Sample-based stocktaking: count less, stay compliant?

What is sample-based stocktaking and when is it allowed?

In sample-based stocktaking only a randomly selected share of stock positions is physically counted. A recognised statistical procedure then extrapolates the total inventory value from that sample. Under recognised accounting principles the method is permitted as long as its evidential value matches a full physical count.

For that equivalence the standards require a confidence level of at least 95 % at a relative sampling error of no more than 1 % of the total value. If both limits are met, the extrapolated valuation is treated as equal to a complete count for balance sheet purposes.

Which extrapolation methods exist?

Two approaches dominate in practice: free extrapolation (mean estimation) and difference estimation. Which one fits depends on how well the book quantities match physical reality.

Difference estimation is the most common approach because it uses the existing book values: only the deviation between target and actual is counted. As these differences usually vary little, far smaller samples suffice than with free extrapolation.

• Stratify the population sensibly (e.g. by A/B/C value classes).
• Count high-value items (A parts) fully, sample the rest.
• Screen out zero balances and extreme values in advance.
• Draw the random selection by software, not by intuition.

How large does the sample need to be?

The sample size follows mathematically from the required confidence (95 %), the permitted error (1 %) and the spread of stock values. The more homogeneous the warehouse, the smaller the sample. In practice only 5 to 15 % of positions are actually counted.

Crucially, the absolute sample size grows less than proportionally as the warehouse gets bigger. That is exactly why the method pays off most in large small-parts stores with many low-value positions. If the extrapolation breaches the 1 % limit, the sample must be enlarged or the method changed.

What do auditors look for in the implementation?

For the auditor to accept sample-based stocktaking, the procedure must be documented transparently. The decisive points are a clean random selection, a recognised estimation method and proof that the error limits were kept.

• Description of the chosen extrapolation method.
• Proof of random selection including a drawing protocol.
• Calculated relative error and confidence level.
• Target/actual reconciliation for every counted position.
• Documentation of stratification and full-count classes.

Frequently asked questions