Automate, Validate, Repeat: Essential Steps for Traceability Success with FSMA 204

By Andy Coutts

With FSMA 204 on the horizon, traceability in the food supply chain is critical. Complying with FSMA 204 also provides an opportunity to enhance operational decision-making. The success of a traceability deployment relies on accurately capturing time, temperature, and location data and transforming this into actionable insights.

For food safety leaders working through a traceability implementation, there are a few key things you should know.

Automate Everything

Radio Frequency Identifiation (RFID) articles often depict someone with a handheld reader, but using manual readers isn’t ideal for several reasons:

Inference software can fill data gaps, ensuring traceability even when infrastructure is limited.

• They negatively impact employee productivity.
• The process is unreliable, as expecting employees to scan at the right time consistently is risky.
• They assume the underlying processes are fully understood, which may not be the case.
• Equipment management can be challenging (e.g., “Has anyone seen the extra battery for the reader?”).

While handheld scanners can be useful in specific scenarios, such as locating a particular tag, the more reliable solution is automating the read processes. A traceability project proposal will detail infrastructure setup and tagging solutions. Automated reads can be done using RFID tags or Bluetooth Low Energy (BLE) beacons. RFID tags cost considerably less than BLE beacons, but BLE readers are less costly. The total cost of each solution will depend on the ratio of readers required to tags being read.

Site Visits Are Crucial

Once business objectives are defined, site visits are necessary for information gathering. Access to personnel with a deep knowledge of operations is essential. However, the information gathered should be treated as draft, as this may be the first time the site has implemented automated tracking.

For example, at one facility both the freezer and an adjacent reader picked up the same tags due to a cart blocking a sensor and leaving the freezer door open. These types of issues can be solved but may not be immediately apparent during initial visits. The real data will reveal such problems, often requiring additional visits and course correction.

Measure Twice, Cut Once

Business requirements can’t be fully costed out for FSMA 204 compliance without validating the proposed read infrastructure and tagging plans. This proof consists of two parts:

1. Data Acquisition Test: This tests whether the system captures and stores raw read data correctly. While this is the simpler part, the challenge lies in ensuring that the infrastructure can reliably read the full inventory at scale. For example, a shipping bay may have thousands of tags, while a fridge at a quick service restaurant may only have tens or hundreds. Each setup needs testing to ensure it can handle the expected volume.
2. Integration Tests: This ensures the system processes and converts raw data into meaningful insights. Anomalies may surface during this process, prompting additional site visits, and sometimes extra readers may be required. Project costing should account for these potential adjustments.

Collect and Retain All Data

Automated reader operations generate large volumes of raw read data. Transforming this into useful information requires converting raw data into stateful information, tracking a tag from when it is first attached to a container through to disposal or return.

Storing raw data without premature processing is recommended for several reasons:

• It ensures the data you’re analyzing is the same data that was posted.
• It allows you to test different inference rules for accuracy.
• It provides the ability to reprocess data when needed.
• It reduces operating costs by leveraging cloud services like serverless processing.
• Cloud storage is relatively inexpensive, and retention policies can be applied.

Mind the Gaps

Despite best efforts to track tags and associated information (such as geolocation and temperature), gaps in the traceability process can occur. These may be due to:

• Focused business objectives that install infrastructure only where it’s most needed.
• Budget constraints limiting infrastructure deployment.
• Third-party logistics providers that won’t allow infrastructure installation.
• Missed reads, where a tag moves from A to C but wasn’t read at B.

Inference software can fill these gaps by making educated guesses based on known data points. For instance, if no infrastructure exists inside a production facility, the system can still detect when assets arrive and leave to calculate asset utilization.

Tailor Information Access and Analysis to Stakeholders

Once raw data is processed into useful information, its analysis and distribution should align with the business objectives. At a minimum, data can be integrated with existing data assets. At the maximum, stakeholders may require specific analysis and access as part of the project plan.

Analytics can be complex, depending on the supply chain’s latency, which in food may range from short (e.g., raspberries) to long (e.g., frozen chicken). Time-series databases are often necessary for accurate comparisons between current operational data and prior periods. It’s important that complex analysis is part of the project’s original requirements to ensure the proper storage of information.

These are just a few things to keep in mind when deploying traceability projects that collect time, temperature, and location data automatically. Trusting the raw data, even when it’s initially difficult to interpret, has always proven to be a reliable guiding principle.

About the Author:
Andrew Coutts is a seasoned technology strategist and leader at Nazar Systems, specializing in digital transformation and innovation across complex industries. With more than 40 years of experience in software development and systems integration, he excels in creating solutions that streamline operations and enhance data security.