How to Set Pass Marks and Scoring Criteria for Pre-Employment Assessments

An employer gives 40 questions to 120 job applicants and announces that anyone scoring below 70% will be rejected. Why 70%? If the answer is simply “that is our usual pass mark,” the employer may be using a familiar number rather than a defensible hiring standard.

A pass mark, also called a cut score or cutoff, can have serious consequences. It may determine who advances to an interview, who receives an offer or who is excluded from consideration. The mark should therefore be linked to the knowledge, skills and behaviours required for the job, the quality of the assessment and the intended decision.

This guide explains how employers can develop scoring rubrics, set justified minimum standards, combine assessment results and interpret borderline scores. It includes numerical examples, standard-setting approaches and a practical governance checklist.

The examples are instructional, not validated selection standards. Employers should obtain appropriate psychometric, legal and professional advice when tests have high-stakes consequences, and comply with applicable employment, equality, disability and privacy requirements.

Start by distinguishing four different numbers

Hiring teams often use the words score, pass mark, benchmark and ranking as if they mean the same thing. They do not.

A candidate who answers 75% of questions correctly is not necessarily at the 75th percentile. Nor does a score of 75% mean the candidate has a 75% probability of performing well at work.

A cutoff is a policy and measurement decision. A percentile is a description of relative standing. Both require context.

Step 1: Define what passing is supposed to mean

Before selecting a number, write a clear statement of the decision.

For example:

A candidate who meets the minimum standard should be able to complete routine bank reconciliations, recognise common discrepancies and document unresolved items without material errors.

This is more useful than “the candidate must be above average.”

Ask whether the test is used to:

• Establish minimum job readiness.
• Identify applicants who merit further assessment.
• Compare qualified finalists.
• Detect a specific critical deficiency.
• Support development planning after hiring.

These purposes may call for different decision rules. A threshold for progressing to an interview need not be the same as a minimum standard for a safety-critical task.

Step 2: Map the score to essential competencies

A single total score can conceal important gaps. Before setting a cutoff, identify which competencies are essential and whether weakness in one can legitimately be offset by strength in another.

Consider a fictional accounts officer assessment:

If reconciliation accuracy is essential, an employer may need a separate minimum standard for that component rather than rely entirely on the overall total.

However, every component minimum must itself be justified. Too many arbitrary gates can exclude capable applicants without improving the decision.

Step 3: Use a defensible standard-setting method

There is no single standard-setting method suitable for every assessment. The choice depends on the test format, intended use and available evidence.

Modified Angoff method: judge item-level performance

In an Angoff-style process, trained subject-matter experts consider each question and estimate how likely a minimally competent candidate would be to answer it correctly.

For a fictional five-item test, an expert panel might reach these illustrative probabilities:

The sum is 3.5 expected correct answers out of five, equivalent to 70% of the available items. This is a starting standard-setting estimate, not automatic justification for a cutoff of four or five.

A real process needs a clear definition of the minimally competent candidate, trained judges, discussion of disagreements, documentation and appropriate review of consequences. The five-item illustration is far too short to support a consequential hiring cutoff by itself.

Bookmark-style approaches: judge an ordered item set

For certain item-response-theory-based assessments, experts may review items ordered by difficulty and identify a point representing the required performance standard.

This approach depends on suitable calibration and technical expertise. Employers should not imitate it by merely sorting questions by the proportion answered correctly.

Borderline or contrasting-groups approaches

Where independent, credible performance classifications exist, score distributions can be examined for people judged to be minimally competent or clearly competent.

These methods depend heavily on the quality of the external classifications and the relevance of the reference population. They should not be treated as proof that a cutoff will transfer unchanged to every new job.

Which method should an employer choose?

The strongest approach is not necessarily the most complicated. It is the one whose assumptions and evidence fit the decision.

Step 4: Design scoring rubrics before reviewing answers

For an open-ended work sample, a rubric translates a performance standard into observable evidence.

Suppose an applicant reviews a fictional reconciliation:

A rubric should explain partial credit, common errors and acceptable alternative methods.

Avoid grading on polish alone. A beautifully formatted reconciliation with an incorrect balance should not outrank an accurate, clearly explained result solely because of presentation.

