CLIENT

Servier 

CONTRIBUTION

Competitor Audits
Interviews
UX / UI Design
Brand

Designing an Optimized Search

The DSMM (Direction Stratégique Médicin Mondiale) at Servier, present in over 150 countries worldwide, requested a secure Cloud-based platform to consolidate and share their strategic documents and medical reports worldwide. 

The challenge we determined would be finding the right strategic documents and reports in a vast digital repository, as users often struggle with unclear filters, irrelevant results, and inefficient navigation.

Our goal was to redesign the search experience by implementing intuitive filters and facets, improving relevance, speed, and usability.

Identifying key users

As this was an internally sponsored project, we kicked off with a briefing and several interviews with key stakeholders to listen and learn about their requirements. They would also be the end users, but as this was not just reserved for the C-suite, I hunted down other potential users to understand their expectations of such a platform. An interesting highlight was the ability to save searches and receive notifications on new document additions and links to external industry subscriptions. 

Identifying a red route flow for a MVP

Working closely with technical and business stakeholders, a red route was defined for the MVP; this is the core user flow. Alongside, I mapped out the ideal user journeys to be validated, from document upload and categorization to sharing with specific roles or access levels. Ensuring the experience remained intuitive and focused on key outcomes such as access control and audit trail visibility. This collaboration helped refine flows to match infrastructure constraints, surfaced early technical blockers, and created a shared foundation for an MVP centered on real user needs.

A simplified, non-exhaustive recreation of the project’s red route.

Wireflow

Early wireflow mapping of the entry sequence showing login, error handling, and the landing screen with its core navigation (bookmarks, subscriptions, saved documents, profile, notes, inbox). This initial feature inventory, worked through with stakeholders, became the starting point for the more refined information architecture. 

Content Structure

With a red route defined and the MVP key users established, we then worked to sort content around a user's priority tasks. The platform was structured around a metadata-driven document model, enabling consistent classification, discovery, and management of the documents across the platform, whether they were marketing, medical, or market access. The metadata model wasn't just for classifying documents; it also supported the document lifecycle and approval workflow.

To structure the IA, I ran a workshop to identify and sort content types and establish the document categories into buckets on a shared board, surfacing how different user groups (medical, marketing, compliance) mentally grouped the same content differently. This exercise directly shaped the streamlined structure below, prioritizing the red route while keeping room for the categories that mattered most to non-primary users. 

02 — Metadata model

Metadata is assigned at the document level to ensure consistent classification, discovery and access across the platform. The underlying content structure was defined and agreed upon as:

Cross-platform metadata model

01 — Document lifecycle & approval

Information Architecture

The MVP architecture was streamlined around content discovery and document management with a structured metadata model that included therapeutic area, region, country, access level, document type, and audience. The navigation was organised around a user’s core tasks. A global search provided access to key functions across the platform, while document-level actions such as download, save, and share supported everyday workflows. As we were short on time for necessary additional workshops, subscription management and advanced personalisation features such as customising the homepage were deprioritised for the MVP and considered for future iterations.

The Service Blueprint

A service blueprint was created in collaboration with the engineering team to clarify the various user access levels and their corresponding tasks or functions within the platform. In this case, the blueprint focused on the process of submitting a new document, outlining each step from the user's initial interaction to the system's backend responses. This included touchpoints such as authentication, form validation, file upload, and confirmation feedback, as well as the roles of different user types—such as admins, reviewers, and standard users—throughout the process. Mapping out these interactions helped identify potential friction points, ensure consistency across user flows, and align the technical implementation with the intended user experience.

The Service Blueprint - New Document Submission

The Existing Design Proposal

I was provided with an existing design as a point of departure, which I turned into a wireframe to focus discussions with stakeholders on functional features for a MVP (Minimal Viable Product). Helpful to understand the intended user flow, visual hierarchy, and brand considerations, while also identifying areas that required refinement or clarification.

Designing Search for a Mixed Audience

Four interaction models were on the table for the redesign — query syntax, split facet panel, expanding overlay, and filter bars. Each solves the same problem (get a mixed medical/marketing/compliance audience from blank state to filtered result) with different trade-offs in learnability, screen real estate, and scalability as the taxonomy grows.

I proposed the granular faceted panel as the primary pattern: it's legible without training for a non-technical audience, transparent for compliance reviewers who need to see what's being filtered and why, and extensible as new document categories get added — no redesign needed each time a taxonomy grows. The proposal was accepted and became the core interaction model for v1.0.

Same job: get someone from a blank state to a filtered result set as fast as possible, with four different interaction bets. Switch between them below; each includes a short read on who it favours and what it costs.

The constraint — documents already classified along known dimensions (type, therapeutic area, region...); the taxonomy is fixed today but will grow as the platform scales globally.
The question that constraint raises — which search interaction holds up as categories get added, without needing a redesign each time.
The exploration — this is where your four models go, each captioned with what it costs and who it favors (you've basically already written these captions in our earlier back-and-forth).
The decision — facet panel as primary (legible to a mixed medical/marketing audience, transparent for compliance reviewers), AI natural-language as a secondary layer (extensible without new UI, but never the only path so no one's forced to trust a parser).
The transition — the collapsing hero, because it shows you also thought about the moment-to-moment motion, not just the static states.

Query syntax

Best for frequent, expert searchers who repeat the same complex queries. The costs are: a learning curve for anyone unfamiliar with the operator syntax. Feels like Linear, GitHub search, a research terminal
 

Split facet panel

Best for exploratory browsing — narrowing by trial and error, seeing counts before committing
Costs: permanent horizontal space; less oversized/dramatic than the current hero
Feels like Bloomberg terminal, e-commerce faceted filters
 

Expanding Overlay

Best for keeping the header calm, then rewarding intent with a focused full search moment
Costs: An extra click/keystroke before typing; recent + saved searches need real data to earn their space
Feels likeSpotlight, Linear's Cmd+K, Notion search
 

Filter bars better showcase filters
By making the filters more prominent on the page and by making them available as the user scrolls a filter bar does a better job of showcasing filters. Presenting filters left to right also makes it very easy for users to quickly see what options are available.
5. Filter bars are more mobile and tablet friendly
 

Brand & Type Scale

+ 33 (0) 6 62 81 69 54
penstar@email.com  
Orgeval,  78630,  France.

©  |  2026