Imagine receiving a recall notice on an ordinary Tuesday. Are you actually prepared to respond quickly and effectively, even if you’ve met FSMA 204 requirements? The clocks are ticking.
One is the 24-hour compliance clock everybody has been preparing for. The FDA's Food Traceability Rule gives you 24 hours to produce a spreadsheet covering the critical tracking events for the lot codes in question. If you plan and build for FSMA 204 you will make that deadline. The International Fresh Produce Association (IFPA) describes the goal as compressing a foodborne illness traceback from 5-6 weeks to 5-6 days.
There's no deadline with the second clock, but it matters more. How much of the affected lot is still inside the building, where exactly is it, and can you get hands on it before it ships? Nothing in the rule addresses that.
FSMA 204 builds an audit trail that tracks product crossing a boundary. The FDA's traceability plan example for distribution centers, published in March 2025, spells this out. The compliance requirements in it are captured as warehouse management system (WMS) transactions. Inbound, unloading staff record a warehouse receipt capturing the Key Data Elements for the receiving critical tracking event. Outbound, pickers record products and their lot codes in a warehouse pick transaction. The point of contact the agency lists for the entire plan is an inventory control specialist.
Picking is the second place the record gets checked, but obviously that only applies to product that's actually picked. In a building with tens of thousands of locations, one day's activity may only touch a fraction of them while the remaining locations stay static in your WMS.
Most DCs weren't built to confirm a lot code on every case at the moment it's picked. The IFPA made the argument to the FDA in July, proposing to allow warehouse management systems to infer which lots went out instead. No decision has been made yet.
But let's look at the conditions IFPA suggested. Cycle counts. Inventory reconciliation. Empty slot verification. Linking received pallets to license plates, slots, and replenishment history. Their stated reason for the inventory accuracy piece is to keep the inferred lot range from getting too broad. Which is to say the industry's own proposed shortcut only works if you already know what's in your slots.
Now, let's imagine running that Tuesday recall notice across 12 facilities.
Each one may count on its own cadence, have its own labeling conventions, its own put away processes, or its own WMS. Good luck producing a single sortable and accurate spreadsheet across all of them in 24 hours.
Furthermore, 12 teams searching 12 buildings, each working from a record of different age and different quality, present a huge problem for corporate because there is no comparable measure of how good any of those records are. So recall scope balloons to cover any uncertainty. Product gets pulled that was never affected, out of buildings that may never have received it, because proving otherwise takes longer than the recall allows.
FSMA 204 fails to account for two critical scenarios inside the four walls of a distribution center: static inventory and inventory actively being moved within the warehouse. This is where autonomous data capture technology can fill in that gap.
Autonomous drones can fly the racks and cover reserve and ground stored inventory, where product sits longest, gets handled least, and has the most time to drift away from the record. That inventory is also the most challenging for a person to manually verify because it's tedious, time-consuming, and dangerous. Drones can scan and count on a schedule, so what they provide is a series of recent snapshots rather than a truly live picture.
Scanning systems mounted to forklifts and other material handling equipment track loads as operators handle them, updating the WMS in real time. These attached scanning devices also cover receiving and shipping, which is most important for FSMA 204 compliance. However, their blind spot is the opposite of the drone's because product that's rarely touched doesn't get updated in the WMS; so if something was put away in the wrong location, it could be weeks or even months before someone notices.
Two other approaches are viable solutions, but also have their limitations. Dock door scanning towers cover receiving and shipping events as product moves in and out of the building but they do not track inventory movement inside. Autonomous mobile robots scan static product from the floor but they need clear aisle access and clean floors.
Consider the impact these technologies provide compared to manually counting. The physical inventory count is no longer an event on the calendar and is instead a background process that's running every day. With constant tracking, you have fresh, accurate data you can rely on when a recall notice arrives.
At the facility level, you don't have to rely on a quarterly snapshot. Instead, you have an accurate history of what's inside your facility, and can learn where and when your records actually drift: which zones, which SKUs, which shifts, which put-away habits. That's where you look first when you have little time to comply with the recall.
At the network level, capturing data the same way across your sites is what makes the data accurate and comparable. Twelve buildings with 12 different tracking processes produce 12 datasets that have to be reconciled before anyone can ask a question across them. Twelve buildings that track inventory the same way provide a clean data set you can use to easily locate recalled products.
You should consider running a mock traceback test before FSMA 204 goes into effect so you'll actually know if your operation is ready.
But don't just review the records. Check whether your teams can physically locate and quarantine affected product inside your own four walls the same day, in every building, at once. Once you run that scenario, you will know where your gaps are.
The compliance date was pushed back 30 months to July 20, 2028. Companies that start consistently capturing data, the same way across their network, reach that date with an accurate history, comparable across sites, and recall processes already rehearsed. An operation that starts in 2028 may have a working automation system, but with a discombobulated dataset and inconsistent recall processes.
The technology is available now, whenever you want it. The years are not.
