Numerical reasoning
12 questions · 90 seconds for the first question on a new data set, then 45 seconds for each subsequent question on that set
What it tests. Advanced data interpretation, currency conversions, multi-step percentage changes, CAGR, and isolating variables within large, distracting data sets.
Worked example. A table shows quarterly revenues, operating expenses and headcount for five regional software offices over three years. A question might ask: if the UK office's operating expenses rose 8% while revenue per employee fell 4.5% versus the prior year, what was the absolute variance in net profit margin between the UK and German offices in GBP? You must look up baselines, apply each percentage shift, adjust for exchange rates, compute margins as (revenue minus expenses) / revenue, and subtract.
Common traps. The distractor trap (irrelevant columns waste seconds) and the close-choice trap (options like 1.42m, 1.44m, 1.46m mean small rounding errors lead to a wrong option).
How to handle it. Spend the first 20 seconds of the 90-second window reading the table architecture before any maths. Keep a clean, labelled scratchpad so repeated baseline data does not need recalculating.
Verbal reasoning
15 questions · 60 seconds to read a new passage and answer the first question, then 45 seconds for subsequent questions on that text
What it tests. High-level comprehension, logical deduction, and distinguishing absolute facts from inferences and unproven assertions.
Worked example. A 250-word passage on the UK carbon border adjustment mechanism for steel producers. A statement asking whether a manufacturer can claim tariff exemptions when steel is routed through an intermediate hub in an exempt jurisdiction is answered only from the text; if the text does not support it, the answer is Cannot Say.
Common traps. External-knowledge injection (using facts from outside the passage) and modifier misinterpretation (missing qualifiers like "primarily", "solely", "frequently" or "contingent upon").
How to handle it. Read the question prompt before the passage to target your scan. Treat Cannot Say as highly viable; do not force True or False without explicit logical connective tissue.
Logical / inductive / deductive reasoning
12 questions · 45 seconds per question
What it tests. Abstract spatial reasoning, pattern recognition, simultaneous tracking of multiple rules, and non-verbal problem-solving under time pressure.
Worked example. A grid where shading alternates between solid, grey and cross-hatched, the number of vertices increases by one across each row, and the shape rotates 45 degrees clockwise down each column. You must satisfy all rules at once.
Common traps. The single-rule trap (spotting one pattern and ignoring a second governing colour or size) and fixation (staring at the whole matrix instead of deconstructing it).
How to handle it. Deconstruct systematically: isolate one variable (e.g. an inner line), trace it horizontally and vertically, eliminate non-matching answers, then move to the next variable. If stuck within 25 seconds, eliminate and make an educated guess; a timed-out answer is a complete failure for that adaptive thread.
Situational judgement (occasional)
15-20 scenarios · Scenario-based
What it tests. Alignment with Roland Berger's core values, professional ethics, teamwork, client management and proactive problem-solving.
Worked example. Running a final validation check late at night, you find a structural flaw in baseline data that invalidates two weeks of analysis ahead of a steering-committee presentation, and your stream lead is offline travelling.
Common traps. The hero complex (rewriting the model alone and presenting directly to the client, breaking hierarchy and QA) and passing the buck (doing nothing until the lead returns).
How to handle it. Balance analytical integrity (the error must be fixed) with structured communication (inform your manager immediately with a proposed solution). Always protect the client relationship by managing mistakes internally first.
Personality questionnaire (Korn Ferry Dimensions)
Multiple forced-choice blocks · Untimed but best completed quickly
What it tests. Behavioural traits, workplace preferences, resilience and cultural fit for high-intensity strategy consulting.
Worked example. Choosing, across four professional-style statements, which is most and least like you, designed to detect contradictory profiles.
Common traps. Gaming the profile as an aggressive independent leader (the forced-choice mechanism flags low consistency) and extreme neutrality (an indistinct profile lacking sharp consulting traits).
How to handle it. Be honest but keep the successful-consultant traits in mind: entrepreneurial spirit, analytical curiosity, intellectual humility and strong collaboration. Work quickly; natural behaviour is the most consistent.