EPSO numerical reasoning test: format, pass score and practice
The EPSO numerical reasoning test measures your ability to read a table or a chart and pull the right figure out of it: 10 multiple-choice questions in 20 minutes.
Applications closed
- Application deadline
- Last verified
Notice EPSO/AD/427/26
Official test parameters
The data is given to you and no advanced mathematics is expected: the test is built on percentages, ratios, changes and averages, under tight time pressure.
- Questions
- 10
- Duration
- 20min
- Pass score
- 10/20 (combined)
- Weight
- -
- Scoring
- 1 pt/question
- Eliminatory
- Yes
- Ranked
- No
- Notice
- EPSO/AD/427/26
The pass score is not an individual one: this test is marked together with abstract reasoning, and the combined total for the two tests must reach 10/20.
Verbal reasoning
Numerical reasoning
Abstract reasoning
EU knowledge
Digital skills
EUFTE (written test)
Content based on the notice EPSO/AD/427/26 (C/2026/711). Official parameters last verified: .
Which competitions include this test
Numerical reasoning sits alongside verbal reasoning in almost every EPSO competition, from generalist administrators to assistants and specialists.
The table below sets out, competition by competition, the settings published in the Official Journal: number of questions, duration and, above all, how the pass score is calculated.
That last point is decisive: depending on the competition, numerical reasoning is marked on its own, together with abstract reasoning, or together with verbal and abstract reasoning combined. The three cases do not call for the same preparation strategy.
Administrateurs généralistes (AD 5)EPSO/AD/427/26
Assistants communication (AST 3)EPSO/AST/157/25
Administrateurs audit (AD 7)EPSO/AD/428/26
CAST Permanent (agents contractuels FG I à FG IV)CAST Permanent
Administrateurs fiscalité (AD 6)EPSO/AD/422/25
Administrateurs gestion et connaissance des données (AD 7)EPSO/AD/426/25
Juristes-linguistes CJUE (AD 7)EPSO/AD/424/25
Juristes-linguistes et réviseurs juridiques PE, Conseil, Commission (AD 7)EPSO/AD/423/25
Administrateurs TIC (AD 7)EPSO/AD/429/26
Traducteurs (AD 5)EPSO/AD/414/24 a EPSO/AD/421/24
The competition families
- AD 5
Generalist administrators
- AST 3
Communication assistants (AST 3)
- AD 7
Administrators in the field of audit (AD 7)
- CAST
Contract agents (permanent)
- AD 6
Administrators in taxation (AD 6)
- AD 7
Administrators in data and knowledge management (AD 7)
- AD 7
CJEU lawyer-linguists (AD 7)
- AD 7
Lawyer-linguists and legal revisers for the EP, Council and Commission (AD 7)
- AD 7
ICT administrators (AD 7)
- AD 5
Translators (AD 5)
What it really measures, and what it does not
What the test measures
- Accurate reading of a table: finding the right cell, at the right intersection of row and column, first time.
- The four operations that keep coming back: percentage, percentage change, ratio and average.
- Attention to units and orders of magnitude: thousands against millions, euros against percentages, absolute value against change.
- Time management: at 10 questions in 20 minutes, each question gets two minutes, calculation included.
What it does not measure
- Advanced mathematics: no algebra, no inferential statistics, no geometry.
- Your ability to calculate in your head: the format provides a calculation aid, so the difficulty lies elsewhere.
- Your knowledge of economics: the figures in the table are enough, and a hunch about the real world is no substitute for accurate reading.
- Your memory: there is nothing to learn by heart, everything is in front of you.
The format in detail: the five recurring calculations
The notice of competition publishes no sub-domains for numerical reasoning: it sets the number of questions, the duration and how the pass score works, not a taxonomy. What follows is therefore not an official list, but the typology we draw from our own question bank, which covers most of what candidates meet on the day.
Direct reading
The value asked for is in the table, at a precise intersection. The only difficulty is not picking the wrong row, the wrong column or the wrong unit.
Percentage of a total
What share of a whole a given value represents. The trap is the denominator: the column total, the row total and the grand total do not give the same answer.
Percentage change
The gap between two periods expressed against the starting value. Confusing a change in percentage points with a percentage change is the single most common mistake in the whole test.
Ratios and per capita values
Dividing one quantity by another to compare countries of different sizes. The word to spot is “per”: per inhabitant, per business, per year.
Averages and estimation
The average of a column, or an order of magnitude close enough to rule out three options. Rounding early often saves more time than calculating exactly.
Typology established by Erudiam from its own question bank. The official notice of competition publishes no sub-domains for this test: the only official figures on this page are those in the table and the banner, taken from the Official Journal.
Try questions from the test
The questions below come from our practice bank, in the same format as the test: a data table, a question, four options, with the worked answer and an analysis of every distractor.
Answer key : A
Take the "CO2 (t/inh)" column for the six countries: 5.2, 7.8, 4.1, 5.5, 6.7 and 4.3. A simple average weights each country equally, regardless of population. (5.2 + 7.8 + 4.1 + 5.5 + 6.7 + 4.3) / 6 = 33.6 / 6 = 5.6 t/inh, option A. Option B (6.7) divides by 5 instead of 6, and C (33.6) is the undivided sum.
