Published by
Joris Wijpkema
Published on
May 13, 2026
In Formula One, you invest heavily in engineering the car. Aerodynamics, power unit, fuel efficiency. Then you race it. But the track changes every Grand Prix. Weather shifts. Tire compounds behave differently. What was optimal in Barcelona is wrong in Monaco.
The best teams don’t just drive the car they built. They continuously re-engineer the setup between races, sometimes between sessions.
Supply chains are heading in the same direction. The structural decisions that define your network, where facilities sit, which suppliers you source from, how product flows across echelons, used to be settled once and executed against for years. That model assumed a relatively stable operating environment. The environment is less stable than ever.
Supply chain shocks now demand structural responses. In the first half of 2025, the effective tariff rate on US imports swung from roughly 5% to above 30% and back, driven by successive waves of policy on Chinese goods, steel, aluminum, reciprocal tariffs, and partial rollbacks. Crude oil futures dropped nearly 10% in days on geopolitical shifts. These are not demand fluctuations you can replan around. They change the economics of entire sourcing corridors. The right response is to re-evaluate the network itself: new sourcing relationships, nearshored production, restructured logistics. That requires optimization, not scheduling.
Compute has caught up with the math. Full-granularity optimization, millions of orders, hundreds of thousands of SKUs, hundreds of facilities, used to require heavy aggregation just to be feasible. Design models operated at a coarser level of detail than planning systems, which limited how actionable their outputs were. That constraint is gone. General Motors now runs a network model with 180,000 SKUs on Optilogic, a model that previously had to be chunked into pieces before it could run at all. When design and planning operate at the same level of granularity, the rationale for treating them as separate disciplines weakens considerably.
Cloud-native platforms connect the people who need to collaborate. Network design used to be the domain of a small specialist team working in desktop tools. Results were communicated through static presentations. By the time findings reached decision-makers, the analysis was often months old. Shared team environments and composable applications now put scenario results, KPI dashboards, and decision-support tools directly in the hands of business users across functions and regions. Optimization outputs inform operational decisions in real time rather than sitting in a queue for the next annual review.
APS systems have structural gaps in the convergence zone. APS platforms are good at operational planning: demand forecasting, consensus demand, basic capacity and inventory planning, order promising, MRP, replenishment, S&OP execution. For what they do, they do it well. But five capabilities that supply chain leaders increasingly need sit outside most APS architectures: scenario evaluation at scale, structural what-if analysis, true mathematical optimization (mixed-integer programming, not heuristics), routing optimization, and uncertainty simulation. Without these capabilities, supply chains are structurally suboptimized and leaving money on the table. These gaps exist because the underlying computational approach required is fundamentally different.
The distinction matters because it explains why APS vendors can’t simply add these capabilities. A CPU is general purpose: a small number of powerful cores executing instructions sequentially. It is the backbone of any computing system. A GPU is purpose-built for a specific class of problems: computationally intensive mathematics involving millions of variables solved in parallel. It does not replace the CPU. It handles problems the CPU was never designed to solve efficiently.
APS systems are the CPU of supply chain planning. Network optimization is the GPU. Both are needed. They are architecturally complementary. APS defines the best plan given your network. Optimization defines the best network given your strategy. Increasingly, supply chain leaders need both answers at the same time.
The overlap zone, S&OE and S&OP, is where this plays out. APS owns the core planning processes. But the scenario-heavy, optimization-intensive decisions that sit within that same horizon (supply-demand scenarios, profit/cost/service trade-offs, routing integration, inventory optimization across multiple objectives) require a purpose-built optimization engine. Optilogic’s system of supply chain models, engines, and solvers do exactly that. Connected by DataStar data pipelines that keep models current, the two systems form a complete planning architecture.
In practice, this complementary model applies across a range of high-stakes decisions: tariff and trade policy response, post-merger network rationalization, greenfield market entry, capital investment prioritization, and the shift from periodic network studies to continuous network adjustment. Each of these requires structural analysis that APS was never architected to deliver.
Extend what you already have. A network model built to optimize a single year can be restructured as a 12-to-18-month rolling model with inventory bridges between time periods. Mapped onto your S&OP cadence, it becomes a powerful ongoing planning tool rather than a periodic project.
Fix your most critical APS gap. One organization we work with managed volumes and service levels well in their APS but couldn’t incorporate full financial data into allocation decisions. They were making sourcing and distribution choices without end-to-end P&L visibility. Running Optilogic alongside their existing system closed that specific gap without replacing the platform running day-to-day operations.
The processing power exists. The integration patterns are proven. The math works.
What remains is the harder work: building the habit of asking structural questions on a planning cadence instead of reserving them for an annual review. Training teams to use the degrees of freedom that modern optimization gives them. Setting up data pipelines that keep models current enough to inform real decisions rather than validate past ones. The half-life of a network design decision is shrinking. The organizations that recognize this are the ones closing the gap between strategy and execution, and they will be best positioned when the next disruption arrives.
