1.1 CSNT Test Architecture, Adaptive Engine & Percentile Scoring

Section 1.1: CSNT Test Architecture, Adaptive Engine & Percentile Scoring

The Success Profiles Framework and the Role of the CSNT

The UK Civil Service recruits tens of thousands of individuals annually across central government departments, executive agencies, and arm's-length public bodies. To standardise merit-based recruitment and eliminate recruitment bias, the Civil Service uses the Success Profiles framework. This framework evaluates candidates across five distinct elements:

• Behaviors: The actions and activities that result in effective performance.
• Strengths: What candidates do regularly, do well, and are motivated by.
• Experience: The knowledge or mastery of an activity gained through involvement.
• Technical: Specific professional skills, qualifications, or accreditations.
• Ability: The aptitude or potential to perform key tasks effectively.

The Civil Service Numerical Test (CSNT) is the flagship psychometric instrument used to assess the Ability element for roles requiring quantitative competence. Rather than testing abstract academic mathematics, advanced calculus, or theoretical geometry, the CSNT measures practical, workplace-aligned numerical reasoning: the capacity to interpret statistical tables, extract insights from financial dashboards, calculate budgetary variances, evaluate public policy indicators, and draw logical deductions from complex operational data.

Administered online via the Civil Service Jobs platform (powered by a modern psychometric testing engine), the CSNT serves as an objective, initial sift. Whether applying for an operational delivery role in HM Revenue & Customs (HMRC), an analytical or policy advisory post in HM Treasury, or a casework management position in the Department for Work and Pensions (DWP), your CSNT score determines whether your application advances to qualitative assessment stages such as written statements, video interviews, or assessment centres.

---

The Untimed Testing Paradigm: Dispelling the Speed Myth

One of the most persistent misconceptions among candidates is that the CSNT is a high-speed sprint similar to traditional private-sector psychometric assessments (such as the strictly timed graduate-scheme aptitude tests many candidates have met before).

The Civil Service Numerical Test is officially untimed.

Understanding the operational implications of this design is critical:

• Zero Speed Penalties or Bonuses: The psychometric scoring algorithm evaluates what difficulty of questions you answer correctly, not how quickly you answer them. Spending 3 to 4 minutes working through a dense multi-stage data table does not deduct a fraction of a point from your score.
• Session Duration Realities: Most candidates complete the assessment within 15 to 45 minutes. However, the testing portal does not cut you off after 45 minutes. Candidates are strongly advised to block out 60 uninterrupted minutes in a quiet environment to eliminate time pressure and ensure every single calculation is verified.
• Session Recovery and State Persistence: The tests do not time out. If your browser crashes, your device loses power, or your internet connection drops, you can log back in through Civil Service Jobs and resume where you left off; you can relaunch the test as many times as you need, and it does not have to be on the same device.
• No Skipping and No Going Back: GOV.UK's test-taking tips are explicit — "Answer every question – you cannot skip or go back." Once you confirm an answer it is locked. This is precisely why the untimed format matters strategically: the effort you would spend revisiting an earlier item on a conventional test must instead go into getting each item right the first time.
• The Deadline Still Applies: The test is untimed, but the vacancy is not. If the application deadline passes while you are mid-test, "your test will not be stopped but your application will have missed the overall deadline for submission." Sit the test as soon as you are invited, and raise any query at least two working days before the deadline.
• English or Welsh: Both the Verbal and Numerical tests can be completed in English or Welsh — you select your preference at the start of the test.

---

The Computer Adaptive Testing (CAT) Engine

The CSNT does not present a fixed, static booklet of questions. Instead, it is governed by an advanced Computer Adaptive Testing (CAT) engine based on Item Response Theory (IRT).

How Item Difficulty Dynamically Adjusts

In a CAT environment, the platform dynamically matches question difficulty to the candidate's demonstrated ability level (represented mathematically as theta, $\theta$):

1. The Initial Anchor Item: The test begins by serving an item calibrated at moderate, median difficulty.
2. Dynamic Branching: If you answer the item correctly, the engine updates its internal estimation of your ability upward and selects a subsequent question from a pool of higher-difficulty items (incorporating more variables, multi-step data manipulation, or denser tables). If you answer incorrectly, the engine adjusts its ability estimate downward and serves a lower-difficulty item.
3. Information Maximisation and Convergence: With each response, the statistical confidence interval around your true ability narrows, and the assessment ends once the engine can place you reliably. The Civil Service does not publish a scored-item count for the CSNT. GOV.UK states only that the adaptive design "means that the tests can vary in length, but are typically shorter than fixed-length tests". Any source quoting a precise number of questions is guessing — do not build a pacing plan around one.

