ISO 22000 clause 8.5.2.3: assess which food safety hazards are significant

What clause 8.5.2.3 requires in practice

For every hazard identified, the organisation carries out a hazard assessment to decide whether preventing it, or reducing it to its acceptable level, is essential. Two things are weighed: the likelihood of the hazard occurring in the end product before any control measures are applied, and how serious the harm to health would be, judged against the intended use described under 8.5.1.4. Hazards that come out as needing control are the significant food safety hazards. The method used is described, and the results are kept as documented information. The clause does not name a scoring system, a matrix size or a threshold, so the organisation may choose its method as long as it covers both factors and can be explained.

Assess the identified hazard, not a vague category

Use the hazard and location established through hazard identification. If the description is too broad, the team may struggle to assign meaningful likelihood or severity. Resolve the underlying information before relying on a score.

Consider the evidence relevant to occurrence in the end product before applying control measures. Avoid treating an existing or proposed measure as proof that the underlying hazard is insignificant. That reasoning can remove the very justification for maintaining the measure.

Evaluate severity in the context of intended use and relevant consumers. The assessment concerns adverse health effects, not simply the commercial cost of a complaint or the inconvenience of rejecting product.

Describe and apply a suitable methodology

The standard requires the methodology to be described and the results maintained as documented information. It does not prescribe a universal numerical matrix, threshold or scoring formula. Choose an approach that addresses the required considerations and can be explained.

If categories or scores are used, define what they mean and how evidence supports them. Otherwise, two team members may use the same number to describe different judgments. Review borderline or inconsistent results with appropriate technical input.

Record the basis sufficiently for later review. A coloured cell without reasoning can be difficult to reassess after a material, product or knowledge change. The method should support decisions, not create an appearance of precision beyond the evidence.

Use significance to guide the next decision

Identify the significant hazards that need appropriate control. The next step is selecting and categorising control measures, not automatically declaring every significant hazard a CCP.

A hazard can require a combination of measures. The assessment establishes the need for prevention or reduction; the selection process examines how to achieve it and how the measures should be managed. Keeping these stages separate makes the reasoning easier to defend.

Why clause 8.5.2.3 matters to the business

Significance decides where effort goes. If the assessment is vague, resources are wasted on trivial hazards while serious ones receive the same light attention. A transparent method also protects the business in disputes: when a customer questions why one hazard is controlled and another is not, the organisation can show its reasoning. Severity judged against the real consumer, including vulnerable groups, links the assessment to genuine harm rather than to commercial inconvenience. A sound method that is applied consistently also reduces arguments inside the team about scores.

How to implement clause 8.5.2.3

Choose a method the team understands, define its terms, and apply it to every identified hazard. Keep the language of the method plain, so that supervisors and operators who join later meetings can follow why a hazard was judged significant and challenge it sensibly.

1. Pick an approach, which may be a matrix or a structured description, and define what each likelihood and severity level means in your products.
2. Score likelihood as it would be without controls, using process knowledge, history and external information, not the hope that existing controls work.
3. Judge severity by the adverse health effect for the intended use and the relevant consumers, not by cost of a complaint.
4. Have a technically competent person challenge borderline or inconsistent results.
5. List the hazards that come out as significant and pass them on to control selection.
6. Describe the methodology in the documented information, with the results and the reasoning for any judgement calls.
7. Rerun the assessment when intended use, process or knowledge about the hazard changes.

Illustrative food-manufacturing example

Illustrative catering example

A catering team compares two service arrangements involving different intended users and handling conditions. It checks whether a shared severity and likelihood judgment is justified or whether the differences matter to the hazard assessment.

The team does not create different scores simply because the contracts have different names. It records relevant technical differences and uses suitable expertise where the implications require it.

Evidence and common analytical errors

Common nonconformities with clause 8.5.2.3

Auditors tend to find the same patterns when they test how hazard assessment was done.

• The method is not described, so nobody can say why one hazard scored higher than another.
• Likelihood is scored with the existing control already assumed to work, which makes the hazard look insignificant and hides why the control is needed.
• Severity reflects business cost or complaint impact instead of health effect for the intended consumers.
• Only hazards already controlled at a CCP are assessed, leaving other identified hazards without a recorded decision.
• The assessment was done once and has not changed since a new product, customer group or supplier was introduced.