In a job of hundreds of pages it is not always possible to evaluate every sentence independently. A sampling audit examines one part and reaches a judgement on the whole. The validity of the method depends on one thing: that the chosen part really represents the text.
Definition
It is a judgement on the whole text made by examining part of it. The method is statistical. It has conditions.
When it is used
It is used in high-volume jobs and in acceptance evaluation. It is not needed in small jobs. Scale is decisive.
Difference from full revision
Sampling is an evaluation, not a review step. It does not replace revision. The distinction is preserved.
Not a replacement for revision
The whole text is still reviewed. Sampling is for measuring quality. The two tasks are different.
Representativeness
The chosen section must reflect the whole text. Otherwise the result misleads. The condition is fundamental.
Selection method
The section is chosen at random or by a defined rule. Arbitrary selection spoils the result. The method is written down.
Not choosing easy sections
Choosing simple sections improves the result but does not show the truth. The selection is made impartially. Discipline is preserved.
Taking from different sections
The sample is taken from the beginning, the middle and the end of the text. This reveals the effect of fatigue. Spread is ensured.
The importance of the last section
Finding density may rise towards the end of the text. This section is always covered. The rule is applied.
Including risky sections
Numeric fields and technical sections are added to the sample. The risk is there. The scope is widened.
Sample size
The amount to be examined is defined in advance. A very small sample is not enough for a judgement. A measure is set.
Relationship with volume
In large jobs a tiered size may be used instead of a proportional one. The method is kept practical. The rule is written down.
Generalising the result
The sample result is scaled to the whole. The method of generalisation is defined. The calculation is written down.
When a critical finding appears
If a critical finding appears in the sample, the examination is extended. The result is not generalised directly. The rule is separate.
Extending the examination
If high density is found, the sample is enlarged. If necessary, a full examination is carried out. The steps are gradual.
Recording the result
Which section was examined is recorded. The record can be verified later. Transparency is ensured.
Sharing with the client
The method and the sample are explained to the client. A hidden method does not build trust. Openness is preserved.
Qualifications of the evaluator
The person examining the sample has the same language and subject competence. A less competent person cannot evaluate. The condition is firm.
Application in recurring jobs
In regularly flowing jobs sampling can be done periodically. Not every job is measured separately. The load is balanced.
Honest use of the method
Sampling is used not to hide the fact that no full examination was made, but to reach a reliable judgement with limited resources; when the impartiality of selection is lost, the method stops being measurement and becomes a defence. That is why the selection rule is written down. Discipline is preserved.
Limits of measurement
Sampling does not show that there are no findings in the sections not examined. It gives only an estimate of the general level. The limit is stated openly.