Where feasible, have two reviewers independently score a sample of responses and reconcile meaningful differences. Maintain a record of any changes to the answer key.

Step 5: Distinguish compensatory and non-compensatory decisions

A compensatory model allows strength in one area to offset weakness in another. A weighted total is a common example.

A non-compensatory model requires a minimum standard in one or more essential areas, even when the overall score is high.

Consider two fictional candidates:

Candidate A: (92 × 0.50) + (65 × 0.30) + (70 × 0.20) = 79.5.

Candidate B: (70 × 0.50) + (85 × 0.30) + (80 × 0.20) = 76.5.

Candidate A has the higher weighted score. But if the role genuinely requires a justified minimum communication standard of 75, Candidate A would not meet that requirement. If a justified technical minimum were 75 instead, Candidate B would not meet it.

These hypothetical thresholds illustrate how decision rules affect outcomes. They are not recommendations for any actual job.

Governance rule: Define the essential standards and weighting before reviewing candidate results.

Step 6: Do not confuse assessment difficulty with seniority

A harder test is not automatically a better test for a senior position.

A junior analyst may need accurate calculations from well-defined data. A senior analyst may need to evaluate conflicting evidence, explain uncertainty and recommend action under incomplete information.

A senior-level test should reflect the decisions and accountability of the role, not merely include obscure formulas or faster arithmetic.

This distinction matters when an employer uses the same question bank across entry-level, professional and management vacancies.

Step 7: Handle borderline scores responsibly

Suppose the cut score is 30 out of 40, and a candidate scores 29. A strict rule may classify the candidate as below standard, but that one-point difference may be smaller than the uncertainty in the assessment.

Possible responses, where justified and applied consistently, include:

• Review the scoring and answer key for errors.
• Examine whether technical or accessibility issues affected administration.
• Use an equivalent additional work sample for a material competency.
• Apply a documented borderline review procedure.
• Consider complementary evidence when the decision framework permits it.

Do not silently lower the cutoff for a favoured applicant. Any review procedure should be established in advance and available consistently to comparable candidates.

Measurement uncertainty is not a free pass

Every assessment has limitations, but uncertainty does not mean employers must ignore genuine weaknesses. It means decisions should reflect the quality and precision of the evidence, especially near a boundary.

A displayed score of 74.8 should not be treated as meaningfully different from 75.0 merely because a dashboard reports decimal places.

Step 8: Check item quality before relying on pass marks

A cutoff is not credible if the questions or answer keys are defective.

Review:

Correctness: Are answer keys technically and factually accurate?

Ambiguity: Could a reasonable candidate interpret the question differently?

Difficulty: Does challenge arise from the intended competency rather than confusing wording?

Duplication: Do multiple items test essentially the same answer with superficial changes?

Answer-pattern bias: Are correct options consistently longer or more detailed?

Currency: Does the question rely on a law, standard or software version that may have changed?

Accessibility: Does presentation create barriers unrelated to the skill?

Exposure: Have items circulated so widely that familiarity may distort results?

If a flawed item is discovered after testing, establish a consistent documented correction process. Assess the effect on scores and decisions rather than quietly editing results.

Step 9: Review fairness and potential adverse impact

A cutoff can disproportionately exclude some groups even when the same numerical rule is applied to everyone. Employers should examine whether the threshold is job-related, necessary for the decision and supported by appropriate evidence.

Review whether the assessment contains unnecessary language complexity, culturally specific assumptions, accessibility barriers or speed demands unrelated to the role.

Where lawful and appropriate, analyse selection outcomes for relevant groups. Use caution with small samples and consult applicable local rules.

A group difference does not automatically establish unlawful discrimination; conversely, an absence of observed difference does not establish that the assessment is fair.

Step 10: Document the standard and monitor outcomes

Maintain a standard-setting record that includes:

1. The role and assessment version.
2. The intended decision and competency requirements.
3. The scoring rubric and essential minimums.
4. The method used to propose the cutoff.
5. The experts involved and their rationale.
6. Relevant reliability and validity evidence.
7. Fairness, accessibility and candidate-impact reviews.
8. Approval date and responsible decision-makers.
9. Borderline, retest and technical-incident procedures.
10. The schedule for reviewing the standard.

Revisit the threshold when job requirements, test content, candidate populations or delivery conditions change.

