top of page

Alternatives to Traditional Food Safety Auditing

A Review of

Anyone in the restaurant, hotel or food manufacturing industries with responsibility for food safety will have heard the current buzzwords in relation to assurance services; ‘AI’, ‘Blockchain’, ‘Big Data’, ‘Remote Assessment’ and that plague on all our social media streams …. ‘algorithms’.


Also - in a post-pandemic, high inflation world, any services that do not directly generate a profit need to be as cost-effective as possible, and increasingly, traditional auditing - once the mainstay of the assurance industry, is being seen as inefficient as well as being vulnerable to international travel restrictions.


So any system that utilises remote assessments to remove travel costs and AI to minimise people costs should be attractive in a sector where margins are under severe pressure. But the challenge is that very few of the innovative assurance services currently on offer actually work in practice.


The limitations of traditional auditing are well known; in addition to the costs they can drive non-value adding behaviours; such as people creating documents for the sole purpose of passing audits, or even following auditors to give prior warning of their visits. Furthermore, by sending an external person to inspect an operation, there’s also the risk that the accountability for maintaining standards can become diluted, with the site managers blaming the auditors or even the head office safety team for any failures identified. However, the limitations of the alternatives to on-site audits are becoming clearer as businesses are starting to pilot innovative schemes.


To understand why alternatives to traditional audits have struggled to gain traction, we need to revisit the benefits of the traditional, on-site third-party audit. I’ve been listening to customers who’ve been considering the switch to remote systems, but who eventually decided to stick with traditional auditing in order to better understand the value of on-site audits. What they’ve been telling me is that while they were concerned about the rising costs and the unintended activity related to traditional auditing, they also appreciated the external oversight of their standards. Even the most diligent business can have a crisis and having an external organisation as part of your QA systems provides a stronger due-diligence defence as well as helping to maintain customer confidence. This benefit is reinforced if the audits are part of an accredited certification scheme. An additional benefit is that many Quick Service Restaurant (QSR) operations are run by franchisees, and the brand owners appreciate having an independent pair of eyes monitoring their operational standards.


An indirect benefit of traditional auditing that’s often overlooked - is that a structured audit performed by a trained and calibrated auditor provides reliable data that can be analysed and compared to produce valuable ‘Key Performance Indicators’ (KPI’s) that can be used for performance management or to identify trends. As the old Peter Drucker quote says; “If you can’t measure it, you can’t manage it.”


I’ve been involved with several forward-thinking food businesses who’ve been experimenting with mining their internal ‘data lakes’ to see if this big data can be re-purposed to provide food safety KPI’s. Unfortunately, the many variables in the reliability of the data, combined with the sheer volume has made this approach extremely challenging.


So what have been the main issues slowing the adoption of remote assessments using AI and big data? Looking at remote viewing technology first, this was going to be the future of auditing but quickly ran into concerns about personal privacy. While CCTV has been used for many years to monitor the actions of staff and customers in banks and shops, it is a very different proposition to extract measurable data from CCTV images. More recently, portable viewing technology such as body cams and smart-glasses with integrated cameras have been tried, but the patchy nature of cellular and WIFI signals in remote areas, or when walking into cold-rooms for example, has made this approach impractical in many situations.


Moving on to AI, the primary limitation is that in most cases it’s not Artificial Intelligence at all. True AI that involves machine learning – the ability of a computer to learn from past data without the need for programming – is still the reserve of high budget research project such as IBM’s Watson or Alphabet’s Deep Mind. Mass market AI systems such as Google Cloud AI are becoming available for high value areas such as customer contact centres, but it will be a while before this proprietary technology trickles down to niche application such as food safety. So for now, ‘AI’ is just a marketing term for the application of basic algorithms to perform routine data processing tasks. The limitation to an approach based on algorithms is that you need to know the calculations that should be applied to the data in advance, in order to programme the system - and this requires an in-depth knowledge of the specific operations of the food business and the relationships between their different data sets. This makes it practically impossible to develop an effective ‘off-the-shelf/plug and play’ food safety system for use by customers.


Moreover, the biggest limitation with a proprietary assurance system is that it’s essentially an internal auditing system, lacking the independent oversight of a trusted auditing organisation - and so may not provide the level of confidence required by customers or enforcement organisations. So what is the future of computer assisted remote assurance for food safety? I don’t want this article to be an advertorial for the current project I’m working on with Supply Chain In-Sites – but it does serve as a useful demonstration of the probable trajectory.


At SCI we’ve been pioneering a hybrid approach to food safety assurance. Data from a carefully designed dataset is collected from the customer’s operations via a mobile App used by site management. This avoids the issues of excess and unreliable data when sourcing from data lakes, whilst also keeping the focus of responsibility firmly on local management. The data is then analysed using experienced auditors via desktop review. These experts apply their experience by using relationships within the supplied data to verify the reliability of the information. As we gain more experience through repeated desk-top reviews, we’re able to translate the expert analysis into algorithms, enabling a degree of automation to reduce costs. The resulting ‘verified data’ is then used to generate the customer dashboards and provide the essential management KPI’s.


In addition, sample ‘spot-checks’ are made to selected sites to perform an in-person review of the supplied data to monitor the validity of the overall system. This is the final stage of the verification process and is a parallel to the way financial audits operate - when external auditors review a company’s accounts. This hybrid approach combines the independent oversight of a traditional audit with the efficiency of remote auditing, whilst also providing a high level of confidence in the data. In the future, when real AI systems become available at commercially viable costs, we can reduce the ‘expert review’ stage as the system will use machine learning to constantly update the algorithms, and in addition Blockchain technology can be applied to provide data security and traceability - but for now we believe a combination of artificial and human intelligence is the optimal approach.


To summarise as an equation …..


AI – HI = AS (Artificial Intelligence minus Human Intelligence equals Artificial Stupidity)

0 views0 comments

Recent Posts

See All