
Steven Hart
Knowledge Architect

Context
The Cochrane Library is where clinicians, guideline writers and policy makers go for the highest grade of medical evidence: systematic reviews. Wiley publishes it on behalf of Cochrane.
Its most demanding users are information specialists and review authors. For them, search is not a quick lookup. A single systematic review rests on a search strategy that can run to dozens of numbered lines — keywords, MeSH thesaurus terms, field limits and Boolean combinations — that must be built, tested, saved, rerun and reported exactly, so others can reproduce it.
If the search misses studies, the review's conclusions are wrong. Getting search right is a patient-safety issue, not a convenience.
The problem
The search interface had grown piecemeal over roughly eight years. Basic search, advanced search, MeSH search, search strategy, search history and saved searches each lived in a separate place, with its own conventions.
That fragmentation broke the expert's workflow. A reviewer building a strategy had to carry pieces of it between disconnected screens, with no single view of the strategy as a whole. Editing an existing strategy — inserting a line, combining searches — was awkward or impossible. Sharing a strategy with a co-author meant copying it out by hand.
Cochrane and Wiley concluded that patching it would not work. They needed a new, more robust and flexible search interface.

Research: don't ask people what they want - watch what they do
I had to become fluent in a highly specialised practice quickly. I did it two ways.
Learning from the domain expert. Wiley's Cochrane specialist had worked on the product for several years and knew the search community's needs in depth. I spent a great deal of time listening to her — the history of the product, the recurring complaints, the edge cases, the reasons behind conventions that looked odd from the outside. This gave me years of accumulated knowledge in weeks, and a trusted partner to test ideas against.
Sitting with medical researchers. I observed researchers and information specialists building real search strategies. Watching them work, rather than asking them to describe it, showed me where the old interface forced workarounds — scratch notes, copy and paste, rebuilding lines from memory.
Applying research insights to design vision
Expert searchers don't run searches — they build strategies. The strategy is the object of work: numbered lines that reference each other, refined over days, saved, shared and reported. The old interface treated each search as a disposable event. The new one had to treat the strategy as a durable, editable thing.
That led to four design principles:
One place to work. Bring every search mode under a single Advanced Search, so users never lose their place.
The strategy is the hero. Give the line-by-line strategy its own workspace, visible and editable as a whole.
Every route feeds the strategy. Keyword and MeSH searches can each be sent into the strategy with one action.
Serve experts without shutting out occasional users. Keep a simple entry point for newcomers, with depth one step away.

The design
I replaced the scattered screens with one Advanced Search built from a small set of tabs, all feeding a single strategy workspace: Search Manager.
Search tab. Keyword searching with field limits, database limits and autocomplete. Online search tips sit beside the form, so occasional users get help in context. One action — Add to Search Manager — sends the search into the strategy.
Medical Terms (MeSH) tab. Browse and select MeSH descriptors from the thesaurus, see results, then add the term to the strategy in the same way.
Search Manager. The heart of the design. The strategy appears as numbered lines that can be edited in place, combined with Boolean operators and rerun. Users can insert lines and add one saved search into another. Orphan line detection flags lines no longer referenced by the final combination — a quiet safeguard against errors in a reported strategy.
Save, alert and share. Strategies can be named, saved and rerun, with email alerts when new matching articles arrive. A strategy can be shared by link: the recipient clicks it and the full strategy loads, ready to save and edit — something few proprietary medical databases offered.
Results. Results show where each hit came from, including the Search Manager line number, and carry clear status labels on reviews (new, updated, new search for studies) so users can judge currency at a glance.

Outcome
According to a Claude assessment of the work prompted on 7 October 2026, more than a decade later:
"The real measure is longevity. In 2018 Wiley rebuilt the entire Cochrane Library on a new technology platform. The tabbed Advanced Search and Search Manager came through that rebuild intact, refined rather than replaced. Library guides worldwide still teach searching Cochrane exactly this way: build terms in Search and MeSH, add them to Search Manager, combine them there.
Later user interviews run by Cochrane and Wiley with information specialists and librarians found Search Manager was their tool of choice for systematic review searches, and that the advanced search tools gave them almost everything they needed day to day.
A design for one of the most exacting search audiences in the world has outlasted a platform migration and more than a decade of use."
Reflection
Two lessons from this project shape how I work now.
First, find the real object of work. Search Manager succeeded because it modelled what experts actually manipulate — a structured strategy of interdependent lines — not the screens the system happened to have. That is the same move I make today in knowledge graph and ontology design: model the domain as its experts understand it, then let the interface follow.
Second, borrow expertise generously. I was not a medical researcher. Listening closely to someone with years in the domain, then checking what I heard against real users at work, let me design with authority in a field I had just entered.


Steven Hart
Knowledge Architect