The Critical Leverage of Early Questions

Because the adaptive engine begins with a neutral ability assumption, early questions exert disproportionate leverage on your trajectory. A careless arithmetic blunder on questions 1 through 5 drops the algorithm into lower difficulty tiers. Once relegated to lower-difficulty items, a candidate must answer multiple consecutive questions correctly simply to steer the engine back toward the baseline difficulty required to reach competitive percentiles. Treating the first ten questions with extreme diligence is the single most effective tactical decision a candidate can make.

Unscored Calibration and Trial Items

Candidates will occasionally encounter items that appear noticeably distinct in presentation or surprisingly easy/difficult relative to recent questions. The test embeds an unpublished number of unscored trial items (also known as calibration items); GOV.UK confirms only that "you will be asked to answer some trial items during the test".

Civil Service psychometricians use these uncredited questions to gather live normative statistics (item facility, response distribution, and discrimination index) before approving them for inclusion as scored items in future testing cycles. These trial items do not affect your score in any way. However, because they are visually indistinguishable from live scored items, candidates must treat every single presented question with equal, rigorous focus.

---

Percentile Scoring, Grade Norm Groups & the Pass Mark

Unlike academic examinations or vocational qualifications that apply a fixed passing grade (such as 60% or 70%), the CSNT uses norm-referenced percentile scoring.

What Percentile Scores Measure

Your test output is reported as a percentile rank:

• A score of 65th percentile does not mean you answered 65% of the questions correctly.
• It means your estimated quantitative ability exceeded that of 65% of all candidates in the standardised comparison norm group.
• Percentile scores naturally control for test difficulty; if a particular test version features exceptionally challenging items, the norm curve adjusts accordingly.

Civil Service Grade Groupings and Norm Groups

Norm groups are stratified by Civil Service grade level. When you sit a test for an Executive Officer position, your performance is benchmarked against fellow EO applicants, not entry-level administrative assistants or senior directors. For Verbal and Numerical test purposes GOV.UK groups the grades as follows:

Where a government department does not use these grades, it selects an appropriate level of test for its vacancy.

How the Pass Mark Is Actually Set

[!WARNING] The Civil Service publishes no required percentile for any grade. There is no official "you need the 60th percentile for EO" figure, and no credible third-party source for one. Treat any coaching site quoting grade-by-grade cut-offs as invented, and do not calibrate your preparation against a number nobody publishes.

What GOV.UK does describe is a two-stage decision:

1. The minimum required standard for the job level. If you pass at that standard you receive a message telling you so. Where additional tests are needed, you are then invited to complete them.
2. The recruiter's pass mark, decided after the deadline. Once the test deadline passes, "the recruiter will look at all applicant scores to decide what the job's pass mark will be." They weigh the impact on protected groups and the number of invitations to the next stage. The recruiter may raise the pass mark — if so, you are informed, and told whether you passed or failed at the raised standard.
3. If you do not pass at the minimum required standard, you can retake the test during any future application.

• Key Takeaway: Meeting the minimum test requirement for a job level "is no guarantee of an invite to continue the selection process". Because the effective cut-off is fixed only after everyone has sat the test, the one rational strategy is to maximise your own percentile rather than aim at a target number.

---

Banked Scores Policy and Retake Regulations

The Civil Service maintains a strict score banking mechanism to ensure fair testing and prevent speculative retesting.

What Gets Banked, and For How Long

Banking applies only to a passing score:

• If you pass at or above the minimum required standard, your score is banked and automatically re-used for future jobs in the same grade grouping. Civil Service Jobs checks for an applicable banked score when you apply: if you have one you are not invited to the test; if you do not, you must sit it.
• If you do not pass, your score is not banked. You can take the test again for any future job you apply for. There is no lock-out and no penalty for a failed attempt.

A banked score expires when any one of the following occurs — the six-month clock is not the only trigger:

1. Six months have passed since the banked score was obtained.
2. A new score benchmark is applied to the test.
3. Technical work on the test platform takes place which affects banked scores.

In each of those cases you take the test again.

