Point-of-view memo by Jaseem Quraishi

Point-of-view memo

Beyond Unified Design Systems: Architecting for Operational Efficiency in Multi-User Portals

For Engineering leaders overseeing frontend architecture and strategy in product-led SaaS companies. · 20 Oct 2026

The position

For multi-user operational portals, optimizing for distinct user workflows often requires targeted UI/UX divergence from a common design system, rather than strict adherence to a single, unified component library.

Why it matters

Engineering leaders frequently prioritize maximizing component reusability and design system uniformity, which are valid goals for scaling development. However, for product-led SaaS companies relying on highly efficient operational portals, a rigid adherence to uniformity can inadvertently introduce friction into specialized user workflows. This friction translates directly into reduced user productivity, increased training costs, and the need for costly workarounds, ultimately impacting the operational effectiveness and strategic value of the software.

The argument

While the pursuit of a unified design system and maximal component reusability is often a sound architectural goal, my experience suggests this approach can become a liability when building complex multi-user operational portals. Different user groups—like managers, brokers, staff, or salespeople—often have fundamentally distinct workflows and information consumption patterns. A manager might need aggregate oversight and approval flows, while a broker requires granular client-specific data and rapid action UIs. Attempting to force these divergent needs into a strictly uniform UI can lead to compromised efficiency for critical operational tasks.

In my work developing a comprehensive suite of operational portals, including a CRM, client, auditor, manager, and QA portals, I directly encountered the challenge of serving these varied user needs. Building these systems with technologies like React and Next.js, I observed how efforts to reuse components broadly across all portals, while initially appealing, often necessitated complex workarounds or led to suboptimal user experiences for specialized tasks. For instance, implementing real-time dashboard updates over WebSockets for analytics and notification counts frequently demanded tailored information presentation that a generic component couldn't easily accommodate without significant customization, which then eroded the reusability benefit.

My conclusion is that a more effective architectural strategy involves identifying where strategic UI/UX divergence is not just permissible, but essential. This means consciously designing for distinct workflow optimizations, even if it implies a slightly larger, but more targeted, set of components or variations. It's not about abandoning a design system entirely, but about architecting a system that can gracefully adapt to the genuine operational differences without compromising the core user's ability to perform their job efficiently. This approach requires careful consideration of scalability and maintainability from the outset, acknowledging that 'one size fits all' often translates to 'one size fits none perfectly' for highly specialized operational users.

This perspective extends to integrating complex data services into user-facing applications. For example, when implementing backend APIs for an AI report-analysis feature and presenting those processed insights within the client portal, the specific nature of the insights and the actions users need to take on them often dictated UI elements that prioritized clarity and immediate action over strict adherence to a general component pattern. This deliberate divergence, when managed architecturally, enhances user effectiveness and the overall value of the system.

What this is grounded in

I built the CRM frontend used daily by managers, brokers, staff and salespeople, plus client, auditor, manager and QA portals, the Shopper app and company sites using React and Next.js. I implemented real-time dashboard updates over WebSockets for report changes, analytics and notification counts. I also implemented 2 backend APIs for an AI report-analysis feature, where approved reports are saved to the database, batched weekly to the Claude API, and the analysis is stored and shown in the client portal.

What would change my mind

If a single, highly generalized component library could demonstrably serve the unique operational efficiency and information presentation requirements of all distinct user groups—managers, brokers, staff, and salespeople—across all my built portals without requiring significant workarounds or custom implementations for critical features.