A complete worked example: customer service assessment

A fictional employer builds a 20-point exercise to evaluate how applicants handle customer issues.

The employer identifies correct application of policy as essential and defines behavioural anchors for each level.

Candidate X scores 5, 2, 5 and 5: 17/20.

Candidate Y scores 4, 5, 4 and 4: 17/20.

Both have the same total, but Candidate X's policy application is materially weaker. If the employer has established a justified minimum of four points on policy application, Candidate X would not meet that specific requirement despite a high overall score.

The exercise illustrates why competency-level evidence matters. It does not prove that four points is an appropriate cutoff for any real employer.

Common pass-mark mistakes

Using 70% by habit. Familiarity does not make a threshold valid.

Choosing the cutoff after seeing results. This creates opportunities for bias and inconsistent decisions.

Treating a percentile as percentage correct. They describe different things.

Averaging incomparable scores. Combining scales without justified interpretation can create misleading totals.

Ignoring critical competencies. A high total may conceal a serious deficiency.

Setting too many gates. Multiple unnecessary minimums can exclude candidates without clear benefit.

Assuming precision from decimals. Fine-grained displays do not eliminate measurement error.

Using a cutoff from another role. Job demands and candidate populations differ.

Failing to revisit the standard. Tests, roles and regulations evolve.

Using Talentrenza Assess responsibly

Employers can explore Talentrenza Assess for relevant assessment categories and employer or organisation workflows.

Before using an assessment score as a pass/fail decision, verify which current platform features support the intended scoring, pass-mark configuration, reporting and candidate management workflow.

The existence of a configurable pass mark does not establish that the selected threshold is scientifically justified. Employers remain responsible for the relevance of the test, the scoring policy, accessibility, fairness and final hiring decisions.

Employer pass-mark checklist

• Define the job-related meaning of passing.
• Identify essential competencies and acceptable trainable gaps.
• Confirm assessment content and answer-key accuracy.
• Choose a suitable standard-setting approach.
• Document expert judgement and scoring anchors.
• Set weights and component minimums before testing.
• Avoid unsupported universal pass marks.
• Establish borderline and technical-incident procedures.
• Review fairness and accessibility.
• Train assessors to interpret scores correctly.
• Record approval and version information.
• Monitor outcomes and revisit standards when necessary.

Frequently asked questions

What is a good pass mark for a pre-employment assessment?

There is no universal percentage. The threshold should be based on the skills required for the role, the assessment's quality and a defensible standard-setting process.

Is 70% an acceptable recruitment test pass mark?

It may be appropriate in a particular assessment if supported by evidence, but 70% is not automatically justified. Employers should not choose it merely because it is a familiar number.

What is the difference between a pass mark and a percentile?

A pass mark is a decision threshold. A percentile describes a candidate's standing relative to a defined reference group. A 70th-percentile result does not mean 70% of items were answered correctly.

Should every competency have a minimum score?

No. Use separate minimums only where they are justified by essential job requirements. Unnecessary gates can reduce fairness and flexibility.

Can an employer change a pass mark after candidates take a test?

Changing thresholds after seeing results creates significant governance risks. Any exceptional correction should follow a documented, consistent process rather than favour particular candidates.

How should employers handle candidates just below the cutoff?

Apply a pre-established borderline review policy, check for scoring or administration errors and consider appropriate additional evidence where the decision framework permits.

Do weighted scores guarantee better hiring decisions?

No. Weights must reflect job requirements and assessment quality. An arbitrary weighted formula can be less useful than a clear competency profile.

Does a high test score guarantee job performance?

No. Test results provide limited evidence about assessed competencies. Future performance also depends on experience, support, context and capabilities not captured by the test.

Conclusion

A pass mark should represent a meaningful standard, not a convenient round number. Employers should define the job-related meaning of passing, build accurate scoring rubrics, choose an appropriate standard-setting method and review how thresholds affect real applicants.

Careful scoring can make hiring decisions clearer and more consistent, but numerical precision must not replace professional judgement or fair process.

Employers exploring structured assessment delivery can review Talentrenza Assess and verify how its current assessment and reporting features support their selection workflow.

Further reading

For the broader assessment framework, see Pre Employment Assessment Tests Guide.