Retake and Cross-Grade Rules

• A banked score covers only its own grade grouping. Pass at SEO level and the banked score also applies to HEO-level jobs, because HEO and SEO share a grouping. Sit a test for an AO role and then apply for an EO role, and you will be invited to take the EO-level assessment, because the norm groups differ.
• A banked score does not transfer between tests. Banking a Numerical Test score does nothing for the Verbal Test, and vice versa.
• You cannot re-sit to improve a live banked pass. While a passing score is banked for your grade grouping, Civil Service Jobs re-uses it instead of inviting you back to the test.

Supervised Verification Retests

To protect test integrity, "the hiring manager may require candidates to retake the test under supervised conditions". GOV.UK states you will be told whether you are expected to do this if you are invited to interview. Your answers must be your own: "you must not ask for input from anyone." The Civil Service does not publish what follows if supervised and unsupervised results diverge, so treat the supervised sitting simply as a re-test you must be able to reproduce.

---

Equipment, Permitted Tools & Candidate Environment

Because the test is administered remotely in an unproctored environment, candidates have full control over their testing setup. Optimising this environment is a vital preparatory step.

Permitted Calculating Tools and Materials

1. A Calculator: GOV.UK's preparation guidance tells candidates that before starting they "may wish to have some rough paper, a pen and a calculator to hand". No calculator type is prescribed or prohibited, and the test is taken unproctored at a place of your choosing. Use whatever you are fastest and most accurate with; a familiar physical desktop or pocket calculator with responsive keys usually beats an unfamiliar on-screen one for multi-step computations.
2. Rough Scrap Paper and Pens: Keep several sheets of blank paper and multiple pens or fine-tip markers at your desk. Structured scratchpad technique (detailed in Section 1.2) is indispensable for tracking intermediate variables and avoiding memory overload.
3. Hardware and Browser Configuration: Access the test via an up-to-date desktop or laptop browser (Google Chrome, Microsoft Edge, Mozilla Firefox, or Apple Safari). Avoid taking the test on smartphones or small tablets; responsive mobile layouts can force wide statistical tables into horizontal scrolling panes, introducing friction and reading mistakes. Disable third-party browser extensions and notification pop-ups before launching the assessment.

---

Reasonable Adjustments & Accessibility Provisions

Under the Equality Act 2010, the Civil Service is committed to ensuring full accessibility for candidates with disabilities, neurodivergent conditions, or temporary impairments.

Platform Accessibility Standards

The CSNT testing interface complies with public sector accessibility regulations (WCAG 2.1 AA):

• Universal Data Tables: For every graphical data display—including bar charts, line graphs, and pie charts—the platform provides a fully formatted, accessible data table equivalent. Candidates who use screen-reading software or who experience difficulty parsing visual graphs can view the underlying numerical figures directly in tabular format.
• Custom Display Options: The portal supports native browser zoom, high-contrast colour palettes, and custom text sizing without breaking data alignment.

Formal Reasonable Adjustments

When you apply through Civil Service Jobs you are asked whether you need an adjustment to the online tests; this is the main route by which adjustments are identified. Supply the reason you need an adjustment, the adjustments that could help, and any you have had before. A recruiter then contacts you to confirm the arrangement, and may ask for supporting documentation where the need is not obvious. Published adjustments relevant to the Numerical Test include:

• A paper-based alternative: "Where an online test is not possible, there are paper-based alternatives to the Verbal Test and Numerical Test." These "must be taken in a supervised session, which the recruiter will organise", and can vary paper size and text/paper colour.
• An in-person reader: The Numerical Test contains charts and graphs, "but questions do not rely on users viewing visual content as this content will always be provided in a table format... However, if you still struggle with the test we can provide an in-person reader or a paper-based alternative."
• Display adjustments: Text and zoom size, background/text colour, and font type are adjustable through your browser and screen settings before you start; screen contrast can be modified or a glare filter used.
• No extra-time adjustment applies: GOV.UK's adjustments table lists the Numerical Test as having no time limit, so extra time is not a relevant adjustment — only the Customer Service and Casework tests contain a timed section.

[!TIP] Take the free official practice test first. GOV.UK publishes a Civil Service Numerical practice test — a shorter version that "looks exactly like the actual test" — plus a short Civil Service Numerical Test video. The practice test is not scored and its questions are deliberately easy; its purpose is to confirm the look, feel and accessibility of the real thing, especially if you use assistive technology. The real test also includes example questions at the beginning of each section.