Dairy plant MES and batch traceability. Silo to pallet, in an Australian plant.

Why dairy is the hardest trace problem in F&B.

When was the last time your plant ran an unannounced mock recall and finished it before lunch?

The Australian dairy sector processed 8,315 million litres of milk in 2024-25, down 0.7% on the previous season due to drier southern conditions and flooding on the NSW Mid North Coast (Dairy Australia, In Focus 2025). USDA forecasts 8.8 million tonnes for 2025. Cheese alone takes about 35% of that milk pool. The processor map is concentrated: Lactalis Australia, Saputo Dairy Australia, and Bega Group hold the largest shares of the $5 billion milk and cream processing market.

Inside any of those plants, the traceability problem looks different from a packaged-foods plant. A biscuit line links a flour pallet to a finished case in something close to a straight line. A dairy plant blends thirty farm receipts into one silo, runs the silo across two filler lines and three SKUs, and ships the pallets to four customers. Forward trace from farm to pallet, and reverse trace from pallet back to farm, both have to handle the many-to-many topology. That is the problem MES exists to solve.

What silo-to-pallet actually requires.

The phrase gets used loosely. A useful definition: at every step where a raw or in-process material changes container, identity, or batch number, the link between what went in and what came out is captured and stored. Every step. Not every important step.

For a fluid dairy plant the steps are predictable. Raw milk receipt at the tanker bay: farm of origin, tanker volume, test results, receiving silo. Silo blending: which silo received which receipts, which day, in which order. Pasteurisation cycle: HTST cycle parameters, time-temperature integration, divert events, the silo or silos drawn from. Inter-process storage: any movement between silos. Filling line: which filler ran the SKU at which time, drawing from which silo. Packaging: case codes, pallet codes, the link from case to pallet. Despatch: which pallets left on which truck, against which customer order.

The traceability output is two queries that must run quickly:

• Forward trace. Given a raw milk receipt that turned out to have an antibiotic residue, list every finished pallet that may have been affected, and which customers received them.
• Reverse trace. Given a pallet code from a customer complaint, list every raw milk receipt that contributed to it, with farm and date of origin.

Both queries should resolve in minutes, not days. Plants that run mock recalls quarterly know the gap between their actual response time and the regulator's expectation. Plants that do not run them are about to find out under pressure.

The standards stack.

Three standards do most of the work in an Australian dairy MES rollout. None of them mandates MES specifically. All of them shape what the MES has to deliver.

FSANZ Food Standards Code 3.2.2.

The baseline food safety practices standard for all Australian food manufacturers. The relevant clauses cover receipt, processing, storage, and despatch records sufficient to identify the immediate supplier and the immediate customer of every batch. One-step-back, one-step-forward. The standard does not specify the format or system, but the practical reality is that a dairy plant cannot meet it on paper at the speed a real recall demands.

ISA-88 batch control.

The lingua franca of batch processing. The procedural model (recipe procedure, unit procedure, operation, phase) lets the PLC and MES layers carry batch identity end to end without bespoke per-recipe code. ISA-88 is not dairy-specific, but a dairy plant whose PLC code is structured to the standard makes the MES integration meaningfully easier. The ISA-88 article goes deeper on this.

ISA-95 enterprise integration.

The layering standard that defines how the MES talks to ERP above and SCADA/PLC below. ISA-95 part 3 is the relevant section for the operations management activities the MES owns: production, quality, maintenance, inventory. A dairy plant that grows from one MES to multiple sites typically inherits an ISA-95-aligned corporate architecture, and the conversation about traceability becomes a conversation about which level of the stack each piece of data lives at.

The retailer overlay.

Coles Supplier Portal and Woolworths Vendor Compliance both have evolving expectations around traceability evidence. Neither prescribes a specific system; both expect that a recall query from the retailer can be answered the same business day. Plants supplying private-label cheese, butter, or fluid milk hit this layer first, often before they have hit a FSANZ audit on the same scope.

The MES layer — what actually runs.

The realistic platform shortlist for a mid-sized Australian dairy is shorter than the vendor maps suggest.

AVEVA MES.

The successor to the Wonderware MES suite, integrated with AVEVA System Platform on the SCADA side and AVEVA Historian on the long-term data side. A natural choice for plants already standardised on Wonderware. The integration story is mature; the licence model is tag-tiered and rewards plants with stable tag counts (see the Ignition vs Wonderware spoke for the licensing detail).

Inductive Automation Ignition with a SQL-backed MES layer.

