All Articles
Company

Why We Built Loamy: The Phone Tree That Broke During a Recall

California Central Valley farm field at sunset with dark loam soil rows

Two days. That is how long it took a produce shipper I was working with to identify the affected lots after a bagged salad recall. The recall notice arrived early on a Tuesday morning. By Thursday afternoon they had their answer. Forty-eight hours of phone calls, email threads, spreadsheet reconciliation, and warehouse staff pulling physical boxes to read lot codes under fluorescent lights.

The certificates existed. Every supplier they worked with sent certificates of analysis. There were pesticide residue results, lot identifiers, origin declarations, handling condition confirmations. All of it existed. The problem was that it lived in six different email inboxes, a shared drive folder with three years of unsorted PDFs, and one person's notebook where she had written down supplier lot codes by hand at the receiving dock because the WMS field for that information was never configured. When the call came in asking "which of your outbound shipments contain product from this grower, these dates," there was no system to answer it. There was only people, memory, and time.

The Moment the Phone Tree Broke

What I watched happen over those two days was what I now think of as a phone tree under pressure. The supply chain coordinator called the packing shed. The packing shed called their grower relationship manager. The grower relationship manager found the harvest records for the relevant date range and called back. Meanwhile the shipping team was pulling bills of lading to identify which orders included product from that grower in that window. Someone in QA was trying to match the lot codes on the pulled bills of lading to the lot codes on the certificates in the email archive. They were working with four different lot code formats across three suppliers, none of which matched the lot codes in the WMS. The matching was manual. The margin for error was high.

By the time they had a confident answer on which shipments were affected, the 24-hour window that will soon be mandated under FSMA 204 had already passed twice over. Product that should have been on hold was still at retail. The remediation effort expanded because the scope of the recall could not be confirmed quickly enough to contain it.

I had been working in supply chain software for several years before this, mostly on logistics and inventory tools. Watching this event unfold made it clear to me that the core problem was not the certificates themselves. It was the absence of a structured connection between certificate data and lot identity. The data existed. The structure to make it queryable did not.

Why the Problem Persists

After that recall event I spent several months talking with food safety officers, supply chain coordinators, and operations managers at produce shippers and prepared-food manufacturers across California's Central Valley. What I found was that the situation I had observed was not unusual. It was the norm for operations that had not yet invested in a dedicated traceability system, which, before FSMA 204, included most operations below a certain size.

The reasons were predictable. Dedicated traceability systems were expensive, often requiring significant integration work with existing WMS or ERP systems that themselves did not capture lot-level data at the required granularity. Compliance was not yet mandatory for many operations. And the primary business driver for traceability (recall response speed) was an event most operations hoped would never apply to them. So the investment was deferred, and certificate management remained a file-and-forget exercise rather than a structured data collection activity.

FSMA 204's 24-hour records response requirement changes that calculus for covered operations. The question is no longer whether you will ever need to answer an FDA lot trace request quickly. The question is whether your records will support a 24-hour answer when that request arrives. For operations still relying on email inboxes and spreadsheets, the honest answer is: probably not.

What We Decided to Build

Priya Srinath and I had been thinking about this problem from different angles. I had been watching it play out in supply chain operations. She had been building data pipelines for document-heavy compliance workflows, including earlier work on traceability infrastructure at a fresh produce logistics platform. When we compared notes, the opportunity became clear: the bottleneck was not the hardware or the certificates themselves, it was the step between a certificate PDF and a structured, queryable lot record.

Certificate of analysis parsing is a classic unstructured document problem. Every supplier has a different certificate format. Some are professionally laid out PDFs from accredited labs. Others are Word documents with inconsistent field names. Some include lot codes in headers, some in footers, some in the body of the test results table, and some only on accompanying packing slips that are not always submitted with the certificate. The information is there, but extracting it reliably at scale requires more than keyword matching.

We built Loamy around a document parsing layer that extracts the key fields from supplier certificates regardless of format, and feeds that extracted data directly into a traceability graph. Each certificate becomes a node with typed fields: the lot code, the origin farm or facility, the harvest or production date, the certificate type, the test results relevant to FSMA 204's Key Data Element requirements, and the expiry or validity date. Those nodes connect to lot records, which connect to receiving events, which connect to shipping records. The graph is the answer to the recall query: when a lot is identified in a recall notice, the graph returns the full chain of affected downstream shipments and the supporting certificate records for each lot in the chain.

What Loamy Is and Is Not

Loamy is a tool for food producers and supply chain teams who need to manage supplier certificate intake and build a structured traceability record from that intake. It is not a regulatory authority and it does not certify compliance. We help you get your records into a form that supports FSMA 204 compliance. Whether your records actually satisfy an FDA inspector's requirements depends on the completeness and accuracy of the certificates your suppliers provide and the records your team creates at each Critical Tracking Event.

We are also not a food safety testing laboratory, a GFSI certification body, or a substitute for your HACCP plan. The traceability layer we build sits alongside your existing food safety management system. It is specifically focused on the chain-of-custody record problem: who gave you what, when, under what documented conditions, and where did it go after you shipped it out.

We are a small team in Davis, California. Charles, Priya, and Marco Benetti, our Head of Compliance, who brings years of practical FSMA 204 implementation experience from his time with a produce inspection consultancy. Three people who spend most of their days talking with produce shippers, fresh-cut processors, and food service suppliers about their certificate workflows and building the software that makes those workflows tractable. That is the context from which Loamy comes.

Why Davis

Davis is not an accidental choice. The University of California Davis campus is one of the leading agricultural research institutions in the country. The Central Valley, which produces a significant portion of the nation's fresh produce, is an hour east. The community here has a density of people who understand agrifood supply chains from the ground up: researchers, extension agents, grower association staff, food safety consultants who have been working FSMA compliance since the rule was first proposed. When you are building software for produce supply chains, being located in this ecosystem means the feedback loop is short. We can drive to a packing shed and watch a certificate arrive at the receiving dock in the same afternoon we are debugging the parsing pipeline.

The phone tree that broke during that 2024 recall was not an isolated failure of one operation. It was a preview of the compliance gap that FSMA 204 enforcement is now surfacing at operations across the country. We built Loamy because we believed the solution was tractable and the cost of not solving it was becoming real. We are still early, and we are honest about that. But we are working on the right problem.