To balance the stock account, every company must perform a physical inventory of its warehouse stocks at least once in a fiscal year. The time and costs of a physical inventory can be considerably reduced if only stocks of individual materials are determined and the count result extrapolated for all stock management units. However, such conclusions (from several counts of all stock management units) are only allowed when certain criteria are fulfilled:

• The scope of the stock management units involved in the physical inventory must be large enough.
• The selection of stock management units to be counted must be "representative".

Mathematical/statistical processes are used to perform the inventory sampling:

• A sample to be counted is randomly selected from the stock management units to be inventoried. The size of the sample depends on the probable degree of confidence that is valid for the projection of individual count results on all the stock management units involved.
• An extrapolation is carried out for all stock management units included in the inventory sampling, based on the count results posted.
• If the variance between the extrapolated value and the book value is small, and if it is probable that there was only a very small error in the extrapolation, then the inventory sampling has been successful.

In the case of a successful inventory sampling, it is assumed that the estimated variances from the book inventory are so small that they can be safely ignored. Only the stock of items, which are actually counted, is corrected in the same way as for the classic physical inventory. The stock of the remaining stock management units does not change.

If the inventory sampling is not successful, then a complete physical inventory must be performed for the uncounted stock management units.