Vinci had deep domain expertise in corporate pensions but no user-facing platform. The business challenge was clear: build a platform that lets HR specialists administer plans and brokers sell them. The user challenge was less obvious: these two audiences had different goals, workflows, and mental models.

HR specialists worked in batch mode. They processed hundreds of employees at a time: new enrollments during onboarding windows, quarterly invoice reconciliation, annual compliance reporting. Their work was cyclical and exception-driven. They needed to see everything, identify outliers, and resolve systematically.

Brokers worked in deal mode. Each sale was a unique configuration: understand the client's needs, model contribution scenarios, compare against competitors, close the deal. Their work was episodic and persuasive. They needed to move fast, iterate visually, and generate presentation-ready materials.

We ran competitor analysis to generate hypotheses, then tested them in interviews with 2 HR specialists and 5 brokers. The research confirmed the core insight: a single configurable product would please neither audience. The dual-platform decision emerged in week two, before any design work began.

The design work separated the two platforms while sharing foundations. Each major screen embodied a principle derived from research: HR needed batch operations and exception visibility; brokers needed deal speed and visual persuasion.

Testing happened in working sessions, not formal usability studies. We ran weekly show-and-tell sessions with the same reference users who had participated in interviews. Each session reviewed 3-4 screens; feedback was incorporated by the next week.

The HR table view was initially too dense. We had optimized for information density but sacrificed readability. Testers asked for adjustable column widths and row density controls. We added both; HR specialists customized their views immediately.

The broker proposal builder lacked a comparison reset. Users could create scenarios but struggled to return to baseline. We added a "reset to original" control that preserved the conversation flow.

The wizard step indicators were confusing. HR specialists did not understand which steps were complete versus in-progress. We moved from dots to explicit status labels: "Pending," "In Progress," "Complete."

By week eight, reference users stopped critiquing and started asking about rollout dates. That was the signal: the platforms had crossed from "external design" to "their tools."

Shipped two distinct platforms on a shared data foundation in three months. The dual-platform decision, validated through research, prevented a single-product compromise that would have pleased neither audience.

100+ research insights anchored every priority call. When debates arose about feature sequencing, we referenced the interview data. Research gave us ground to stand on.

The shared design system enabled the timeline. Component reuse across HR and broker platforms meant neither surface was an afterthought. Without shared foundations, three months would have shipped one platform well and one poorly.

User story quality improved after the rewrite. The product manager's second-draft stories — written after internalizing user motivations — drove clearer priority decisions than the technical specifications she started with.

Stakeholder engagement remained healthy despite the research protocol. The mitigation strategy — checkpoints, intervention protocols, cross-checks — preserved both data quality and the client relationship.

What worked: The dual-platform decision actually simplified stakeholder conversations. When we presented research showing the two audiences had non-overlapping primary tasks, the "build two" recommendation felt inevitable rather than expensive. The data made the case.

What was hardest: Managing stakeholder involvement in research without damaging the relationship or biasing the data. The protocol — explicit checkpoints, intervention triggers, cross-checks — was more work than expected but worth it.

What I'd do differently: Get the product manager into user research sessions from day one. By the time she rewrote stories the second time, she understood the audiences. Bringing her into interviews earlier would have saved a rewrite and caught priority calls we made twice.

What surprised me: The shared design system was not just efficient — it was essential. Three months for two platforms was only possible because components could be reused. I had expected efficiency gains; I had not expected the project to be impossible without them.