Early, structured and traceable preservation of technical facts, conditions, documents and samples as a reliable basis for later technical and legal evaluation.
Evidence collection aims to document technical matters in a way that allows them to be traced, analyzed and evaluated in a robust and defensible manner at a later stage.
Where appropriate, statistically supported sampling approaches can be used to select investigation samples in a traceable and representative manner. In doing so, I consider general principles of sampling, including approaches also used by European customs and tax authorities for representative sampling.
The sample size is determined according to the requirements of the individual case, contractual specifications and applicable standards. Retention samples additionally preserve the possibility of later re-testing and further evaluation.
Established methods for sample design may include DIN ISO 2859-1 for attribute inspection, DIN ISO 3951-1 for variables inspection, as well as further approaches from the ISO 28591 – 28598 series, ISO 11648, or general statistical methods. The appropriate method depends on the testing objective, the lot structure and the specific boundary conditions.
The preparation and execution of sampling also depend significantly on the storage and packaging concept. Relevant factors include the number of storage locations, palletized storage, the number of pallets, the number of packaging levels down to the smallest packaging unit, and the number of products per package.
To achieve the highest possible level of representativeness, multi-level randomization has often proven effective. Random numbers are used across the different levels of the storage and packaging concept in order to make the selection methodologically traceable and robust. Depending on the level, different procedures for generating random numbers may be used, where appropriate also with reference to standards such as ISO 24153.
Robust sample taking includes not only the selection of the specimens themselves, but also the control of the surrounding conditions. This includes documentation of the storage condition, where relevant including climate measurements, photographic documentation of the sampling process, clear identification and labeling of the samples, suitable packaging for shipment or storage, and, where necessary, transport with climate monitoring, for example by means of sensors, together with securing measures such as seals.
Early coordination with warehouse operators and other involved parties has also proven effective in practice. It is equally important to establish suitable on-site areas for sample taking. Sampling records and complete documentation are just as important as compliance with applicable legal, health and safety requirements.
Samples and results obtained on this basis are particularly suitable for statistical evaluation. This makes it possible to derive robust conclusions about very large lots even from relatively small samples. In many cases, the binomial distribution is an appropriate mathematical model, especially for outcomes such as defective / non-defective or conforming / non-conforming. Such tests are often based on Bernoulli experiments with two possible outcomes, a constant probability of success and mutually independent observations.
Depending on the data structure, results may also be evaluated using other distributions, such as the normal distribution, or by applying more general statistical methods.