A. (5.2+7.8+4.1+5.5+6.7+4.3)/6 = 33.6/6 = 5.6.
B. 6.7 = /5 instead of /6
C. 33.6 = sum without division
D. 5.5 = rough rounding
Answer key : C
Take the "2023" column for the Germany row (17.6) and the Estonia row (21.0). Germany’s ratio to Estonia’s is calculated as Germany divided by Estonia, in that order: 17.6 / 21.0 = 0.8381, so 83.8% rounded to one decimal. A ratio carries no unit: Germany’s pay gap represents 83.8% of Estonia’s. Reversing the order would give 21.0 / 17.6 = 1.19, or 119%, a value not among the options.
A. 82.0% under-counts.
B. 83.0% rounds down.
C. 17.6/21.0 = 83.81% → 83.8%.
D. 85.0% over-rounds.
E. 86.5% is too high.
Answer key : D
In the column "2021", read the row "Netherlands": 81.7. In the column "2024", the same row shows 83.8. A change in percentage points is a subtraction: 83.8 − 81.7 = 2.1, that is +2.1 pp, so the correct answer is D. The main trap is B (+1.8 pp), which you get by reading the column "2023" (83.5) instead of "2024", since 83.5 − 81.7 = 1.8.
A. +1.5 pp under-counts.
B. +1.8 pp reads 2023 instead of 2024: 83.5 − 81.7 = 1.8.
C. +1.9 pp under-counts.
D. 83.8 − 81.7 = 2.1 pp.
E. +2.5 pp over-rounds.
Answer key : E
On the Italy row, read the "Female" and "Male" columns: 38.4 and 34.7. The gap is measured in percentage points, by subtraction: 38.4 minus 34.7 = +3.7 pp. Do not use the "Total" column (36.5), which does not answer the question, and do not confuse this with relative change (38.4 / 34.7, about +11%), a very different ratio from the point gap asked here.
A. +1.0 under-counts.
B. +1.5 under-counts.
C. +2.5 under-counts.
D. +3.0 under-counts.
E. 38.4 − 34.7 = 3.7 pp.
Answer key : E
In the column "Bike-sharing fleet", read the row "Amsterdam": 12,000. In the column "Cycle paths (km)", read the row "Amsterdam": 767. A ratio is a division, in the stated order: 12,000 / 767 = 15.6453, which is 15.65 rounded to two decimals. No option shows 15.65, so the correct answer is E. Trap A (0.06) comes from the inverted division: 767 / 12,000.
A. 0.06 = swap (767/12,000)
B. 12.00 = bikes/1,000 wrong
C. 18.18 = Paris (20,000/1,100)
D. 19.51 = Copenhagen (8,000/410)
E. 12,000 / 767 = 15.6453 ≈ 15.65. None of the listed options matches 15.65, so the correct answer is E.
Answer key : D
Take the "Average car speed (km/h)" column for the six capitals: 14, 24, 20, 22, 23 and 21. A simple average weights each city equally, regardless of its size. (14 + 24 + 20 + 22 + 23 + 21) / 6 = 124 / 6 = 20.67, rounding to 20.7 km/h, option D. Option A (24.8 km/h) divides by 5 instead of 6, and C (20.6 km/h) rounds down incorrectly.
A. 24.8 = /5 instead of /6
B. 124.0 = sum without division
C. 20.6 = wrong rounding
D. (14 + 24 + 20 + 22 + 23 + 21) / 6 = 124 / 6 = 20.6666 ≈ 20.7 km/h.
Questions written by Erudiam from official EU sources; EPSO does not publish past papers.
Common mistakes and method
The costliest mistake is confusing percentage points with percentages. Going from 20% to 25% is a rise of five percentage points, but a 25% increase. Both answers are almost always among the options.
The second is calculating before reading the question in full. Many questions ask for a comparison or a ranking rather than a value, so the right option can be found without a complete calculation.
The third is units. A table in thousands and an option in millions look the same apart from a factor of a thousand: checking the column heading before setting up the calculation costs three seconds and saves the question.
The method that works
read the question, pick out the two or three cells that matter, estimate the order of magnitude, eliminate, and only calculate precisely when two options are left.
How Erudiam trains this test
- Every worked answer shows the full calculation, not just the right value: you see where the denominator comes from, and why each of the three distractors matches an identifiable method error.
- Spaced repetition (the FSRS algorithm) brings back the types of calculation you get wrong, rather than making you sit one full test after another.
- The readiness index (ERI) combines coverage, consistency and success rate to tell you whether you are ready, on this test as on the others in the competition.
- Real-conditions tests reproduce the time pressure: that is where the difference shows between being able to calculate and being able to calculate in two minutes.
Frequently asked questions about the numerical reasoning test
10 multiple-choice questions in 20 minutes in the competition for generalist administrators, according to the official notice of competition (reference at the foot of the page). The settings differ from one competition to another: the table on this page compares them.
There is no individual pass score for this test in this competition: it is marked together with abstract reasoning, and the combined total for the two tests must reach 10 points out of 20. A weak numerical score can therefore be offset by a strong abstract score, and the other way round.
No. The test covers percentages, ratios, changes and averages, all from data provided to you. What separates candidates is accurate reading of the table and time management, not their level in mathematics.
We make no claim here. The official settings we recorded set a language regime for European Union knowledge, digital skills and the written test, but not for numerical reasoning. Your individual invitation and the notice of competition concerned remain the only reference.
In almost every one, alongside verbal reasoning and abstract reasoning. The table on this page lists the competitions we track, with their exact settings and how their pass score is calculated.
By working one type of calculation at a time rather than sitting full tests back to back: identify the one that trips you up (change, denominator, unit), drill it on its own, then take a test under real conditions again to confirm that it holds under time pressure.
