Sampling vs Full Review

1. Introduction

Sampling and full review are two methods used by organisations, auditors, investigators, employers, lawyers, and courts to examine documents, records, transactions, employee data, or other large datasets.

• Sampling means examining a carefully selected portion of the available material and using the results to assess the larger population.
• Full review means examining every relevant item in the population.

The choice between the two depends upon the purpose of the review, the volume of material, the risks involved, the applicable law, and whether examining only a sample could overlook an important issue.

In employment and legal contexts, sampling may be useful for routine compliance audits, payroll checks, or large-scale document review. However, where an individual employee's rights, alleged misconduct, discrimination, fraud, or a specific disputed transaction is involved, a fuller investigation may be necessary.

2. Meaning of Sampling

Sampling is a method under which a reviewer examines a representative or risk-based selection of records rather than every record.

For example, an employer has 20,000 payroll records. Instead of checking all 20,000 records, it may select 1,000 records based on a statistically appropriate or risk-based methodology.

Sampling may be:

1. Random sampling – records are selected randomly.
2. Stratified sampling – records are divided into groups and samples are taken from each group.
3. Systematic sampling – every nth record is selected.
4. Risk-based sampling – records presenting greater risks receive greater attention.
5. Judgmental sampling – reviewers select records based on professional judgment.

3. Meaning of Full Review

A full review, sometimes called 100% review, involves examining every item within the defined scope.

For example, if an employee's disciplinary case concerns 50 specific emails, the investigator may review all 50 rather than selecting only 10.

Full review is generally more appropriate where:

• the dataset is small;
• the consequences of missing one item are serious;
• allegations concern a particular individual;
• fraud or misconduct is suspected;
• discrimination or retaliation is alleged;
• the relevant evidence is easily identifiable;
• legislation, regulations, court orders, or internal policies require comprehensive examination.

4. Sampling vs Full Review

5. Legal Principles Relevant to Sampling and Full Review

There is generally no universal rule requiring every employment or legal investigation to examine every document. Courts commonly focus on whether the procedure was fair, reasonable, relevant, and adequate in the circumstances.

The appropriate standard can therefore depend on:

• the nature of the allegation;
• the material being reviewed;
• the importance of the evidence;
• the opportunity given to the affected person to respond;
• the reliability of the investigation;
• whether relevant evidence was ignored;
• whether the investigation was biased or predetermined.

6. Important Case Laws

1. State of Uttar Pradesh v. Saroj Kumar Sinha (2010)

The Supreme Court of India emphasised the importance of a fair disciplinary inquiry and the requirement that the inquiry officer act fairly rather than mechanically.

Relevance:
Where an employment investigation relies on a limited sample of documents, the employer should ensure that the material selected is sufficient to fairly determine the allegations. Sampling should not become a method of avoiding relevant evidence.

2. Roop Singh Negi v. Punjab National Bank (2009)

The Supreme Court stressed that disciplinary proceedings are quasi-judicial in nature and that findings must be based on material and evidence rather than mere allegations.

Relevance to sampling:
A sample cannot automatically establish misconduct merely because several sampled records show irregularities. The evidence must reasonably support the particular allegation against the employee.

3. Union of India v. Prakash Kumar Tandon (2009)

The Supreme Court considered principles concerning disciplinary proceedings and the importance of a proper evidentiary foundation.

Relevance:
Where a review is conducted through sampling, the methodology and evidentiary basis should be capable of supporting the conclusions drawn from the reviewed material.

4. ECIL v. B. Karunakar (1993)

The Constitution Bench of the Supreme Court dealt with procedural fairness in disciplinary proceedings and recognised the importance of giving the employee an opportunity to respond to adverse material.

Relevance:
Whether evidence was discovered through sampling or full review, material relied upon against an employee may need to be disclosed and an appropriate opportunity to respond must be provided.

5. Managing Director, ECIL v. B. Karunakar (1993)

The decision is particularly significant for the principle that procedural safeguards cannot be disregarded merely because an employer considers the evidence sufficient.

Relevance:
An employer cannot rely on the existence of a review methodology alone. If the sampled evidence is used to impose disciplinary consequences, procedural fairness remains important.

6. State Bank of India v. Ramesh Dinkar Punde (2006)

The Supreme Court considered the evidentiary requirements applicable to departmental proceedings and the role of evidence in establishing misconduct.

Relevance:
Sampling may assist an employer in identifying patterns or irregularities, but conclusions concerning an individual employee should be supported by relevant evidence.

7. Karnataka Public Service Commission v. B.M. Vijaya Shankar (1992)

The Supreme Court recognised the importance of fairness and proper procedure in administrative decision-making.

Relevance:
A review methodology should be applied consistently and fairly rather than selectively targeting particular employees or records.

8. V.C., Banaras Hindu University v. Shrikant (2006)

The Supreme Court examined principles of natural justice and fairness in disciplinary/administrative decision-making.

Relevance:
If an employer's sampling methodology excludes material that could materially affect an employee's defence, the adequacy and fairness of the process may become an issue.

7. When Sampling Is Appropriate

Sampling can be appropriate where:

A. The dataset is extremely large

For example, reviewing millions of routine payroll or attendance records may make a 100% review impractical.

B. The objective is compliance testing

An organisation may sample records to determine whether its general payroll or HR procedures are functioning correctly.

C. The population is reasonably homogeneous

If records have similar characteristics, a properly designed sample can provide useful information about the wider population.

D. The consequences of an error are relatively limited

Routine administrative verification may reasonably rely on sampling.

E. Statistical reliability is important

A statistically designed sample can provide a measurable level of confidence about the larger population.

