A microbiological criterion consists of (Codex Alimentarius, 1997):

• a statement of the microorganisms of concern and/or their toxins/metabolites and reason for that concern;
• the analytical methods for their detection and/or quantification;
• a plan defining the number of field samples to be taken and the size of the analytical unit;
• microbioligcal limits considered appropriate to the food at the specified point(s) of the food chain; and
• the number of analytical units that should conform to these limits.

A microbiological criterion should also state:

• the food to which the criterion applies;
• the point(s) in the food chain where the criterion applies; and
• any actions to be taken when the criterion is not met.

The sampling plan is defined by the number of samples tested from each food lot, and the size of each sample. As the number and size of samples increases, the probability of detecting the hazard in a contaminated lot of food increases, but so does the cost of sampling. Therefore sampling plans are designed to optimize the likelihood of hazard detection given finite resources. The purpose of this tool is to explore the impact of various types of sampling plans, in conjunction with microbiological limits, on the safety of a food product.

The unit of a sampling plan is a ‘lot’

Sampling plans use as their unit a ‘lot’ of food. A lot is defined as a “grouping of a product manufactured during a certain period of time or under the same conditions, or a consignment of food arriving at a border” (van Schothorst et al., 2009). Different lots can be expected to contain different mean concentrations of hazard, and the distribution formed by these concentrations is often assumed to be log-normal. When transformed to the log scale, this distribution follows a Normal curve, described by the “between-lot” arithmetic mean concentration and the standard deviation. Similarly, the mean hazard concentration among samples of food within each lot can also be considered to be log-normal and described by the “within-lot” arithmetic mean concentration and standard deviation when shown on a log scale. 

Alternatively, sampling plans can be analysed by assuming uniform or triangular distributions.

The sampling plan will influence the likelihood that a lot of food having a certain level of contamination will be accepted or rejected, and the proportion of lots that are included in the sampling plan (and thus risk rejection) influences the impact that sampling has on the safety of the food that reaches the marketplace, i.e. the risk reduction achieved.

A key distinction among sampling plans is whether the test result is based on the mere detection of the hazard, or on the level (concentration) of the hazard in the tested sample. "Presence/absence" plans are more common when dealing with hazards associated with severe health risks, for example Clostridium botulinum. Concentration-based schemes are preferred for less hazardous pathogens or for indicator organisms. These plans have a higher tolerance for contamination, and thus result in a less stringent “Level of Control” than presence/absence plans.