In Formula One, you invest heavily in engineering the car. Aerodynamics, power unit, fuel efficiency. Then you race it. But the track changes every Grand Prix. Weather shifts. Tire compounds behave differently. What was optimal in Barcelona is wrong in Monaco.
The best teams don’t just drive the car they built. They continuously re-engineer the setup between races, sometimes between sessions.
Supply chains are heading in the same direction. The structural decisions that define your network, where facilities sit, which suppliers you source from, how product flows across echelons, used to be settled once and executed against for years. That model assumed a relatively stable operating environment. The environment is less stable than ever.
Supply chain shocks now demand structural responses. In the first half of 2025, the effective tariff rate on US imports swung from roughly 5% to above 30% and back, driven by successive waves of policy on Chinese goods, steel, aluminum, reciprocal tariffs, and partial rollbacks. Crude oil futures dropped nearly 10% in days on geopolitical shifts. These are not demand fluctuations you can replan around. They change the economics of entire sourcing corridors. The right response is to re-evaluate the network itself: new sourcing relationships, nearshored production, restructured logistics. That requires optimization, not scheduling.
Compute has caught up with the math. Full-granularity optimization, millions of orders, hundreds of thousands of SKUs, hundreds of facilities, used to require heavy aggregation just to be feasible. Design models operated at a coarser level of detail than planning systems, which limited how actionable their outputs were. That constraint is gone. General Motors now runs a network model with 180,000 SKUs on Optilogic, a model that previously had to be chunked into pieces before it could run at all. When design and planning operate at the same level of granularity, the rationale for treating them as separate disciplines weakens considerably.
Cloud-native platforms connect the people who need to collaborate. Network design used to be the domain of a small specialist team working in desktop tools. Results were communicated through static presentations. By the time findings reached decision-makers, the analysis was often months old. Shared team environments and composable applications now put scenario results, KPI dashboards, and decision-support tools directly in the hands of business users across functions and regions. Optimization outputs inform operational decisions in real time rather than sitting in a queue for the next annual review.
APS systems have structural gaps in the convergence zone. APS platforms are good at operational planning: demand forecasting, consensus demand, basic capacity and inventory planning, order promising, MRP, replenishment, S&OP execution. For what they do, they do it well. But five capabilities that supply chain leaders increasingly need sit outside most APS architectures: scenario evaluation at scale, structural what-if analysis, true mathematical optimization (mixed-integer programming, not heuristics), routing optimization, and uncertainty simulation. Without these capabilities, supply chains are structurally suboptimized and leaving money on the table. These gaps exist because the underlying computational approach required is fundamentally different.
The distinction matters because it explains why APS vendors can’t simply add these capabilities. A CPU is general purpose: a small number of powerful cores executing instructions sequentially. It is the backbone of any computing system. A GPU is purpose-built for a specific class of problems: computationally intensive mathematics involving millions of variables solved in parallel. It does not replace the CPU. It handles problems the CPU was never designed to solve efficiently.
APS systems are the CPU of supply chain planning. Network optimization is the GPU. Both are needed. They are architecturally complementary. APS defines the best plan given your network. Optimization defines the best network given your strategy. Increasingly, supply chain leaders need both answers at the same time.
The overlap zone, S&OE and S&OP, is where this plays out. APS owns the core planning processes. But the scenario-heavy, optimization-intensive decisions that sit within that same horizon (supply-demand scenarios, profit/cost/service trade-offs, routing integration, inventory optimization across multiple objectives) require a purpose-built optimization engine. Optilogic’s system of supply chain models, engines, and solvers do exactly that. Connected by DataStar data pipelines that keep models current, the two systems form a complete planning architecture.
In practice, this complementary model applies across a range of high-stakes decisions: tariff and trade policy response, post-merger network rationalization, greenfield market entry, capital investment prioritization, and the shift from periodic network studies to continuous network adjustment. Each of these requires structural analysis that APS was never architected to deliver.
Extend what you already have. A network model built to optimize a single year can be restructured as a 12-to-18-month rolling model with inventory bridges between time periods. Mapped onto your S&OP cadence, it becomes a powerful ongoing planning tool rather than a periodic project.
Fix your most critical APS gap. One organization we work with managed volumes and service levels well in their APS but couldn’t incorporate full financial data into allocation decisions. They were making sourcing and distribution choices without end-to-end P&L visibility. Running Optilogic alongside their existing system closed that specific gap without replacing the platform running day-to-day operations.
The processing power exists. The integration patterns are proven. The math works.
What remains is the harder work: building the habit of asking structural questions on a planning cadence instead of reserving them for an annual review. Training teams to use the degrees of freedom that modern optimization gives them. Setting up data pipelines that keep models current enough to inform real decisions rather than validate past ones. The half-life of a network design decision is shrinking. The organizations that recognize this are the ones closing the gap between strategy and execution, and they will be best positioned when the next disruption arrives.
Fill out the form to unlock the full content
.jpg)
