Custom Optimization for Process Manufacturing

Introduction

Process manufacturing runs on physics. Viscosity, reaction yields, tank aging, and sequence-dependent changeovers are not parameters you set once and forget. They are constraints that shift run to run, week to week, and they decide whether the plan you publish on Monday survives contact with the plant on Tuesday.

A custom optimization model is built around those constraints. Tank routing, recipe yield variability, by-products, shelf life, and directional changeover penalties become equations the solver actually uses, not footnotes a planner manages on the side. That fit is what makes the plan executable.

The reason this matters is that the default option for mid-market process manufacturers is configurable Advanced Planning and Scheduling (APS): a six-figure license plus a multi-year implementation, sold on the promise that configuration will absorb whatever your plant looks like. In discrete manufacturing, that model works imperfectly but it works. In a chemical, coatings, or dairy plant, it leaves you with clean S&OP numbers and a shadow Excel system that does the real planning.

The Four Primitives Process Plants Violate

Most configurable APS systems are built on four assumptions. Process manufacturing violates all four at once.

Item identity: one tank stream splits into multiple grades; there is no single "parent SKU" that captures the physics.
BOM: recipes are formulations with variable yield, not a static parent-child tree.
Routing: setup is a directional penalty matrix, not a fixed "4-hour changeover."
Work center: a holding tank is a countdown clock and a mass-balance constraint, not an on/off machine slot.

The template is structurally wrong for how your plant runs. You are not missing a checkbox in configuration.

What "Configuration" Actually Produces

Vendors paper over these gaps with custom fields, user-defined attributes, and consultant-led mapping workshops. In practice, the result looks the same in every deployment we review. Changeover matrices get flattened into averages. Then the planner re-sequences the week in Excel. By-products get modeled as negative consumption on a phantom item. Then someone reconciles them in Excel when the secondary tank fills. Tank constraints become manual work-center block-outs owned by whoever knows which header is tied up. Shelf life becomes a flag on a report after the batch has already aged out in WIP. You paid for an end-to-end platform. What you operate is a macro plan with executive-friendly numbers and a parallel spreadsheet layer that encodes the physics the optimizer never sees.

When the template cannot model your physics, your planners model it in Excel instead of the system you bought.

Custom Modeling Once, Inside the Math

A custom optimization model inverts where the hard work happens. Tank routing, sequence-dependent changeovers, recipe yield variability, by-products, and shelf life become equations solved together with demand, inventory, and capacity. They are not offline rules maintained by planners every Monday morning. When you run production through an end-to-end supply chain optimization engine like WonForge, the solver sees the coupled system: reactors, tank farm, packaging lines, and shipment dates in one pass. It can take a strategic 3-hour changeover on Wednesday because it knows that choice prevents a 12-hour catastrophe on Friday. It also knows which holding tank will still be occupied when that transition lands.

Custom optimization moves the modeling work from recurring spreadsheets into the model once, where the solver can actually use it.

The Real Cost Is the Plan, Not the License

The expensive part of a configurable APS plus Excel deployment is not the annual maintenance invoice. It is the plan itself. When the optimization engine cannot see sequence-dependent changeovers, by-product tanks, tank-farm blocking, or shelf-life decay, it does not optimize around them. It optimizes a simplified problem. Your team patches the output by hand. Capacity bleeds out in washouts nobody priced into the macro schedule. Safety stock balloons. The plan never executes anyway, so working capital sits there compensating for a fiction. Margin leaks on rush orders that looked feasible in a grid but triggered a Friday caustic boil-out and weekend overtime. The S&OP deck looks clean. The decisions that actually move money are made downstream in spreadsheets with no view of the full supply chain. The smaller software invoice from a custom model is real, but it is secondary. The primary saving is recovered every week the plan runs: fewer phantom changeovers, less schedule churn, and decisions priced on plant physics instead of open calendar slots.

A simplified optimizer does not save money. It hides where margin is leaking until freight, scrap, and overtime show up somewhere else.

Frequently Asked Questions

What is custom supply chain optimization?

A custom optimization model is a mathematical description of your specific plant, run on a commercial solver. The methods, solver, and cloud infrastructure are standard. What is custom is the specification: your tanks, your recipes, your changeover matrix, your shelf-life rules, encoded as equations and constraints the solver actually uses.

Why do process manufacturers plan in Excel if they already bought APS?

Because the APS optimizer optimizes a simplified version of the plant. Sequence-dependent changeovers, by-product tanks, tank blocking, variable recipe yields, and shelf-life clocks are either averaged away or managed as manual overrides. Planners use spreadsheets to encode the real rules the license cannot solve.

Is custom supply chain optimization cheaper than a configurable APS?

Yes, for most process plants, on license and total cost of ownership. The bigger savings are operational. When physics live inside the optimization model, the plan executes with less hand-patching, less schedule churn, and less margin lost to washouts, WIP aging, and rush-order surprises. License cost is only part of the equation; plan quality is the recurring return.

Conclusion

The question is where the custom modeling work lives. In a configurable deployment, it stays in spreadsheets, year after year, owned by planners who patch what the template cannot solve. In a custom optimization model, it happens once, inside the math, where the solver can trade off the full plant: reactors, tanks, packaging lines, inventory, and shipments together. Want to see whether custom optimization fits your plant? Click Check Your Fit and we will take it from there.

Custom Optimization is for Process Manufacturing