Operating Models

Shared Services, CoEs, and Federated Models: Choosing the Right Structure

When to centralize, federate, or distribute capabilities — decision criteria, cost-speed trade-offs, and real patterns

8 min read

Here's the uncomfortable truth: most organizations choose their operating model structure based on politics, not performance. The CFO pushes for shared services to cut costs. Business units demand autonomy and fragment capabilities. IT creates centers of excellence that nobody uses. Meanwhile, the actual work — delivering value to customers — suffers in the crossfire. The result? Organizations oscillate between centralization and decentralization every few years, burning millions in transition costs while never quite getting the structure right. This isn't a failure of strategy; it's a failure to understand how capabilities, value streams, and operating models actually interact in practice.

With hybrid work reshaping how organizations operate and economic pressure demanding both cost efficiency and speed to market, the old 'pendulum swing' between centralized and distributed models isn't sustainable. Business architects now have the frameworks and tools to make these decisions based on capability analysis rather than organizational politics — but only if we understand the true decision criteria.

Key Takeaways

  • Map each L1 and L2 capability to its natural governance pattern — some capabilities demand centralization while others require distributed execution
  • Use the frequency-variation matrix to determine structure: high frequency, low variation capabilities are shared service candidates; low frequency, high variation capabilities should remain federated
  • Calculate the 'collaboration tax' for each operating model choice — include coordination costs, decision delays, and knowledge transfer overhead in your ROI analysis
  • Build federated models around value streams, not organizational silos — structure follows the flow of value creation, not reporting relationships
  • Establish clear decision rights and escalation paths for each capability before implementing any structural change — ambiguous accountability kills shared services and CoEs alike

The Capability-Structure Mapping Framework

Not all capabilities are created equal when it comes to organizational structure — some naturally centralize while others resist it.

The Business Architecture Guild's BIZBOK provides clear guidance here, though most organizations ignore it. Foundational capabilities like Enterprise Risk Management or Regulatory Compliance naturally centralize because they require consistency and specialized expertise. Customer-facing capabilities like Sales or Customer Service resist centralization because they need local market adaptation. The mistake is trying to force a universal structure across all capabilities. Start by mapping your L1 and L2 capabilities against three dimensions: required consistency level, specialization depth, and local adaptation needs. Capabilities that score high on consistency and specialization but low on local adaptation are natural shared service candidates. Those with high local adaptation needs belong in federated models. The middle ground — moderate scores across all dimensions — is where centers of excellence add value by providing standards and expertise without controlling execution.

The Economics of Shared Services: Beyond Labor Arbitrage

Most shared services business cases focus on FTE reduction, but the real value comes from capability standardization and process optimization.

Traditional shared services models chase labor cost reduction — move work to lower-cost locations or consolidate redundant roles. This approach fails because it treats shared services as a cost center rather than a capability enhancer. The more sophisticated approach analyzes the total cost of capability delivery across the organization. Include technology licensing costs, training expenses, compliance overhead, and — critically — the hidden costs of inconsistency. When each business unit runs its own Finance function differently, you multiply audit costs, increase error rates, and create integration nightmares for M&A. Calculate what we call the 'consistency dividend' — the value created when a capability operates the same way across the enterprise. For many organizations, this consistency value exceeds the labor savings by 2:1 or more.

  • Map all direct and indirect costs of current distributed capability delivery
  • Quantify inconsistency costs: audit multipliers, integration complexity, training duplication
  • Model the learning curve effect — shared services typically take 18-24 months to achieve projected efficiency
  • Account for transition costs including change management, technology integration, and temporary dual-running

Centers of Excellence: The Expertise Distribution Model

CoEs work when they're designed as capability accelerators, not control centers — the key is balancing standards with autonomy.

The most successful centers of excellence operate as internal consulting organizations rather than corporate mandates. They develop standards, provide training, and offer advisory services while leaving execution accountability with the business units. This federated expertise model works particularly well for capabilities that require specialized knowledge but local execution — think Data Analytics, Digital Marketing, or Customer Experience Design. The TOGAF framework's concept of 'federated governance' applies perfectly here. The CoE establishes architecture principles, reference models, and approved technology stacks. Business units choose how to implement within those guardrails. The CoE measures adoption rates and business outcomes, not compliance percentages. When a business unit wants to deviate from standards, they propose an alternative and justify the business case. This creates innovation without chaos.