to articles instead of reading them.

With any project of this scale, caching is always a nightmare. Ok to be honest, on a project of ANY scale caching is always a nightmare. So to keep things running smooooth, cache tags were a genuine turning point for the team. Before, they were managing a mix of time-based revalidation windows across different content types, some set to five minutes, some to a day, maintaining a document that tracked what needed to be invalidated when something changed. It was fragile, confusing for editors, and led to a constant inflow of "I published, where's my content", which is never fun.

Caching was always a pain. With a lot of related content, a person is linked to an article, a project is linked to the same article, you have a homepage slice showing the latest news. You have to keep the whole site structure in your head to know what to invalidate when. It never really worked well.

Thomas Iacopino, Web Team Lead

Cache tags removed all of that. The team now uses them everywhere. They no longer invent naming conventions, maintain revalidation tables, or explain to editors why their change isn't live yet.

The editorial layer

Here's where things get interesting. Eurac's web projects have about 100+ editors creating content in 3 languages, all powered by a small team of three developers. The team doesn't use granular permissions per editor, but instead, they have combined roles on a very high level to keep things smooth as far as content access and operations go.

The trust-based system at Eurac is not an accident, it's a deliberate choice. Editors are grouped by institute and given access to the models they actually need. They can read other models to create relationships, but they cannot delete content created by someone else. That single guardrail works for them without needing to create a super complex permission based structure on top of their SSO. This also means things are more "generalized" and easier to document, so onboarding takes one to two hours. After that, editors are free to publish.

What has genuinely helped editor confidence is how consistent the DatoCMS interface has stayed over the years. Martin called us boring and vanilla in the nicest way.

The interface is boring, and I mean that in the best way. Editors learn it once and then it just works. With systems like X, Y, and Z, every update reshuffled the UI. With DatoCMS, they always know where the preview button is, where the publish button is. That stability makes them trust it.

Martin Palma, Developer

Content creation itself is a mix. For the magazine, editors often draft in Word and paste into the structured content fields, which works well for long-form articles that then need image placement and translation. For more modular, block-heavy pages, direct input into the CMS is faster. The team handles most translations internally, and some editors still use Word's review and comment features for that workflow before the final version lands in Dato.

DatoCMS as a hub

This is the part that surprised us when we first heard about the Eurac setup. DatoCMS is not just where content lives. It is the connective layer between a bunch of external systems too.

Researchers at Eurac have profiles on ORCID, the widely-used researcher identification service. They manage their publications in Converis, an internal research database, rather than having access to the DatoCMS project. Employee records live in Microsoft Dynamics. None of these people log into the CMS, and yet their data is there enriched and structured to be connected to the rest of the content on the site.

The three Next.js backend projects mentioned earlier each connect to one of these external systems. Every day, GitHub Actions triggers a sync. Publications and project data are pulled from Converis. Employee profiles and team data are fetched from Microsoft Dynamics. That content lands in DatoCMS, where researchers can then log in and enrich it, adding context, stories, and detail that the raw database records don't contain.

Interestingly, the content ALSO flows the other way. When a researcher has a publication that needs a DOI, a permanent identifier used for citations and academic ranking systems, the team generates the DOI and builds a landing page for it inside DatoCMS. And researcher publication data gets pushed back out to ORCID on their behalf, so researchers never need to manually update their profiles.

DatoCMS became an event-driven hub. It's not just a place where we store content. It fetches from closed systems, enriches the data, and connects all these little islands together. At some point the ICT department came to us asking for an API to get enriched employee profiles back out of it. That told us everything.

Thomas Iacopino, Web Team Lead

For search, the team uses Meilisearch. Webhooks push new and updated content from DatoCMS to Meilisearch on publish.

For newsletters, the team built a custom Brevo integration via a DatoCMS plugin. Editors compose newsletters entirely within DatoCMS, picking from content that already exists on the site. The backend generates the HTML template and hands it to Brevo for delivery. Editors never touch Brevo directly. Statistics are fetched back into the interface so they never need to leave a single system.

The write-once-publish-everywhere model means institutes don't need separate logins to a newsletter tool, don't need enterprise plans, and don't need to copy and paste content between systems. It all lives in the same interface they already know, and gets channeled where it needs to be, because that's the "omnichannel" approach that actually makes sense when working on a content project of this size.

Looking ahead

Thomas and co. are currently building out an AI-powered "discover our research" feature that will let visitors ask questions using natural language and get back relevant projects, publications, people, and blog posts. The structured data already in DatoCMS makes this significantly more plausible to work with than it would be otherwise.

The text-to-speech plugin we mentioned earlier is also close to being live on the frontend. Long articles will get an audio version generated automatically, uploaded to the media gallery, and surfaced to readers as a listening option.

They also have a broader vision that has guided the whole project from the start - to have no dead ends. Every page should lead somewhere. A blog post leads to the author. The author leads to their project. The project leads to a service or a related story. DatoCMS being a relational, structured content system is what makes that kind of content graph possible at scale, across three languages, managed by a team of three.