Study guide

1.1 CSNT Test Architecture, Adaptive Engine & Percentile Scoring

Civil Service Numerical Test CSNT

Track your progress as you read

Finished sections get a check mark.

1 Chapter 1: Test Structure, Platform & Scoring Strategy

1.1 CSNT Test Architecture, Adaptive Engine & Percentile Scoring 1.2 Core Numerical Strategy & Mental Estimation

2 Chapter 2: Data Interpretation — Tables, Charts & Visual Data

2.1 Reading & Interpreting Complex Tables 2.2 Bar Charts, Grouped Bars & Component/Stacked Graphs 2.3 Line Graphs, Time-Series & Trend Analysis 2.4 Pie Charts & Proportional Representation

3 Chapter 3: Advanced Data Interpretation & Variance Analysis

3.1 Multi-Source & Composite Data Synthesis 3.2 Budget Variance, Performance Metrics & Dashboards 3.3 Index Numbers & Relative Performance Measures

4 Chapter 4: Percentages, Percentage Changes & Financial Applications

4.1 Fundamentals of Percentages & Fractional Equivalents 4.2 Percentage Increase, Decrease & Successive Changes 4.3 Percentage Points vs. Percentage Change 4.4 Reverse Percentages, VAT & Financial Calculations

5 Chapter 5: Ratios, Proportions & Scaling

5.1 Ratio Simplification & Equivalent Ratios 5.2 Sharing Amounts in Ratios & Allocation Problems 5.3 Direct & Inverse Proportion and Scale Conversions

6 Chapter 6: Averages, Rates & Productivity

6.1 Measures of Central Tendency & Spread 6.2 Weighted Averages & Composite Group Statistics 6.3 Speed, Distance, Time & Travel Rates 6.4 Work Rates & Operational Productivity

7 Chapter 7: Sequences, Currency & Practical Arithmetic

7.1 Number Sequences & Pattern Recognition 7.2 Currency Conversions & Exchange Rates 7.3 Word Problems, Linear Equations & Public-Sector Constraints

1.1 CSNT Test Architecture, Adaptive Engine & Percentile Scoring

Key Takeaways Explain This Topic AI

The Numerical Test assesses calculation, evaluation and interpretation of numerical information within the Ability element of Success Profiles.

The test is untimed; time taken does not affect the score, but the application submission deadline still applies.

Question difficulty adapts to performance and test length can vary; avoid invented fixed counts, early-error penalties or percentile caps.

Percentiles compare performance with a representative group and are not the percentage of responses correct.

Only a minimum-standard passing score is banked for the same grade grouping; it expires after six months or earlier on a new benchmark or affected-platform work.

Last updated: October 2026

What the Civil Service Numerical Test assesses

The Civil Service Numerical Test (CSNT) assesses the ability to calculate, evaluate and interpret numerical information. It is used within the Ability element of Success Profiles, alongside whichever other selection methods are relevant to a vacancy. The other Success Profiles elements are Behaviours, Strengths, Experience and Technical skills; a numerical-test result does not by itself establish performance in all five.

Questions present graphs, tables or other numerical information, followed by a question and several possible answers. The useful preparation skill is translating the stated information into a justified calculation or comparison. This guide's worked examples are independent practice, rather than official live test content or a guarantee of the topics in any particular assessment.

The primary candidate-policy source is GOV.UK's preparation guidance for the Verbal and Numerical tests . Check your invitation as well, because the vacancy's requirements and submission deadline still apply.

Untimed does not mean without a deadline

The Numerical Test is untimed , and the time taken does not affect the final score. GOV.UK reports that most people take 15–45 minutes and recommends allowing at least an hour . These figures describe typical completion and planning advice; 45 minutes is not an enforced test cutoff, and an hour is not a mandatory minimum sitting time.

The application deadline is separate. If it passes while you are still taking a test, the test itself is not stopped, but your application has missed the overall submission deadline. Start early enough to resolve problems. GOV.UK asks that queries or help requests be submitted at least two working days before the deadline to allow a response.

Practical implications include:

Read the instructions and complete the examples before relying on a familiar practice routine.

Answer carefully before confirming: you cannot skip questions or go back .

If interrupted, you can reopen the Numerical Test and continue where you left off. The candidate adjustment guidance says it does not time out.

If connectivity is lost, restore the connection and reopen the test through the application centre. A restart should not be assumed to provide a new attempt or a clean response history.

The tests can be taken in English or Welsh, selected at the start. Choose the language in which you can accurately interpret the instructions and data.

Adaptive question selection: what the sources establish

The test is computer adaptive . Candidate guidance says the next question may become harder after a correct answer or easier after an incorrect one. Test lengths can vary, and the Civil Service does not publish a fixed scored-question count in that guidance.