8. When Full Review Is Preferable

Full review is generally more appropriate when:

A. A specific employee is accused of serious misconduct

If an employee's employment may be terminated, reviewing all reasonably relevant evidence may be more appropriate than relying solely on a sample.

B. Fraud is suspected

A sample may identify suspicious activity but may fail to reveal the complete extent or pattern of the alleged fraud.

C. Discrimination is alleged

If an employee alleges discriminatory treatment, relevant comparator records may need to be examined comprehensively rather than through an unexplained sample.

D. The dataset is small

If there are only 50 relevant documents, reviewing all of them may be more efficient and defensible than creating a sampling methodology.

E. A particular document may be decisive

Where even one omitted document could materially alter the outcome, a full review may be necessary.

9. Sampling in E-Discovery

Sampling is particularly important in electronic discovery because organisations may possess millions of emails, documents, chat messages, and other electronic records.

A review team may use:

• keyword searches;
• technology-assisted review;
• predictive coding;
• random sampling;
• statistical sampling;
• relevance sampling;
• quality-control sampling.

The purpose is generally to reduce the volume of material requiring manual review while maintaining reasonable confidence that relevant material is not systematically missed.

However, sampling methodology should be documented carefully.

10. Sampling and Employment Investigations

Employers frequently possess large quantities of:

• attendance records;
• payroll records;
• expense claims;
• performance records;
• emails;
• access logs;
• productivity records;
• workplace communications.

A sample may be appropriate for a general compliance audit.

However, if the review changes into an investigation of a particular employee, the employer should reconsider whether the original sample remains adequate.

For example:

An employer samples 500 expense claims and finds 20 irregularities. This may justify further investigation. It does not necessarily establish that a particular employee committed misconduct merely because some sampled records were irregular.

11. Risk of Sampling Bias

Poorly designed sampling can create serious problems.

For example, suppose an employer examines only records from:

• high-performing employees;
• one department;
• one geographical location;
• one particular month.

The resulting conclusions may not accurately represent the entire workforce.

Therefore, sampling should consider:

1. Population definition
2. Sample size
3. Selection method
4. Representativeness
5. Risk level
6. Error rate
7. Confidence level
8. Documentation of methodology

12. Sampling Does Not Mean Ignoring Evidence

An important distinction is between:

Sampling for general assessment and sampling the evidence used against an individual.

Sampling may be entirely reasonable for discovering whether a general problem exists.

But once specific evidence identifies a serious allegation, investigators may need to expand the review.

A sensible approach is often:

Initial sample → identification of risk → expanded review → targeted/full investigation.

13. Documentation of Sampling

An organisation should document:

• why sampling was chosen;
• what population was reviewed;
• how the sample was selected;
• the sample size;
• the methodology used;
• the results;
• exceptions discovered;
• limitations of the sample;
• whether further review was undertaken.

This documentation can be important if the review is later challenged.

14. Sampling and Natural Justice

Natural justice generally requires fair decision-making.

Two important principles are:

1. Nemo judex in causa sua – no person should be a judge in their own cause.
2. Audi alteram partem – the affected person should have a fair opportunity to be heard.

Sampling does not remove these principles.

If an employer relies upon sampled records to take disciplinary action, the employee should ordinarily have an appropriate opportunity to challenge the relevant evidence and explain the circumstances.

15. Sampling and Data Protection

Large-scale employee-data reviews can also raise privacy and data-protection concerns.

A review should generally consider:

• necessity;
• proportionality;
• purpose limitation;
• data minimisation;
• confidentiality;
• access controls;
• retention;
• security.

Sampling may sometimes reduce unnecessary access to employee information because fewer records need to be examined manually. However, it does not automatically make a processing activity lawful.

16. Advantages of Sampling

Advantages

• Lower cost
• Faster review
• Useful for very large datasets
• Reduces reviewer workload
• Can identify patterns
• Useful for compliance testing
• Can provide statistically meaningful results when properly designed

Disadvantages

• Relevant evidence may be missed
• Poor sampling can produce misleading conclusions
• Sampling bias may occur
• Serious individual misconduct may require additional investigation
• The methodology may be challenged
• Rare events may not appear in the sample

17. Advantages of Full Review

Advantages

• Greater evidentiary completeness
• Lower risk of missing an important record
• Appropriate for smaller datasets
• Useful in serious investigations
• Easier to explain where every relevant item was examined

Disadvantages

• Expensive
• Time-consuming
• Requires more investigators/reviewers
• Can create information overload
• May be disproportionate for routine compliance exercises

18. Practical Decision Framework

An organisation can consider the following questions:

Step 1: How large is the dataset?

Step 2: What is the purpose of the review?

Step 3: What happens if relevant evidence is missed?

Step 4: Does the review concern an individual employee?

Step 5: Is misconduct, fraud, discrimination, or retaliation alleged?

Step 6: Is there a statutory or procedural requirement for comprehensive examination?

Step 7: Can a statistically valid sample answer the question?

Step 8: Does the initial sample reveal issues requiring an expanded review?

If the consequences of missing evidence are significant, the organisation should consider moving from sampling to a more comprehensive review.

19. Conclusion

Sampling and full review serve different purposes. Sampling is primarily a method of efficiently assessing a large population, while full review provides comprehensive examination of the defined material.

There is no universal rule that sampling is always sufficient or that full review is always required. The appropriate method depends on the size of the dataset, purpose of the review, seriousness of the allegations, evidentiary risks, applicable legal requirements, and procedural fairness.

In employment investigations, sampling may be useful for identifying patterns and compliance problems, but it should not automatically substitute for examination of relevant evidence concerning an individual employee. Where serious consequences are possible, a risk-based expansion from sampling to fuller review can help ensure that the decision is based on adequate and reliable material.