In this guide
How an Auditor Decides What to Count
An auditor counting stock selects by value, not at random, because inventory populations are almost always skewed: a small proportion of lines carries most of the money. Value-weighted selection puts the effort where the risk is, and it is why a sample covering a fraction of the lines can still cover a substantial majority of the value. High-value lines are typically examined completely rather than sampled at all. The remaining population is stratified, and lines are drawn from each band so that no band goes entirely unlooked-at. A separate judgemental layer sits on top: items an auditor selects because something about them invites attention, such as a line that has not moved, a description that appears on many lines, or stock held somewhere unusual. What a sample can conclude is a statement about the population within stated limits. What it cannot do is prove that everything not selected was correct, and a report that implies otherwise is overstating its own evidence.
Value-Weighted and ABC Selection
Inventory populations are almost always skewed, and selection is built around that fact rather than in spite of it. Ranking by value contribution is the first step: every line is multiplied by its rate, the population is sorted, and the cumulative value is read down the list. What emerges in most businesses is that a small proportion of lines carries the large majority of the money, and a large proportion carries very little. A, B and C bands formalise the ranking. The A band holds the lines carrying the bulk of the value, the B band the middle, and the C band a long tail of items that together amount to comparatively little. Coverage is then set per band rather than uniformly, so the effort follows the money instead of following the line count. The top band gets full coverage because sampling it saves almost nothing and risks a great deal. Where a handful of lines carry a substantial share of the total, missing one of them leaves a material amount unverified, and the time saved by sampling within that band is trivial against the exposure it creates.
Random and Systematic Selection
Value weighting decides how much attention each band receives; something else has to decide which items within a band are examined. Randomness protects against bias, and the bias it protects against is not usually dishonesty. An auditor left to choose freely will gravitate toward items that are accessible, well labelled and easy to count, and those are systematically the items least likely to be wrong. A random draw within a band reaches the awkward ones. Systematic selection takes every nth item from a sorted list, which is simple to perform and easy to document. Its weakness is that any periodicity in the underlying order can align with the interval and produce a sample that is not representative at all. A list sorted by location, where each location holds the same number of lines, is the classic trap. Combining either with value weighting is what makes the whole approach defensible. The bands determine the coverage, the random or systematic draw determines the items within them, and the two together produce a sample that reaches most of the value while remaining unbiased inside each stratum. Documenting which method was used in which band is what allows the selection to be reviewed afterwards.
Judgemental Selection and Its Limits
Alongside the structured sample sits a layer of items chosen because something about them invites attention: a line that has not moved for years, a description repeated across many lines, stock held at an unusual location, an item whose recorded rate looks wrong, or anything the previous cycle flagged. This layer frequently produces the most useful findings, because it is directed at where the auditor's experience suggests problems live. What judgement can support is a finding about the items examined. If a judgementally selected line is wrong, it is wrong, and that stands on its own. What judgement cannot support is any statement about the population, because the items were not selected in a way that makes them representative of anything. A judgemental sample is not statistical and should never be presented as though it were. Reporting them together is where reports go wrong. The structured sample supports the conclusion about the balance; the judgemental additions support specific findings. Merging them into one coverage percentage overstates the statistical basis of the conclusion, and separating them costs nothing but a line in the method section. A reader can then weigh each layer for what it actually is.
Coverage the Report Has to State
Coverage is a disclosure, not a detail, and a report that omits it is asking to be relied on further than its evidence reaches. The percentage of value covered is the headline figure, stating what proportion of the total stock value was actually verified. The second figure is the number of lines examined, and it belongs beside the first rather than instead of it, because the two are routinely very different: a heavily skewed population lets a small proportion of lines carry a large proportion of value, and quoting only the line percentage understates the work while quoting only the value percentage overstates the breadth. Reporting both lets the reader judge the shape of the sample rather than accept a single flattering number. The uncovered residue then has to be characterised. It matters whether the unverified portion consists of many small lines, which is ordinary, or of a few substantial ones that could not be reached, which is a limitation. A report stating high value coverage while a material line went unverified has answered the arithmetic and not the question.
Setting a Sample You Can Defend
Match the coverage to the assertion being made. A conclusion about the total value of inventory needs a sample weighted heavily toward value; a conclusion about whether the counting process works needs a sample spread across locations, categories and handling points including the low-value ones. Those are different samples, and using one to support the other is the most common defect in this area. Decide which conclusion is wanted before deciding what to count. Document the basis before counting rather than after. The selection method, the value threshold above which items are examined completely, the stratification of the remainder, and any judgemental additions should all be recorded in advance, because a basis described afterwards cannot be distinguished from a basis chosen to fit the result. Full coverage is the only defensible answer in a few situations: a small population where sampling saves little, a contested balance where a sample will simply be disputed, and any case where the sample would leave a materially significant individual line unverified. Inventory audit scoping settles this question before the team travels.
