Case study — NFDI4Health · 2022
FAIR is easy to agree with and hard to interface. What does findable, accessible, interoperable and reusable mean on a screen a researcher uses under time pressure?
The redesigned start page. Visual clutter went out, and the search field took on the work: quick access to recent additions and COVID-19 studies, the searches this person ran last, and what others look for most.
Screens show anonymised catalogue content · click a screen to enlarge
A catalogue for German health research
The Health Study Hub catalogues clinical, public health and epidemiological studies: quality-assured metadata, study materials and harmonised data from many sources, in one place.
Researchers in these fields lose time to the same problem repeatedly: study information is fragmented across institutions, described in inconsistent standards, and often impossible to reuse without a phone call. FAIR principles exist to fix exactly that.
But a catalogue that nobody can search, and a submission form nobody finishes, does not make anything findable. So the project started with research rather than screens: what do these researchers actually do, and where does the current platform lose them?
Metadata does not become findable by being complete. It becomes findable when someone finishes entering it.
Two opposite audiences, one catalogue
Research first, then the redesign
I led the user research — interview guide, sessions, evaluation and prioritised recommendations — and then designed the redesign it argued for: start page, search and filtering, the study dashboard and the submission flow.
The research was the leverage. In a consortium of experts, a documented pain point ends a debate that a design opinion cannot. Every recommendation went back to the client with the evidence attached and a priority next to it.
From there the work was ordinary product design under an unusual constraint: the metadata schema was largely fixed, so nearly all improvement had to come from structure, sequence and wording rather than from cutting fields.
Three decisions, each from a documented pain point
01
I ran an interview series with researchers and platform users, transcribed and coded it, and turned the result into something a consortium can act on: problems grouped by theme, rated by priority, each with a recommended action.
That artefact did the political work of the project. It moved the conversation from “I would prefer” to “this is where people got stuck, and here is how often”, and it set the order in which we fixed things.
02
Studies were hard to find because the filter system was both confusing and overloaded — it asked casual visitors to think like the schema, and still did not give specialists the precision they wanted.
The answer was two modes in one interface: simplified filter queries carry the common case, and an expert mode exposes include, equals and exclude logic per resource type, with filter sets that can be saved, loaded and reset. Applied filters stay visible as chips, so a long result list never becomes unexplainable.
03
The single biggest pain point had nothing to do with search: a study had to be entered in one sitting, or the progress was gone. For a form of this depth that is a guarantee of abandonment.
Two changes fixed it. A dashboard gives every contributor their own studies with a status — draft, in review, published — so work in progress exists as a thing you can come back to. And the form itself became a stepper with grouped sections, mandatory fields marked as such and explanations where the schema demands a term nobody says out loud.
Form abandonment dropped by 80% after the redesign.
FAIR, made operable
Studyhub changed how clinical and public health studies are found, documented and reused: scattered metadata became a catalogue researchers can search two ways and contribute to without losing an afternoon of work.
Documented pain points settled discussions that competing expert preferences could not.
Serving specialists in a second layer let the default stay simple instead of averaging both audiences into one mediocre search.
Splitting one long form into resumable steps beat any attempt to shorten a metadata standard.
Every abandoned form is a study that never enters the catalogue at all.
The fastest way to make data findable was to stop losing the people entering it.
It is the clearest case I have of research paying for itself. The 80% did not come from a visual redesign — it came from an interview in which someone described losing an hour of work, and from taking that seriously enough to restructure the flow around it.
It also set how I work in expert domains: I do not need to become an epidemiologist, but I do need their vocabulary, their sequence and their failure cases before I am allowed to simplify anything.