In June of 2019 Primority set out to build an AI powered food supply chain monitoring tool called AI Scan as part of an Innovate UK funded project in collaboration with the University of Portsmouth. When the project started, it was clear that there was an opportunity to access data about the food supply chain. This data was freely available under Creative Commons or Open Government licensing from various trusted sources.
More than 50 international data sources were identified. It was decided to focus on the largest and most credible sources as follows:
US Food and Drug Administration
UK Food Standards Agency
Canadian Food Inspection Agency
Food Safety Authority of Ireland
Food Standards, Australia & New Zealand
EU Rapid Alert System for Food and Feed (RASFF)
The main challenge was that the data was stored in different formats in each data source. There was no standard way of describing a food safety alert which made it difficult to read and interpret this information. This is where AI (Artificial Intelligence) became useful. After gathering and cleaning the data from each of the sources it was transformed into a format which AI could work on.
AI machine learning was used to ‘read’ the information and identify company names, product names, brands, and food safety hazards. Through an iterative process, machine learning was used to increase accuracy of entity identification to more than 90% in most cases.
Having identified the names of the companies involved and the food safety hazards other information of interest, such as import alerts and regulatory notices, was captured. The resulting data was compiled into a database of the world’s food companies so that the compliance history of each organisation could be analysed, and risk assessed. Only then was it possible to analyse hazards in the food supply chain.
A configurable risk model was then created that could be used to benchmark risk across a food company’s suppliers. This made it easy to identify supplier performance over time. This is important because risks constantly change in the food supply chain, as does regulation, ownership of businesses and environmental factors like war, famine, drought, and politics. All of which can have an impact on food supply chains.
To implement a configurable risk management system, it is important consider the risk appetite of the customer. After a detailed analysis, key factors were identified on how to create a risk scoring model that was flexible and created a consistent view on supplier performance. To do this the following were assessed:
Number of food safety events over the past 5 years
Number of allergen events over 5 years
Number of import refusals over 5 years
Number of regulatory actions over 5 years
Expired Certification
Other factors considered were:
The impact of the actual event on our customers business
How long ago the event happened (more recent is more important)
What factors are important to the customer (weighting of risk events)
Using the information gathered a risk scoring system was created which resulted in an overall score that was completely automated.
The result is an AI-powered supply chain monitoring tool with built-in risk scoring which can be used to benchmark supplier performance with respect to compliance and the food safety performance of each food company based on data from trusted sources.
In addition, many supply chain issues, which otherwise may go undetected, do not end up on a regulatory database. To address this problem AI was applied to web searches using the Google Search API with a customer-configurable list of keywords of interest. This helps detect other issues which our customers may want to be aware of. For example, if the owner of a particular supplier was prosecuted for fraud, food safety or child slavery. This approach yielded additional intelligence from non-trusted sources that could help detect a wider set of issues.
Finally, a deep analysis of the food safety hazard data was required to detect ‘unusual goings on’ that we called anomalies. A unique anomaly detection algorithm was developed which looks for patterns like ‘newness’, i.e., incidents that have not occurred before, or incidents that have not happened for a long time. This resulted in an anomaly score being assigned to food safety incidents which may be able to help with early detection of important supply chain issues before they affect our customers business.
AI Scan is the kind of tool that FDA refers to in it’s “New Era of Smarter Food Safety”. It is aligned directly with the strand of the blueprint for “Tech Based Data Analytics (Using AI to Scan External Information)”. Tools like this will be more commonplace in future because they help identify and reduce risk in the food supply chain.
It is hoped that you found the information presented here useful and thought provoking. At Primority we have a vision for what the future food supply chain looks like, not only in the future, but now. If you have any questions, feel free to reach out.
About Primority
Primority Ltd is a cloud technology company based in Atlanta and the UK who focus on digitally integrated food safety and quality management systems. Our cloud-based 3iVerify software is designed to help food companies and food importers easily manage all aspects of food safety compliance through the following modules:
Vendor / Supplier Approvals / FSMA and FSVP Compliance.
Automated Supplier Certification and Document Management.
Raw Material / Food Product Approvals and Specifications.
AI Powered Automatic Supplier Monitoring (AI Scan).
Internal GFSI Standards based Document Control.
Data Capture to Support Traceability, Goods In checks, HACCP and Quality.
Fully Integrated Corrective and Preventive Actions
Dashboard and Reporting Tools
3iVerify’s modular nature and monthly subscription model ensures that customers can buy only what they need and add on additional modules as their needs change and evolve. This makes our solution affordable, quick, and easy to deploy and delivers benefits quickly.