Every manufacturing operation falls into one of two categories: discrete or process. The distinction shapes how a plant is laid out, how quality is measured, what its MES and ERP need to do, what regulations apply, and where computer vision fits on the line.
The short version: discrete manufacturing builds countable units; process manufacturing transforms materials in bulk. A pickup truck rolling off the line is discrete. A tank of latex paint is process. Each requires a different operating model.
Discrete Manufacturing
Discrete manufacturing produces individual, countable items assembled from a list of parts. Each unit can be picked up, serialized, inspected, shipped, and taken apart again. The defining traits:
- Output is a unit. Cars, smartphones, appliances, aircraft parts, medical devices, furniture, engines, windows, rail cars, packaged finished goods.
- A Bill of Materials (BOM) drives production. The product is specified as a hierarchy of parts and sub-assemblies. The job of the plant is to bring those parts together correctly.
- The work is mostly assembly, machining, forming, and inspection. Conveyors, robots, fixtures, presses, CNC machines, welders, screwdrivers, vision systems.
- Quality is dimensional, cosmetic, and functional. Is the part within tolerance? Is the surface clean? Does it work?
- Tracking happens by serial number or part number. Each unit is identifiable.
- Rework is possible. A defective unit can often be disassembled, repaired, and put back in flow.
- Throughput is measured in units per hour.
Examples of companies running primarily discrete operations: Pella (windows and doors), Rivian (vehicles), and Patrick Industries (manufactured housing and RV components). The plant floor is a sequence of stations, each adding or transforming a piece, with WIP moving through in identifiable batches or lots.
Process Manufacturing
Process manufacturing produces goods through chemical, physical, or biological transformations of raw materials. Output is bulk - measured in tons, gallons, liters, or kilograms - not in countable units. The defining traits:
- Output is a continuous or batch quantity. Chemicals, paints, coatings, fuels, pharmaceuticals, food ingredients, beverages, plastics, paper, cement, processed steel, gases.
- A recipe or formula drives production. Inputs are combined in proportions, often with temperature, pressure, time, and reaction conditions specified.
- The work happens inside equipment, not on the floor. Reactors, mixers, distillation columns, fermenters, ovens, kilns, extruders, pipes and vessels.
- Quality is chemical, physical, and microbiological. Composition, purity, concentration, viscosity, pH, particle size, microbial count.
- Tracking happens by batch or lot number. Once material is in the tank, individual identity is lost; the batch is the unit of traceability.
- Rework is limited or impossible. You can't unmix a batch. Off-spec material is reblended, downgraded, or scrapped.
- Throughput is measured in volume or mass per unit time.
Examples of operations that are primarily process: USG (building materials, gypsum board), Chobani (dairy), and U.S. Steel at the melt and casting stages. The plant floor is mostly piping, vessels, and control rooms with operators monitoring SCADA screens and historians rather than walking the line.
The Hybrid Reality: Mixed-Mode Manufacturing
Almost no real operation is purely one or the other.
The most common pattern is process upstream, discrete downstream: a continuous chemical or food process produces bulk material, which is then portioned into discrete packages, bottles, cartons, or pills. A yogurt plant is process up to the cup filler and discrete afterward. A pharmaceutical plant runs a batch process to produce a liquid, then fills, caps, labels, and packs discrete vials.
Other operations are batch-discrete: produce in batches that retain a batch identity but are also countable as units (semiconductors, specialty chemicals in drums, baked goods on a rack).
What matters isn't picking the textbook label. It's identifying which mode dominates at each section of the plant, because that determines what equipment, software, and metrics make sense there. A common practical answer is to run separate MES configurations or modules for the process side and the packaging side of the same site.
Why the Distinction Matters
The discrete vs. process line is not an academic taxonomy. It changes what's hard.
Changeovers. Discrete plants change over by swapping tooling and updating an MES routing - minutes to hours. Process plants change over by emptying lines, cleaning vessels (Clean-in-Place / CIP), validating the clean, and warming back up - often hours, sometimes a full shift, especially in food and pharma where contamination across products is a regulatory issue.
Quality data. Discrete quality is mostly observable on each unit - measure it, inspect it, log a pass/fail. Process quality is mostly inferred from samples - pulled from the line, sent to a lab, returned hours later, applied retroactively to whatever was produced in that window.
