Changeover and allergen sequencing records behind co-packer scheduling

Which records describe a changeover?#

A changeover at a co-packer is described by at least five records that are rarely kept together: the production run log, the sanitation record, the allergen verification result, the pre-operational inspection and the label or packaging verification. Each one is created by a different team, often on a different form.

Scheduling usually runs on none of them. In many plants the sequence lives in the head of one experienced planner who knows which products can follow which, how long a wet wash on line two really takes, and which brand insists on a full clean even when the allergen profile would not require it.

• 1Run logs: line, SKU, lot, start and stop times, and downtime codes.
• 2Sanitation records: clean type performed, crew, start and finish, chemicals and sign-off.
• 3Allergen verification: visual inspection results and any allergen-specific test results.
• 4Pre-operational inspection: release of the line before the next product starts.
• 5Label and packaging verification: confirming the right label and allergen statement for the next run.
• 6Holds and deviations: product held, reworked or rejected after a changeover issue.

The fields in a changeover matrix#

A changeover matrix is a table with one row for every ordered pair of products a line runs, holding the clean type that pair requires and the time it actually takes. Order matters: switching from a product without an allergen to one with it usually needs less cleaning than the reverse.

Build the matrix at the level the plant schedules. If many SKUs share a formula and allergen profile and differ only in pack size, group them into a product family so the matrix stays small enough to use.

How to build the matrix from actual run logs#

Building the matrix from actual run logs means pairing every product end with the next product start on the same line, then attaching the sanitation and verification records that fall between them. The steps below work in a spreadsheet before any software is involved.

Expect the first version to have gaps. Pairs that have never run show no history, and pairs that ran once show a single value. Mark those as estimates and let new runs fill them in.

• 1Export run logs by line with SKU, lot and start and stop times.
• 2Sort by line and time, and pair each run's end with the next run's start.
• 3Attach the sanitation record and verification result that fall between each pair.
• 4Look up each SKU's allergen class from the formula or specification system.
• 5Calculate actual changeover time from last good unit to first good unit.
• 6Group by line and product pair, and review outliers with the sanitation lead.
• 7Publish the matrix to planners and update it as new changeovers close.

Why allergen sequencing logic needs a record#

Allergen sequencing logic needs a record because it is the reason a schedule looks the way it does, and the reason is easy to lose. Running products without an allergen before products with one, grouping products that share an allergen, and scheduling full cleans at natural break points are common practices, but each plant and each brand applies them differently.

Food safety plans under FSMA may set allergen controls that shape sequencing, and many brand customers add their own requirements in the co-manufacturing agreement. Write the rules down beside the matrix, with the source of each rule, so a new planner or a scheduling tool can follow the same logic. Check your own food safety plan and customer requirements rather than relying on general practice.

What makes changeover data unreliable?#

Changeover data becomes unreliable when the start and end of the changeover are not recorded as separate events. The table lists the common failures and a fix for each.

Illustrative: a snack co-packer replaces tribal knowledge#

Illustrative: a fictional co-packer runs bars and powdered mixes for several brand owners on three lines. Its ERP holds orders and lots, run times are typed into a spreadsheet by shift leads, and sanitation forms are on paper.

The plant manager adds a changeover code with from and to SKU to the run spreadsheet, moves the sanitation form to a tablet with clean type and finish time, and builds a matrix by line. The matrix shows that one powder line loses far more time switching between two brands' products with the same allergen profile than expected, because one brand's agreement requires a full clean every time. Planning starts grouping that brand's runs together.

When a scheduling software vendor proposes an AI sequencing pilot, the co-packer already has the changeover matrix and written sequencing rules the tool needs. The COO reviews the vendor's data terms before sharing any history.

Who controls changeover records at a co-packer?#

Changeover records at a co-packer are usually its own operating records, while the formulas, specifications and brand names belong to the brand owners. The co-manufacturing agreement decides the details, and some agreements restrict how the co-packer may use any information about the brand's products.

That split matters for any use outside the plant. Changeover times, clean types, verification results and sequencing rules can often be shared with brand identities and formulas replaced by allergen class and product family codes. Formulas themselves stay out unless the brand agrees in writing.

How SourceX approaches co-packer run and changeover records#

For a co-packer, SourceX opens the SourceX five-step transaction (Supply, Rights, Preparation, Approval, Delivery) by asking which systems hold run logs and sanitation records, how far back they go and whether each changeover names the products on either side. The co-packer keeps ownership of anything it later licenses.

Rights work centers on the co-manufacturing agreements. In preparation, brand names, formulas and employee details are removed or coded, and the co-packer approves the final scope. Its SourceX Evidence Packet then shows provenance, which brand agreements were checked, what use is permitted, how employee names were handled and who released the data.