Ignition is not a packaged MES, but its SQL-bridge architecture and module library make it a credible base for a custom MES, especially in plants whose existing controls team is already comfortable with the platform. Per-server unlimited tags helps when the dairy plant inevitably grows its instrumented signal count over the rollout's life.

Silio (Pac Technologies).

Pac Technologies' batch-management product, currently deployed in production at a feed mill in Beaudesert QLD (Omron PLC backbone, EtherCat IO, Tencia ERP via SQL). Silio is designed for batch and recipe-driven food and feed manufacturers that want operator-in-the-loop traceability without the licensing weight of a corporate MES. See the Silio product page for the deployment detail and the Beaudesert project page for the engineering write-up.

Inherited corporate standards.

Plants under Lactalis, Saputo, or Bega ownership are often constrained to a corporate-standard MES — Rockwell FactoryTalk Production Centre, SAP MII, or a parent-company-bespoke layer. The integrator's job in those plants is to make the local controls layer talk cleanly to a stack the local team did not choose. The work is in the handshake.

A real rollout pattern.

A mid-sized Australian batch-and-blend operation engaged Pac Technologies for a semi-automatic blending system across three stations (Micro Room 1, Micro Room 2, Bulk Blending) at Beaudesert QLD in 2017. The architecture: Omron PLCs, EtherCat remote IO, RFID operator authentication, barcode-driven ingredient selection, gravimetric dosing, and integration with the Tencia ERP via SQL. The site is animal feed rather than dairy, but the trace shape is structurally identical: many raw ingredients into a batch, recipe-driven dosing, per-batch traceability writing back to the ERP.

The pattern that worked, and which transfers cleanly to a dairy rollout:

• Operator gate at every batch start. RFID auth ties the operator's identity to the batch record, not just the workstation. Audit-defensible from day one.
• Barcode-driven ingredient selection. The operator scans the ingredient sack or silo line, the PLC verifies the recipe expects that ingredient at this phase, and only then opens the dosing valve. The interlock prevents the most common human-error class without slowing the line.
• Gravimetric dosing with tolerance. Mass-based, not volume-based, with a recipe-defined tolerance band that the PLC enforces. Out-of-band events are logged with timestamp, operator, and corrective action.
• SQL-write to ERP at batch close. Not at batch start. The MES holds the batch record locally during the run, then writes the verified-complete record to ERP at close. ERP rollback during a failed batch becomes a non-event.

The same architecture, applied to a dairy plant, sees the ingredients become raw milk receipts and CIP cycles, the operator gate becomes the tanker driver's receipt log, and the SQL-write becomes the feed into the corporate MES or recall system. The shape is the same. The complexity moves with the volume of receipts per silo and the speed of the filler lines downstream.

The recall simulation most plants skip.

A traceability system that has never been tested under recall conditions is a paper system, regardless of how much it cost.

A useful annual exercise: pick a random pallet code from the despatch dock. Hand it to the QA manager and the production manager. Ask two questions. Which farms contributed to this pallet, and which customers received the rest of the same production batch? Time the answer to the minute. Anything over four hours indicates the trace data is not where the trace query needs it to be. Most plants find their first attempt sits between one and three business days. That is the gap between the system on the diagram and the system in operation.

Three failure modes turn up in nearly every mock recall:

• The pasteuriser cycle log lives in the PLC and nobody knows how to extract it. The trace ends at the silo and restarts at the filler, with a black box in the middle.
• The despatch system writes pallet-to-customer at the end of the day, not at the moment of despatch. Mid-day recalls have to chase the gap manually.
• The raw milk receipt records are in the lab system, not the production system. Joining them requires an export, a spreadsheet, and a coffee.

Each gap is solvable. The recall simulation is what tells you which gap to solve first.

Common questions.

Sources and further reading.

Industry and standards references for the claims above. Retrieved 18 May 2026.

• Dairy Australia. Australian Dairy Industry in Focus 2025. dairyaustralia.com.au
• USDA Foreign Agricultural Service. Australia: Dairy and Products Annual. fas.usda.gov
• Food Standards Australia New Zealand. Standard 3.2.2 Food Safety Practices and General Requirements. foodstandards.gov.au
• International Society of Automation. ISA-88 Batch Control Standard. isa.org
• IBISWorld. Milk and Cream Processing in Australia. ibisworld.com

This article sits under the Food & Beverage Automation guide. For the batch-control side of the conversation, see the ISA-88 article. For the OEE layer that sits above MES, see the OEE & SCADA article.