Regulatory exposure. Discrete is governed by general quality systems (ISO 9001) and industry-specific ones (IATF 16949 in automotive, AS9100 in aerospace, ISO 13485 in medical devices). Process is governed by composition-driven and safety-driven regimes - FDA 21 CFR Part 11 for pharma electronic records, HACCP for food safety, EPA reporting for emissions, OSHA Process Safety Management for highly hazardous chemicals. The regulatory weight in process is usually heavier.
MES architecture. A discrete MES is built around routing, dispatch, and unit genealogy. A process MES (often called a Manufacturing Operations Management system or a batch execution system, following the ISA-88 standard) is built around recipes, batch phases, and electronic batch records. Trying to run one on a platform built for the other ends in expensive workarounds.
What yield means. In discrete, yield is units accepted divided by units started - usually a high number, with the loss visible as scrap or rework parts. In process, yield is theoretical recipe output minus losses to evaporation, transfer, off-spec, and reblend - and the baseline yield is built into the recipe.
Improving Manufacturing with Computer Vision
Computer vision plays a real role in both modes. But it does different work in each, and the value is concentrated differently.
In discrete manufacturing, vision AI is a natural fit across the line. Most quality questions are inherently visual: is the part dimensionally correct?, is the surface clean?, is the assembly complete?, is the label right?, is the operator following the work instruction? Vision AI directly answers those questions on every unit at production speed. Common deployments include:
- Surface and dimensional inspection on each unit
- Assembly verification - every component present, correctly oriented, before the next station
- Print, label, and barcode verification
- Robotic guidance for pick-and-place and bin-picking
- PPE and safety zone monitoring
- Asset and WIP tracking on the floor or in the yard
As just one example, this is where BNSF uses Roboflow to automate yard inventory across rail facilities, and where dozens of other operators across automotive, aerospace, rail, and building products run vision AI in production.
In process manufacturing, vision also plays a real role, concentrated at the boundaries. Here are some of the many areas that matter:
- The discrete tail end - packaging, filling, labeling, cartoning, palletizing. Once bulk material becomes individual packages, all the discrete vision use cases (fill level, label, seal, count, completeness, code reading) apply.
- Process boundary observation - tank and silo levels, foam height, flame and burner monitoring, leak detection, conveyor jam detection, color and clarity checks at sight glasses, particulate detection.
- Safety, compliance, and operations support - PPE checks at zone entries, restricted-area monitoring, lockout/tagout verification, manual logbook digitization, valve and gauge reading.
For a hybrid plant, the vision AI footprint typically grows along the discrete portion of the line first, then extends back into the process side for safety and boundary monitoring.
The Next Step
If you're scoping operations technology: MES, quality systems, vision AI, the data backbone underneath - start by mapping which sections of your plant are discrete and which are process. That map tells you:
- Which sections need recipe and batch logic, and which need BOM and routing logic
- Where the regulatory weight sits (and which audit trails are non-negotiable)
- Which quality questions are observable on a unit (vision-friendly) and which require sampling and lab analysis (sensor- and analyzer-friendly)
- Where vision AI will pay back quickly (the discrete portions and the boundaries), and where it plays a supporting role (deep inside the process equipment)
Roboflow works with operators on both sides of the discrete / process line. Over half the Fortune 100 builds with Roboflow, and the platform runs more than 55 billion model inferences a year across critical industries.
If you're scoping where vision AI fits in your plant, see our customer stories for how teams on both sides of the discrete/process line ship it in production, or start a project on the same platform.
Frequently Asked Questions
1. Can a discrete product be a component in a process formula?
Generally, no. Process formulas use raw ingredients like chemicals or liquids. However, discrete items (such as bottles) are often the final destination for products made via process manufacturing (such as soda).
2. What is the main difference between a BOM and a Recipe?
A Bill of Materials (BOM) is a list of distinct parts used for assembly. A Recipe or Formula is a set of ingredients and process conditions (temperature, pressure) used to transform raw materials into a new substance.
3. Is Batch Production discrete or process?
It can be both. In discrete manufacturing, it means making a group of items (e.g., 500 green chairs). In process manufacturing, it means creating a specific quantity of a formula (e.g., a 500-gallon vat of shampoo).
4. Why is process manufacturing more difficult to track?
Because ingredients are blended in bulk, you cannot easily distinguish one specific drop of liquid from another. Process manufacturers rely on batch or lot tracking to maintain traceability for safety and compliance.
Cite this Post
Use the following entry to cite this post in your research:
Contributing Writer. (Mar 11, 2026). Understanding Discrete vs. Process Manufacturing. Roboflow Blog: https://blog.roboflow.com/discrete-vs-process-manufacturing/