The difference between recalling a SKU and withdrawing a lot is the difference between a seven-figure write-off and a contained event. The capability that converts one into the other is lot-grain custody — knowing, from the record, which lot went where — not better crisis PR. Here is the math, with every assumption on the table so you can rerun it with your numbers.
The model, published
One mid-market brand, one implicated SKU, everything co-manufactured. The assumptions are deliberately ordinary; change any of them and the ranking below does not change, only the spread.
| Assumption | Value |
|---|---|
| Annual volume, this SKU | 2,400,000 units |
| Cost of goods per unit | $1.80 |
| Retail price per unit | $4.49 |
| Units in the channel at any moment (6 weeks of supply) | 280,000 |
| Lots in the channel at any moment | 26 |
| Units in the implicated lot still in the channel | 10,800 |
| Retail chargeback and handling, per affected store | $180 |
| Stores carrying the SKU | 3,100 |
| Weeks of lost shelf velocity after a full pull | 4 |
Brand-damage cost is excluded on purpose. It is real, it is the largest line, and it is the least modelable — so this model understates the case for grain rather than overstating it.
Three grains, three bills
SKU grain. You cannot say which lot went where, so the answer to the retailer is "pull everything." All 280,000 channel units are destroyed at cost ($504,000). Every one of 3,100 stores executes a pull ($558,000 in chargebacks and handling). Four weeks of lost velocity on 46,000 units a week at your margin runs another six figures. The bill starts at roughly $1.2 million before a single brand-impact dollar.
Lot grain. The trace answers the actual question: lot 24 shipped in 312 cases to two of your three retailers' distribution centers, and the record shows which stores received it. The destruction set is 10,800 units ($19,440). The store pull touches the subset that received the lot — call it 400 stores ($72,000). The shelf largely survives because the SKU stays on it. The bill lands near $95,000 — about 8% of the SKU-grain number.
Unit grain. Serialized capture narrows further: units of lot 24 already sold through are known, so the destruction set is only what is still in the chain, and the retailer conversation gains a precision that changes its tone entirely. The marginal saving over lot grain is real but smaller; lot grain is where the curve bends.
The retailer conversation is the model's hidden variable. "Pull everything, we're not sure" costs you standing that outlasts the incident. "Pull these 312 cases, here is the record" is a different relationship.
Where grain actually comes from
Trace grain is not a software feature you buy during the incident. It is a property of what your co-manufacturer encodes on the case and what gets captured at each hop afterward. If the case label carries lot and the receiving scans are recorded as events — conformant EPCIS 2.0 events, each carrying what, when, where, why, and how — then the withdrawal-scope query is a lookup. If the lot lives in a batch record at the plant and nowhere downstream, then grain dies at the first handoff and no dashboard resurrects it.
That is why the capability is bought before the incident or not at all. The record accrues in ordinary operation; the incident merely reads it. A brand with no supply-chain department — the reader this brief is written for — does not staff this. You buy the outcome: capture happens where your co-man and your distributors already scan, and the trace answers arrive as answers. The consumer-signal read covers what the same marks can do for you on their best day; this is what they do for you on the worst one. For where the elapsed time actually goes when a trace crosses organizational boundaries, see the mock-recall record at the franchise boundary — a different industry, the identical failure shape.
FSMA is the floor, never the reason
FSMA 204 puts a regulatory floor under exactly this math for covered foods — and the honest read of the record is that the date is a de-risking floor, not a gun to the head, with FDA's own proposed extension and an appropriations enforcement bar both on the public record. Run the program for the withdrawal economics; treat the rule as the reason your board stops debating it.
Run it with your numbers
Take the table above, substitute your volume, your channel depth, your store count. If the lot-grain bill comes out above 20% of the SKU-grain bill, your channel is unusually shallow — and the case still clears. Then get started: the interview takes a few minutes, branches on how your supply chain actually runs, and tells us enough to be useful to a brand like yours.