Illustrative case study. Company names and identifying details are fictional. Technical details reflect real industry practice and current regulation.
Flour lot traceability sounds easy: the tanker arrives, the lot number goes on the record, the flour goes into the silo. But the silo is never empty — each delivery layers onto the heel of the last, and after a few deliveries every pound is a blend of three or four lots: unknowable, inseparable, untraceable. At Pioneer Bakery, a mycotoxin-positive lot went into a silo feeding six weeks of production, and the “lot trace” covered everything the silo had touched: 2 million pounds, because the lot was everywhere and nowhere.
Background: a bakery on bulk flour
Pioneer Bakery (fictional) produced about 150,000 units a day — bread, buns, rolls, 80 employees — on bulk flour delivered by tanker, two to three deliveries a week, into two 50,000-pound silos. The receiving procedure recorded the lot number from the miller’s paperwork, and the lot number went into the traceability system. On paper, every production run could be traced to its flour lots. In practice, the silo made it fiction.
Challenge: the DON-positive lot
A routine incoming test — the bakery had recently started mycotoxin screening on incoming flour — flagged a tanker at 1.8 ppm deoxynivalenol (DON), above the bakery’s 1.0 ppm specification and FDA’s advisory level. The tanker had already been unloaded into Silo 1, which was holding about 20,000 pounds from previous deliveries. The flour was blended, inextricably, and the silo was already feeding the production lines.
The traceability exercise that followed exposed the system’s hollowness. The silo contained the current delivery plus the heels of the previous three — four lots, commingled, in unknown proportions that shifted with every draw-down and refill. The “lot trace” for any production run during the silo’s cycle had to include all four lots, because there was no way to determine which flour went into which dough. And the silo fed production for ten days per fill cycle, with the heel carrying over indefinitely — meaning the DON-positive flour, diluted but present, could have touched six weeks of production before the heel turned over completely.
The recall scope calculation was brutal: every product made while the implicated flour could have been in the silo — 2 million pounds across dozens of SKUs — because the traceability system couldn’t narrow it further. The DON levels in the finished product, tested extensively, were actually below concern in most lots (dilution across the silo worked in the bakery’s favor), but the company couldn’t prove which lots were clean, and “probably fine” isn’t a recall decision.
Investigation: the silo as a lot blender
The investigation mapped the silo dynamics with a tracer study — a food-safe dye added to one delivery, tracked through the silo’s draw-down and refill cycles. The results confirmed the worst case: the dye persisted through four complete fill cycles, detectable in the flour stream six weeks after the marked delivery. The silo wasn’t first-in-first-out; it was a continuous blender with a long memory, and every delivery’s identity dissolved into the heel within days.
The receiving records, meanwhile, were perfect — lot numbers logged, COAs filed, the traceability system populated. The system tracked what went into the silo with precision and had no idea what came out. It was traceability of the doorway, not the room. The miller’s lot numbers, carefully recorded, became meaningless the moment the flour crossed the silo threshold.
The mycotoxin screening program — the thing that caught the problem — had a timing flaw that compounded it. The test took 48 hours (external lab), but the tanker was unloaded on arrival because the production schedule couldn’t wait. The flour was always in the silo, in production, before the results came back. The testing was verification of history, not control of the present.
Root cause: the untraceable silo
1. Silo heels destroying lot identity. Continuous refill onto undischarged heels blended every delivery into an unknowable mixture. Lot traceability — the regulatory and commercial requirement — was physically impossible in the bulk system as operated. The system promised what the physics couldn’t deliver.
2. Testing slower than unloading. The 48-hour mycotoxin test couldn’t hold the tanker, so implicated flour entered the silo before results arrived. The control point (testing) was disconnected from the decision point (unloading) by two days and an irreversible physical act.
3. Lot definitions ignoring the equipment. The traceability system defined lots by delivery — the miller’s lots — while the equipment defined the actual lot as “whatever’s in the silo.” The mismatch meant every trace was simultaneously precise (the delivery records) and useless (the silo reality).
Corrective actions: making the silo traceable
Immediate: the 2-million-pound recall went ahead — the company didn’t fight the scope, because the scope was honest: the system genuinely couldn’t narrow it. The silo was fully emptied, cleaned, and the heel practice ended that day.
Within 60 days, the bulk flour system was reengineered for traceability: the silos were converted to run-to-empty operation — each silo is drawn down completely before refilling, with a documented empty verification (silo weight, visual confirmation) as the lot boundary. The two silos now alternate: one feeding production while the other is emptied, cleaned, and refilled with a single defined lot. The “lot” in the traceability system is now the silo fill — a physically meaningful unit, not the miller’s paperwork.
The mycotoxin testing went rapid and pre-unloading: a lateral-flow DON test ($15, 10 minutes) at the receiving bay, every tanker, before the hose connects. Tankers testing above specification are rejected at the gate — the flour never enters the building. The 48-hour lab test continues as verification, but the decision is made in ten minutes. Within 90 days, the bakery extended the run-to-empty discipline to its other bulk ingredients (sugar, oil), finding the same heel-blending problem in the sugar silo and fixing it the same way.
Results: the lot that’s actually a lot
Twelve months later: zero traceability failures, a mock recall tracing a flour lot through the silo system completed in under two hours (the previous best, on the old system, had been “inconclusive after three days”), and two tankers rejected at the gate by the rapid DON test — both confirmed over-specification by the lab, both kept out of the building. The run-to-empty operation, initially feared as a production constraint, actually improved flour freshness: no more six-week-old heels in the mix.
The QA manager’s summary: “We used to trace flour lots the way you trace a raindrop in a river. Now the silo is a glass, not a river — one lot in, one lot out, empty in between.”
Lessons learned: what you’d do Monday morning
Examine your bulk ingredient silos: are they run to empty, or refilled onto heels? If it’s heels, your lot traceability for those ingredients is fictional — every delivery blended into an unknowable mixture, every trace covering weeks of production. Convert to run-to-empty operation with documented empty verification, or accept that your recall scope for those ingredients is “everything.”
Then check your incoming testing timing: does the test result arrive before the ingredient is used? If the tanker unloads before the lab reports, your testing is archaeology, not control. Move to rapid methods at the gate — lateral flow, NIR, whatever fits the hazard — and hold the delivery until the result. Ten minutes at the gate beats a 2-million-pound recall.
Yeah, but actually — the tracer dye study is the image to keep: six weeks of persistence, the marked flour still detectable a month and a half later. That’s what a silo heel does to lot identity — it archives every delivery into a geological record no paperwork can untangle. A silo that’s never empty has no boundaries. Empty it. Verify it. Define the lot by the empty, not the delivery. And test before you unload — because once it’s in the silo, it’s in everything, and “everything” is the most expensive word in recall.