Skip to content

Audit Sample

Audit procedure — instantiates Representative Sampling Design

Selects records from a transaction universe by risk-weighted probability so findings support a bounded assurance opinion, with a documented trail any reviewer can re-walk.

Version
v1 · 2026-08-24 · History
Mechanism #
569
Type
Audit Procedure
Form family
Assessment, Review & Assurance
Solution family
Evidence, Inference & Validation
Problem family
Uncertainty, Evidence & Inference Failure
Problem subfamily
Sampling, Selection, Missingness & Generalization
Origin domain
Accounting & Auditing
Also from
Statistics & Experimental Design
Instantiates
Representative Sampling Design

Audit Sample draws from a defined universe of transactions or records for the purpose of accountability, and its distinctive move is to fuse three things a plain random draw keeps apart: risk-weighted selection (larger or higher-exposure items get a higher chance of being tested), a bounded assurance statement about the whole population, and an evidentiary trail so an outside reviewer can re-walk the exact path that produced the opinion. The sample exists to support a conclusion someone else will scrutinize, so reproducibility and defensibility are load-bearing, not optional. It works over a ledger of recorded values — this is what separates it from its twin, Quality Inspection Sample, which spreads units across a physical production process.

Example

An external auditor must opine whether a company's $80M accounts-payable balance is materially correct (figures illustrative). The frame is the full accounts-payable ledger for the year. Selection uses monetary-unit sampling: each dollar in the balance has an equal chance of being hit, which makes a large invoice far likelier to be selected than a small one — a risk tilt toward the items that could move the number most. About sixty invoices are pulled and tested against purchase orders and receiving records. Every parameter — the sampling interval, the random start, the materiality threshold — is logged in the workpapers.

Errors found are projected back to the population using the same design. The output is a bounded opinion: "no material misstatement in accounts payable at the stated materiality," reproducible by a second auditor from the recorded design rather than resting on the first auditor's judgment.

How it works

  • Define the universe and materiality. The population of transactions and the error size that would matter.
  • Risk-weight the selection. Monetary-unit sampling or explicit risk strata give exposure-heavy items a higher selection probability, while every item keeps a nonzero chance.
  • Test and project. Examine selected items, then extrapolate any errors to the whole population under the design.
  • Record the design. Log the parameters and rationale so the sample — and the opinion — can be independently re-walked.

Tuning parameters

  • Materiality threshold — a looser tolerance shrinks the sample and speeds the audit but raises the chance a real error slips through untested.
  • Risk-weighting strength — how hard selection tilts toward large or suspicious items; too hard and it stops being representative.
  • Sample size versus confidence — a bigger sample tightens the opinion but costs testing hours.
  • Low-risk coverage — how many "clean" low-exposure items to still sample so the opinion reflects the whole ledger, not only its flagged corners.

When it helps, and when it misleads

Its strength is producing a defensible, reproducible opinion that balances coverage against risk — the auditor tests where the money and the danger are, without abandoning the rest of the population.

Its failure mode is sampling risk: the sample can look clean while a material error hides in the untested majority, and tilting too hard toward suspicious items turns an audit sample into a suspicion sample that understates the clean-population error rate. Monetary-unit sampling guards the upside — material items are unlikely to be missed — but leaves small, systematic errors easy to overlook precisely because they attract little weight.[n1] The classic misuse is testing only manager-flagged items and then generalizing the finding to the entire ledger. The guarding discipline is to keep a probability backbone beneath the risk tilt, log every parameter, and project errors honestly rather than reporting only what was caught.

How it implements the components

  • sampling_frame — the transaction or ledger universe the sample is drawn from.
  • selection_method — risk-weighted probability selection, e.g. monetary-unit sampling.
  • sample_design_record — the workpaper trail that makes the design reproducible and the opinion defensible.
  • generalization_boundary — the bounded assurance opinion (population plus materiality) that findings support.

It does not define a target_population process, screen units with an inclusion_and_exclusion_rule, or run a coverage_gap_check across shifts and lines — that is its twin Quality Inspection Sample, which judges a running production process rather than a ledger of recorded values.

Editorial Notes

Form Classification

Form family: Assessment, Review & Assurance

Rationale: The mechanism selects and tests a risk-weighted sample, projects detected errors to the transaction universe, and supports a bounded assurance opinion, so its operative output is an audit finding.

Nearest alternative: Decision, Gate & Allocation — Sampling allocates records into the examined set, but that selection is subordinate to producing an evidence-backed assurance judgment about the population.

Review outcome: Adjudicated after independent review; high confidence.

Origin Attribution

Primary origin: Accounting & Auditing

Origin pattern: Single lineage

Present-day reach: Specialized

Rationale: Audit sampling, including monetary-unit sampling, selects ledger records to support a bounded assurance opinion and reperformable trail.

Related originating lineages:

Review outcome: Independent reviewer agreement; high confidence.

Notes

[n1] Monetary unit sampling (MUS), also called dollar-unit sampling — a probability-proportional-to-size method where an item's selection probability scales with its monetary value, so material items are unlikely to be missed while every item keeps a nonzero chance of selection. Its blind spot is small, numerous errors that individually attract little selection weight.