What Breaks in a Food Brand's Process When It Adds a Second Co-Packer

The second co-packer is supposed to be the easy part

You found a second co-packer because the first one ran out of capacity, or because a retailer wanted regional supply, or because the board asked what happens if the only plant making your product has a bad week. All good reasons. The trial runs went fine. The product tastes the same. Quality signed off on the audit.

Then, somewhere around the third month, the numbers start disagreeing with each other. Cost of goods by SKU moves for no reason anyone can name. Forecast accuracy drops on items that have not changed. A retailer asks a simple question about a lot, and answering it takes two days and four emails.

Every article you will find on this question says the risk is quality consistency, and answers it with audits, SOPs and a supplier scorecard. That risk is real, and it is the visible one. It is not the expensive one.

You had a relationship, not master data

A brand with one co-packer rarely owns its own master data. It owns a relationship that contains master data.

Think about where the following actually live today:

• Item codes and case configurations. The co-man set up your SKUs in its own ERP, with its own item numbers, and your team learned to translate.
• The bill of materials. The formula is yours. The working BOM, with substitutions, packaging components and the scrap factor, is theirs.
• Yield assumptions. Your cost model uses a yield. Which one? Usually the one the co-man quoted in year one, never updated.
• Lot code formats. Whatever scheme their plant uses, printed by their coders, recorded in their system.

None of this felt like a gap, because there was only one copy and it was always available by email. The first co-man was quietly acting as your system of record. You were paying for it in the tolling fee and never saw it on an invoice.

The second co-packer does not create the problem. It reveals it. Now there are two item numbers for the same product, two BOMs that are each correct for their own line, two yield figures, and two lot code schemes. Everything downstream that assumed one source of truth becomes two, and nothing tells you it happened.

Where it shows up first: traceability, with a clock attached

Of all the places this surfaces, traceability is the one with a regulator holding a stopwatch.

FDA's Food Traceability Rule under FSMA section 204 requires anyone who manufactures, processes, packs or holds foods on the Food Traceability List to assign traceability lot codes at defined critical tracking events, keep the key data elements tied to them, and produce an electronic sortable spreadsheet to FDA within 24 hours of a request. The compliance date moved to July 20, 2028, after FDA published a 30-month extension in the Federal Register in August 2025, and Congress directed no enforcement before that date in its appropriations act.

That extension reads as breathing room. For a brand adding a second co-packer it is closer to a warning. Two plants assigning lot codes in two schemes, recorded in two systems neither of which you own, is not a paperwork problem. It is a 24-hour problem. When the request arrives, somebody has to merge two exports with different field names, different date formats and different ideas of what a lot is, and do it correctly on the first attempt.

Even if your product is not on the Food Traceability List, your retail customers increasingly behave as if it were. The question "which lots shipped to which distribution centers" does not care which plant made them.

The quieter breaks

Traceability is the loud one. The others are slower and cost more over a year.

Cost of goods by SKU. Your margin report blends two plants with different tolling rates, different yields and different freight lanes into one number per SKU. The blended number is accurate and useless. You cannot see that one plant is profitable on a SKU and the other is not, which is the only reason you would ever want to know cost by SKU.

Forecasting. Demand history is now split across two ship-from points, often under two item codes. The planner who stitched them together by hand in a spreadsheet is your entire integration layer, and that is a single person on vacation away from a stockout.

Specifications. A spec change now has to land at two plants, on the same date, with both confirming they are running the new version. We have written about how spec counts multiply when each customer asks for a different format in the private label spec management question. The same duplication happens on the supply side, one plant at a time.

Cost comparisons. The reason you added a second plant was partly to have a price comparison. Two co-mans quoting against different definitions of changeover, scrap and run length are not giving you comparable numbers, a pattern we described from the other side of the table in why co-manufacturers lose money on changeovers they quoted correctly.

Why audits and SOPs do not catch this

Supplier audits are designed to answer one question: can this plant make your product safely and to spec. A second co-packer can pass every audit you run and still break your data, because the audit is not looking at who owns the item master. It is looking at the plant.

The SOP response has the same blind spot. You can write a beautiful procedure for onboarding a new co-packer, and it will specify sanitation, allergen controls, label approval and first-article sign-off. It will not say whose item number wins.

This is why the problem tends to be discovered by finance or by a retailer, not by quality or operations. The people looking for the break are looking at the product. The break is in the records about the product.

What not to do first

The tempting response is software. A traceability platform, a supplier portal, a PLM system, or the ERP you have been putting off. Vendors in each category will tell you, accurately, that their product solves part of this.

The order matters. Any of those systems needs a single item master to sit on. Buy one before you have decided what your canonical item, BOM, yield and lot structure are, and the implementation team will ask you those questions anyway, at consulting rates, in the middle of a go-live. Or worse, they will not ask, and the system will faithfully store both co-packers' versions side by side. You will have automated the split.

A brand that sets up a second plant is making the same structural move as a brand taking its first national retail order: adding a counterparty whose records do not match yours. We wrote about the version of this that shows up with retailers in what a DTC brand has to automate before its first big retail PO. The principle carries across. Be able to answer what you made, where, and in which lot before you automate the transmission of any of it.

The questions that decide it

We will not hand you a project plan here, because the right one depends entirely on answers that live in your contracts and your systems. But these are the questions that decide whether this is a two-week cleanup or a six-month program:

• If your first co-packer stopped answering email tomorrow, could you reproduce your BOM, item list and current yields from records you hold?
• Does your co-packing agreement say who owns the data generated about your product, and does it oblige the plant to give it to you in a usable format?
• Can you run a mock trace across both plants, for one SKU, without asking either of them for a spreadsheet?
• Is there one person at your company whose job it is to say which version of an item record is correct?

Most brands answer no to at least two. That is normal, and fixable. The fix is mostly decisions and contract language, with software arriving later and in a much smaller form than the vendors suggest. Adding a second co-packer is a good way to find out how much you owned all along. Usually it is less than the label says, which is a hard thing to digest.

Before the third plant

The cheapest time to fix this is before the second co-packer runs at volume. The second cheapest is now. Every month two plants run under two schemes adds history that someone will eventually have to reconcile, and that history is exactly what an FDA request, a retailer audit, or a buyer's diligence team will ask for.

If you have added a second co-packer, or you are about to, and you want an outside read on which of your records you actually own and which ones you have been borrowing, start a conversation. We will bring the questions, and we will tell you whether the answer is a contract clause and a spreadsheet before it is a system.