Where Centralization Pays and Where It Stalls: A Shared-Services Test for Mid-Market Holdcos

For US mid-market holding companies, the appeal of shared services is easy to understand. A single finance team, one HR platform, consolidated procurement, and a common IT stack promise lower cost and tighter control. In 2026, with lenders and investors still focused on cash conversion and margin discipline, that promise is difficult to ignore. Yet the parent-company premium is not created by the act of centralizing. It is created when the center delivers a service that is faster, safer, or cheaper than the portfolio company could deliver on its own. When it does not, the holding company has simply moved work farther from the customer and added a layer of approval.

The Paper Synergy Trap

Most shared-services business cases look convincing on a slide. Duplicate roles are counted, vendor contracts are combined, and software licenses are rationalized. The projected savings are real in the model, but the model often misses the costs that appear after the transition: longer queue times, lost local judgment, duplicate shadow spending, and management attention consumed by governance. A portfolio company that once resolved a customer credit issue in an afternoon may now wait for a central collections team to review a ticket. A plant manager may lose the authority to approve a critical maintenance purchase. Those delays do not always show up as line-item costs, but they show up in revenue, retention, and morale.

The error is not centralization itself. The error is treating centralization as a default answer rather than a deliberate operating choice.

A Four-Question Test Before Centralization

Before moving any service to the parent, the holding company should apply a consistent test.

First, is the work standardized and low-variance? Transaction processing, payroll administration, tax filing, treasury operations, and cyber security monitoring tend to fit because they reward consistency. Pricing decisions, customer credit exceptions, and plant-level maintenance planning tend to resist standardization because local context matters.

Second, does local knowledge create revenue or protect risk? If a service depends on customer relationships, regional labor conditions, or site-specific regulations, moving it to the center can erode value. If it depends on technical expertise that is expensive to replicate, the center may be the better home.

Third, can the center meet a service level that the portfolio company would recognize as an improvement? The test is not whether the center is cheaper. The test is whether it is cheaper without degrading speed, accuracy, or responsiveness. A shared-services team that saves $400,000 but adds five days to customer onboarding may be a net loss.

Fourth, is the governance model clear? Shared services require a service catalog, an escalation path, and a chargeback or allocation method that portfolio leaders understand. Without those, the center becomes a budget black box and local teams build workarounds.

What Usually Belongs at the Center

In most US mid-market portfolios, the strongest candidates for centralization are functions where scale, control, and technical depth matter more than local intimacy. Treasury and cash management, consolidated reporting, tax compliance, insurance placement, cyber security, payroll platforms, benefits administration, and enterprise vendor contracts often fit. So does the management of common cloud and software agreements, provided the center does not become a purchasing police force.

Procurement deserves a more nuanced approach. Strategic categories such as freight, packaging, and professional services may benefit from central leverage. Site-level consumables and emergency purchases often need local discretion. A useful rule is to centralize the contract, not every purchase order.

What Should Stay Local

Customer-facing and site-specific work should generally remain close to the operating leader. That includes pricing exceptions, customer credit decisions within agreed limits, local sales compensation design, plant maintenance priorities, and community or regulatory relationships. These are areas where the cost of a delayed decision is often higher than the cost of a slightly less standardized process.

Human resources is similarly mixed. Benefits platforms and payroll can sit at the center. Recruiting for a regional plant, managing a local labor relationship, and handling site-level employee relations usually belong with the business. The center can set policy and provide tools without owning every conversation.

The Middle Ground: Shared Platforms, Local Configuration

Many mid-market holdcos get better results from a federated model than from full centralization. The parent provides the platform, the standards, and the subject-matter expertise; the portfolio company configures the platform to its market and retains decision rights within clear limits. A common ERP with local reporting, a shared HR information system with site-level administration, and a central cyber team with local incident response protocols can capture much of the scale benefit without removing operating judgment.

This model requires discipline. The center must publish service levels, costs, and performance. The portfolio company must follow the standards and stop building parallel systems. The parent must resist the urge to approve every exception.

Measuring the Premium, Not Just the Savings

Shared services should be measured on a small set of operating and financial indicators, baselined before the transition and reviewed quarterly. Cost per transaction, cycle time, error rate, and service-level attainment matter. So do operator retention, customer response time, and the percentage of local workarounds. On the financial side, the holding company should track the EBITDA bridge from centralization, including stranded costs, transition costs, and any revenue impact. A savings number that does not reach EBITDA or cash flow is not a premium; it is a projection.

The parent should also watch decision latency. If portfolio CEOs need more approvals after centralization than before, the center is adding bureaucracy. That signal is often more valuable than another benchmarking study.

A Sequencing Rule for 2026

The safest path is not a big-bang reorganization. Move one or two standardized services first, publish the service catalog, and set exit criteria. If the center cannot meet its service levels for two consecutive quarters, pause further centralization and fix the operating model. Add a portfolio-company operating leader to the shared-services council so that local friction is visible before it becomes attrition or lost revenue.

The parent-company premium is real, but it is earned service by service. For US mid-market holding companies, the winning model in 2026 is not the most centralized one. It is the one that can prove, with operating data, that the center is helping the portfolio companies move faster and earn more.