The use of a legally recognized sampling inventory procedure requires that certain tolerance thresholds must not be exceeded when determining the results. For the stratified mean estimation method (as well as for the related stratified projection methods), the relative sampling error and the percentage deviation between book value and estimated value may not exceed, for example, the 1% or 2% threshold.
A sampling inventory is efficient when these tolerance limits are just narrowly undercut with minimal data collection effort. For the user, this creates a dilemma: on the one hand, as few positions as possible should be counted; on the other hand, the number of required positions depends on the unknown warehouse structure and the accuracy of the inventory records. The sample size calculation is based on assumptions derived from the existing inventory records. If these assumptions do not match reality, the estimate may be too inaccurate, and the compliance of the inventory records is denied.
If one or both tolerance limits are exceeded „slightly“ (sampling error between 1.2 and 1.3%, percentage deviation between book and estimated value approximately 2.2%), a review of the identified deviations can be helpful. If the hypothesis can be established that the inventory records are generally accurate (i.e., only minor deviations were found, but also one or more outliers), it is permissible to expand the sample once and to examine additional sampling positions. The number of additional positions to be checked depends on the extent to which the permitted tolerance limits have been exceeded.
The number of additional positions must be calculated in such a way that the deviations found fall below the acceptable thresholds, and the compliance of the inventory records can be confirmed. A prerequisite is that no additional outliers are found in the extended sample.
Finally, it must be emphasized that expanding the sample is only permitted once. If the result again does not meet the required standards, a second expansion of the sample is not allowed and the extension of the sample will not be successful if the deviations exceed the permissible limits again.