Centralized vs Decentralized UX Research Teams

Centralized UX works like an agency, loaning researchers to products and returning them to a resource pool; decentralized embeds UX in product teams. This matters when UX is added after engineering. The footgun is freezing structure rather than evolving it.
WHY IT EXISTS: UX is often introduced after development and product management are already established, so it must fit into existing power structures rather than define them. Organizations also evolve, adding products, features, and people, which constantly reshapes how teams communicate. A formal structure is needed to decide who owns UX budgets, careers, and priorities without isolating researchers from the product work that gives them context.
THE MENTAL MODEL: Think of centralized UX as an internal agency or center of excellence. Researchers report to a UX manager, build shared methods, and get loaned to product teams like consultants before returning to a common resource pool. Decentralized UX is the opposite: researchers join product teams permanently, report to product leadership, and become domain specialists with deep continuity. Neither is a personality type; it is an allocation of reporting lines and time.
HOW IT WORKS: In a centralized model, a UX manager receives requests from product leads, matches projects to available researchers based on skillsets and availability, and reclaims them when the work ends. This creates a single hierarchy that can champion UX at the executive level and guard budgets. In a decentralized model, UX professionals report into product teams rather than a UX manager. A matrix or hybrid model attempts to blend the two, giving researchers a home in UX while they embed with products.
WHEN TO USE IT: Use centralized teams when you need consistent methods across many products, when UX maturity is low and requires a single advocate to defend budgets and headcount, or when project demands fluctuate and you want to float talent to the highest priority work. Use decentralized teams when product domains are complex and require deep accumulated context, or when speed depends on researchers being present in daily product decisions without waiting for assignment.
WHEN NOT TO USE IT: Do not use a centralized model if it turns UX into a ticket-taking service that drops into projects too late and leaves before implementation. Do not use a decentralized model if every product team invents its own research standards and no one maintains a coherent career ladder or toolkit. Do not treat either choice as permanent; the model that worked for three products often breaks at ten.
ONE CANONICAL EXAMPLE: A growing company starts with five centralized researchers who rotate between two product lines. As the company adds a third line with specialized needs, two researchers embed there full time and report to the product director. The remaining three stay centralized, forming a hybrid. Over time, the embedded researchers keep their product reporting but gain a dotted line to the UX director for craft standards and performance reviews.
Source: nngroup.com
Read the original → nngroup.com
Get five bites like this every day.
Tezvyn delivers a daily feed of 60-second tech bites with quizzes to lock in what you learn.