Schuh, G., Prote, J. P., & Fränken, B. (2018). Designing Mulit-Site Operations Networks. 24th International Conference on Production Research. Aachen, Germany
Summary
In a world where manufacturing companies increasingly compete on a global stage, few rely on a single factory or domestic market anymore. Instead, they operate multi-site networks—systems of interlinked plants that serve different markets, produce different product families, or specialize in distinct stages of production. Managing these networks has become one of the most complex strategic challenges for industrial firms. The paper proposes a practical, research-based method to design such networks so they are both strategically aligned and cost-efficient.
Why Multi-Site Networks Matter
Multi-site manufacturing has expanded because firms want to reach new markets, access skilled labor, balance currency or regulatory risks, and tap into local cost advantages. But these multi-site operations introduce new challenges: coordinating dispersed plants, dealing with changing customer demands, and constantly re-evaluating where to produce what. When companies make site decisions in isolation—optimizing each factory rather than the system—they often create a fragmented, inefficient network. Conversely, when companies only chase low labor costs, they risk losing flexibility and market responsiveness.
The central question, therefore, is how to find the right balance between long-term strategic alignment (which sites should do what, based on market and product factors) and short-term efficiency (how to minimize total costs).
Gaps in Existing Approaches
Academic research offers two main families of methods:
- Strategic frameworks – qualitative tools that encourage reflection about where to locate, what to produce, and how to allocate roles across plants. These are easy to use but lack data precision.
- Quantitative optimization models – mathematical tools that simulate costs and performance across alternative network designs. They are powerful but often too rigid, data-hungry, and disconnected from management’s strategic reasoning.
Most companies struggle because these two approaches rarely connect. The article bridges that gap by integrating qualitative strategic input (managerial judgment, site competencies, market needs) into a quantitative optimization model that minimizes Total Landed Costs (TLC)—the combined production, logistics, and customs costs of the entire network.
Set the Data Model (“As-Is” Network)
First, managers define the scope (which products and sites are included) and the objective (e.g., cost reduction, regionalization, or growth). They then build a simplified data model of the current network—how products flow between plants and markets, what capacities exist, and what costs arise.
Key data include:
- Production programs per site and product group
- Process times, machine data, and resource use
- Fixed and variable costs (materials, labor, depreciation, energy)
- Transport and duty costs
- Local costs (rent, management, overhead)
Validation is essential: does the model mirror reality? The paper recommends a two-step calibration—first check capacity accuracy (machines, utilization, personnel), then verify cost realism (total personnel and product costs). Visualization tools—maps, waterfall charts, and treemaps—help managers spot inconsistencies and confirm that the digital twin of the network behaves like the real one.
Depict Future Scenarios
Once the baseline is validated, the model projects possible futures—typically three to ten years ahead. Scenarios can vary in:
- Sales growth per region
- Wage and productivity trends
- Introduction of new sites or product groups
This step helps management visualize how the network might evolve under different business conditions and test “what-if” assumptions before making costly decisions.
Achieve Strategy Fit (Integrate Qualitative Judgment)
Here lies the innovation of the paper. Before optimization, the firm inputs strategic boundary conditions—rules that limit which site can produce which product for which market. This prevents the algorithm from proposing unrealistic or undesirable allocations (e.g., producing highly customized medical devices in a low-skill, low-cost region).
The strategic fit is based on matching:
- Market and product requirements (from the customer side)
- Site abilities and competencies (from the internal capability side)
The authors adapt and extend Schmidt’s “portfolio analysis,” rating both products and sites on two dimensions:
| Market & Product Factors | Site & Capability Factors |
|---|---|
| Market complexity: degree of customization, local regulations, just-in-time needs | Market service: proximity to customers, delivery reliability, flexibility |
| Product maturity: stability of design and process | Product competency: site’s technical expertise and ramp-up ability |
Each factor is rated on a four-point scale (1 = very low, 4 = very high). Products with high market complexity or low maturity must be produced where market service and product competency are also high. This matching defines the solution space for optimization: only allocations that satisfy strategic logic are allowed.
Optimize the Target Footprint (Quantitative Optimization)
With the feasible solution space defined, the analysis begins iterative reconfigurations of the network. Like natural selection, it tries combinations of product allocations across sites, calculating the total landed cost of each configuration and keeping the most cost-efficient generations.
TLC includes all direct and indirect costs—labor, machines, buildings, energy, logistics, customs, and materials. The algorithm continues until it finds a configuration that minimizes total cost within the strategic boundaries.
The resulting “target footprint” shows:
- Which products each site should produce
- Which markets each site should serve
- The expected cost savings and logistics simplification
Managers can then analyze the resulting flows and iterate—testing trade-offs between a cost-optimized versus a market-oriented network to identify the best balance.
Industrial Application and Results
The approach was validated with a large metal-working company that had over ten plants and fifty product groups serving eight regions. The firm’s network had grown through mergers and acquisitions, creating overlapping production responsibilities and redundant logistics flows.
By applying the method, managers used expert input to narrow the solution space (excluding plants lacking specific capabilities) and used software to optimize within those limits. The resulting configuration consolidated production into three main sites, each serving distinct markets.
Even for a single product group, cost savings of 0.7% were achieved; across the entire network, projected savings reached 12.1% in total landed costs—along with reduced transport complexity and better market responsiveness.
Conclusions
The study delivers a balanced, pragmatic approach to multi-site operations network design. Rather than relying solely on intuition or on purely mathematical optimization, it integrates both: strategic managerial judgment (what makes sense) and quantitative rigor (what costs least).
Key contributions include:
- A method for combining qualitative strategy inputs with quantitative optimization.
- Improved managerial acceptance of modeling results, because decisions reflect their own input.
- Faster and more realistic optimization by narrowing the solution space to feasible scenarios.
- Demonstrated savings and better strategic alignment in a real industrial case.
Overall, the paper shows that firms can make smarter multi-site footprint decisions when they treat strategy and cost modeling as complementary, not competing, tools.
10 Practical Insights for Business Owners and Managers
- See your operations as a network, not isolated sites.
Each plant’s role should strengthen the entire system’s competitiveness—avoid “every site for itself” mentalities. - Define your objective clearly before analyzing.
Decide whether the goal is cost reduction, market proximity, resilience, or growth; this determines how to model and interpret data. - Build a realistic digital twin of your current operations.
A simplified but validated model of your plants, products, and logistics is essential before testing scenarios. - Integrate strategy with data—don’t let algorithms drive decisions alone.
Optimization tools can find cost minima, but only management can set strategic boundaries that make sense in practice. - Match products and markets to plant capabilities.
Complex or customized products should stay close to customers; stable, mature ones can move to lower-cost or specialized plants. - Use expert knowledge to guide model constraints.
Involve production, sales, and product managers in rating site competencies and market demands—it improves both quality and buy-in. - Test multiple future scenarios.
Don’t assume today’s conditions will hold. Model what happens if wages rise, demand shifts, or new markets open. - Balance efficiency with flexibility.
The cheapest network isn’t always the best; resilience, lead times, and service levels often justify local production. - Use visualization to communicate decisions.
Maps, flow diagrams, and cost waterfalls help non-experts grasp where money and time are spent—and where savings come from. - Iterate continuously.
Network design isn’t a one-time project. Revisit assumptions regularly as markets, technologies, and costs evolve.
Closing Thought
This study turns a complex academic problem—multi-site manufacturing network design—into a practical management tool. It empowers decision-makers to combine their strategic judgment with data-driven optimization, producing smarter, leaner, and more resilient global operations.