The SHL Cognitive Test in the Commission's Internal Competitions: A Complete Guide

If you are preparing an internal competition at the European Commission, the single most important thing to understand in 2026 is this: the first hurdle is no longer the test you remember. It is an adaptive, interactive cognitive assessment run on the SHL platform, and it works nothing like the classic EPSO multiple-choice reasoning tests. This guide walks through what it is, exactly how it behaves, how it is scored, and how to get ready — all from verified sources and, where noted, our own study of the official practice assessments.

What changed, and who it affects

Three internal competitions moved to the new model in 2026:

• COM/AD/107/2026 — Administrators AD7 (generalist, also open to temporary agents): published on 31 March; its candidates already sat the new cognitive assessment in June.
• COM/AD5/108/2026 — Administrators AD5: applications closed on 30 June; the cognitive assessment is expected around September 2026, with the EU Knowledge MCQ and written test following in October.
• COM/AST/109/2026 — Assistants AST3: the newest of the three, with applications open until 31 July 2026 and tests following shortly after.

In all three, the structure is the same and the order is deliberate: the cognitive assessment comes first, as an eligibility gate. Only candidates who clear it continue to the stages you may know from previous internal competitions — the EU Knowledge MCQ (30 questions, 30 minutes), a written assignment, and the oral. The knowledge and written stages are still there; it is the entrance door that changed completely.

The platform: adaptive, and scored on a curve

The assessment is delivered on SHL's online platform by an external company specialising in psychometric testing. Two properties make it feel very different from a traditional exam.

First, it is adaptive. The difficulty of each question responds to how you are doing: answer well and the questions get harder. Your result is not a raw number of correct answers — it is a percentile, a measure of how you performed relative to a benchmark population of other test-takers. The Commission applies an eligibility line at the 60th percentile: in plain terms, you need to perform better than roughly the bottom 40% of the comparison group to stay in the competition.

Second, it is built around a single 36-minute sitting (in the combined "general ability" form) that mixes three reasoning types — numerical, inductive and deductive — into one continuous, adaptive flow of around 24 questions. The stand-alone sections are shorter and focused: the deductive reasoning section, for example, is roughly 12 questions in 18 minutes.

Inside the assessment: interactive, not multiple choice

This is the part almost no guide describes properly, so it is worth being precise. The SHL assessment is activity-based. Instead of presenting a set of options A to E for you to choose from, it asks you to construct the answer by interacting with the screen. Across the three reasoning types you will:

Numerical reasoning — work with realistic workplace data (allowance rates, project revenues, headcounts, vehicle mixes) and turn it into an action: rank people by dragging numbered badges onto them, approve or reject each tab against a threshold, drag the handles of a pie chart until the percentages match the facts, set the totals and splits of a bar chart, or plot a monthly series point by point.

Inductive reasoning — the close cousin of abstract reasoning, with a twist: you build the answer. Four figures are shown; you construct the fifth by tapping regions of a shape until the pattern is complete. Or you extend a sequence of labelled "bubbles" by literally drawing the connecting line to the next one, once you have worked out the hidden rule.

Deductive reasoning — the section that replaces classic verbal reasoning, and the most visual of all. You are given rules and facts and must make everything fit. In practice, this means tasks such as: rescheduling a meeting across colleagues' calendars where some blocks cannot move; placing people into offices on a floor plan while keeping departments apart and respecting who can sit next to what; working out a month-long duty rota where each person covers several consecutive working days; scheduling a job on a machine or room timeline without clashing with others; or ranking colleagues from a set of positional clues by dragging badges. You solve these by painting time slots, dragging people, selecting dates and placing badges — not by picking a lettered option.

How it is scored, and the rules that trip people up

A few mechanics are worth committing to memory before test day, because discovering them live is expensive:

• Every question is mandatory. You must provide an answer to move forward; the platform will not let you advance an empty question.
• There is no going back. Once you move to the next question, the previous one is locked. There is no review screen at the end.
• No negative marking, but unanswered or unfinished items lower your score — which, combined with the mandatory-answer rule, means pacing is everything.
• The score is a percentile, so you are effectively compared against other candidates, most of whom will meet this format for the first time on the day.

Why the unprepared struggle

From the official practice materials and the published notices, a handful of patterns stand out — the kind that punish a first-timer:

• Rules that bind per line, not per total. An expense-approval task can reject a claim whose total fits comfortably, because one single line exceeds its own rate. The wording is deliberate; read it twice.
• Deliberate boundary cases. Figures set at 2.02× a threshold are treated differently from 1.98×. The items are built around the edge.
• The obvious answer is trapped. In a calendar task, the tempting early gaps are often the one thing you cannot move — an immovable client meeting.
• Directions flip. One ranking asks for 1 = highest; the next, 1 = lowest. The instruction tells you; momentum betrays you.

None of this is hard once you have seen it. All of it is costly to meet for the first time with a timer running and no way back.

The opportunity — and how to train for it

A percentile system has a property worth understanding: preparation compounds. Because you are compared against a population that, for the most part, will discover the format on test day, every bit of familiarity moves you up the distribution. Understanding the mechanics, recognising each item type on sight, and having already burned the classic traps in practice — each of these shifts you higher, and the 60th-percentile line stops being a wall.

That is exactly what we are building. EU·Now is preparing a training platform that recreates the format and mechanics of all three SHL-style sections — deductive, inductive and numerical reasoning — with the verified format, every documented item type, the traps and the method, in English, French, German, Spanish and Italian. It is coming very soon to eu-now.com. Create your EU·Now account today to be among the first to train on it — because with percentile scoring, being early is an advantage that pays off twice.

EU·Now is not affiliated with, or endorsed by, SHL Group Ltd or the European Commission. All practice items on our platform are original, created in the style of the assessment; we never reproduce SHL content. Format details are verified against the official notices of competition, SHL's public product documentation and the official practice assessments — sources below.