Sample-based count: drawing, size and documentation

A sample-based count saves time but has a weak point: the auditor does not see the whole warehouse, only the selection. If the selection does not convince, neither does the result. What matters is therefore less the counting, which is routine, than the draw: which stock was it drawn from, by what rule, by whom and when? This article shows how to set up a sample so that it can be followed. Whether the method comes into question for your accounts at all, and at what size, is for your auditor and head office to decide, not the warehouse.

Prerequisites before the draw

• A complete population: the stock extract of all items at the cut-off time, with quantity and value.
• A reliable day-to-day stock record, because the sample result is projected onto the total stock.
• Written agreement with the auditor on the method, the size and the way of projecting.
• Count instructions that describe the draw and the count separately.

If one of these is missing, a full count is usually the safer route.

Stratification: do not treat everything alike

A carton of screws and a control unit do not carry the same weight for the accounts. It is therefore usual to stratify by value: a few high-value items are counted completely, the large mass only in part.

A fictional example with 12,000 items in three strata:

• Stratum A, 600 high-value items: full count, so 600 positions.
• Stratum B, 3,400 medium-value items: sample of 10 per cent, so 340 positions.
• Stratum C, 8,000 low-value items: sample of 2 per cent, so 160 positions.

Altogether 600 + 340 + 160 = 1,100 positions are counted, about 9.2 per cent of the 12,000 items. At 40 positions per team hour and a second count, that is 1,100 × 2 ÷ 40 = 55 team hours. A full count would come to 12,000 × 2 ÷ 40 = 600 team hours. The percentages are freely chosen; which rates your auditor accepts is something you must ask beforehand.

The draw itself

A selection "by feel" is not a sample. Two methods are common and can be rebuilt in any spreadsheet:

• Random selection: each item in the stratum is given a random number, the list is sorted by it, and the first n lines form the selection.
• Systematic selection: after a randomly chosen starting point, every k-th item is drawn.

The procedure matters. The population is frozen as a file with date and time, the draw takes place in the presence of a second person, and the result list is saved before the count team sees it. After that nothing is swapped. Items with a book stock of zero also belong in the population, because goods may still be lying there. If the team cannot find a drawn item, it is not replaced by a more convenient one but recorded as a difference.

Result, projection and extension

Deviations are recorded per stratum. How the result is projected onto the total stock is something you agree with the auditor in advance; the calculation method belongs in the count instructions, not in a later explanation. The same applies to the extension rule: if the deviation rate of a stratum exceeds an agreed limit, the sample is enlarged or the area is counted completely.

In the example, 17 of the 340 positions in stratum B show a difference, which is 5 per cent (17 ÷ 340). If the agreed limit is 3 per cent, stratum B is extended. The 3 per cent is a freely chosen illustrative value.

What has to be documented

• The population: file, date, time, author.
• Stratum limits and the reasons for them.
• Size of the sample per stratum.
• Method and starting value of the draw.
• The draw list, saved before the count.
• Count result and deviations per stratum.
• The decision on any extension, with date and name.

In short

A sample convinces when every step of the draw is evidenced. AuditMaster draws the sample under a fixed scheme, counts the selected positions and records the draw, the result and the deviations in writing. Whether the method suits your accounts is something you clear beforehand with your auditor.