The Cabinet Office's December 2025 algorithmic-transparency record describes an item-response-theory approach to estimating ability and a stopping rule connected to measurement precision. This explains why the test should not be treated as a fixed booklet in which every candidate answers the same questions.

These descriptions do not provide a reliable candidate-facing formula for converting one early mistake into a specified percentile loss, a fixed number of recovery questions or a maximum attainable percentile. Likewise, do not treat a difficult-looking question as a known bonus worth a fixed multiple of an easy item. Question appearance is not an official score report.

  • Published feature Sensible preparation consequence
  • Difficulty adapts to performance Practise reasoning across different levels of difficulty
  • Time taken does not affect the score Check the data and arithmetic rather than racing an invented item clock
  • The test can vary in length Do not plan around an unsupported question total
  • Questions cannot be revisited Verify the selected answer before confirming it
  • Some trial items are unscored Answer every presented question carefully instead of trying to identify trials

GOV.UK confirms that trial items help calibrate future questions and do not form part of your assessment. Their number is not supplied as a preparation target. A question that feels unexpectedly easy, difficult or different is not a dependable signal that it is a trial item.

Percentiles are comparisons, not percentages correct

The number and difficulty of questions are used to produce a score, which is compared with a representative group of applicants who have taken the test. Your result is presented as a percentile . For example, the official guidance explains that a percentile of 44 means you scored better than 44% of that comparison group.

A percentile of 65 therefore does not mean that 65% of your responses were correct. Nor should the comparison group be assumed to consist only of the people applying for the same current vacancy. Without the scoring and comparison data, a practice-set percentage cannot be converted directly into an official percentile.

For Verbal and Numerical testing, the published grade groupings are:

  • Grade grouping Scope of the grouping
  • AA / AO Administrative Assistant and Administrative Officer
  • EO Executive Officer
  • HEO / SEO Higher Executive Officer and Senior Executive Officer
  • Grade 7 / Grade 6 / SCS 1 These three levels share a test grouping

Where a department uses different grade labels, it selects an appropriate test level. Do not infer the level solely from a job title that resembles a role in another department.

Separate the minimum standard from vacancy progression

If you pass the minimum required standard for the job level, you receive a message. You may then be invited to further required tests, or the recruiter considers applicant scores after the deadline to decide the vacancy's pass mark. The recruiter may raise that mark, taking into account the number of invitations and the impact on protected groups.

Meeting the minimum standard is not a guarantee of an invitation to continue selection . GOV.UK does not provide one universal required percentile for each grade in this guidance. Avoid adopting a coaching site's fixed EO or HEO number as an official cutoff. Read the actual result and the vacancy's subsequent communications.

Banked scores and later applications

Only a score meeting the minimum required standard is banked . An applicable banked score is reused when you apply for another job in the same grade grouping , rather than inviting you to take that test again. A failed score is not banked; you can take the test again in a future application.

A banked score expires when any one of these events occurs:

Six months have passed since it was obtained.

A new score benchmark is applied to the test.

Technical work on the platform affects banked scores.

Thus, six months is an upper duration subject to the published earlier-expiry conditions, not an unconditional promise. An SEO banked pass can apply to an HEO application because they share a grouping. An AO pass is not automatically an EO pass. Banking a Numerical Test result also does not replace a required Verbal Test result.

A live applicable banked pass is reused instead of providing an ordinary opportunity to retest for a higher mark. Separately, the hiring manager may require a supervised retest; GOV.UK says candidates will be told whether this is expected if invited to interview. Your responses must be your own, so prepare to reproduce your reasoning without help from another person.

Equipment, practice and accessibility

GOV.UK suggests having rough paper, a pen and a calculator available. Use a reliable internet connection and follow the instructions for the assessment. The published suggestion is not permission to obtain answer assistance or use unspecified tools that conflict with those instructions.

Tests work on most modern browsers and operating systems. They should work on phones and tablets, but GOV.UK says larger-screen devices work better. Try the official Numerical practice test with your intended setup. It is unscored, deliberately easy and intended to show the look and feel of the test; its result is not a pass prediction. GOV.UK also links an official Numerical Test video .

The reasonable-adjustments guide , updated January 2026, says Numerical Test visual information is also available in table form for screen readers. It describes display adjustments and, when needed, an in-person reader or supervised paper alternative. It says the test is periodically audited against WCAG 2.1 AA; this is not a promise that every individual setup will work without adjustment.

Request an adjustment when applying, describing the difficulty and what may help. If the need becomes apparent later, contact the named person in the advert. Supporting information may be requested, but lack of existing documentation should not deter a request. An untimed Numerical Test does not need an extra-time allocation in the way a timed assessment section does; other access needs still matter.

Loading diagram... CSNT results, conditional banking and vacancy decisions