It is a well-known fact that sampling as well as the analysis uncertainty influence the final analysis result, however, the work concerning development and optimization of methods has traditionally attracted the most attention. The reality is more or less worthless analysis results – in spite of quality-controlled methods – if the sampling is not representative for the basic material.
Within the food sector, a representative sampling is crucial for risk assessments. Also for elaboration of valid nutrient labelling, or when the purpose is to calibrate indirect methods for quality-controlled food, the representative sampling is of the utmost importance.
The Danish Meat Research Institute (DMRI) has previously worked on obtaining up-to-date data on nutritive value in beef and veal. In the current EU project, TRIMSCAN, work is performed on test of a prototype for in-line determination of fat in trimmings from beef and pork. In both projects, planning and execution of a representative sampling have been essential. Broadly, the work has followed DS-3077, which is a matrix independent standard, based on Theory of Sampling (TOS). The process is illustrated in Figure 1 (See attached document) showing the connection between sampling steps and sources of error, three governing principles (heterogeneity, variography, lot dimensionality reduction) and four sampling unit operations (mixing, particle size reduction, composite sampling, mass reduction).