Numerical reasoning (Scales Numerical)
37 questions · 12 minutes (about 19 seconds per question)
What it tests. Data location, rapid fractional arithmetic, compound percentages, currency conversion and chart analysis.
Worked example. Across a Revenue tab and an Exchange-Rate tab, verify 'In Year 3 the Americas division earned more in GBP than UK and Europe combined.' Converting ($110m / 1.25 = £88m) vs (£50m + EUR65m / 1.15 = £106.5m) gives False.
Common traps. The 'tab hang': wasting seconds on the wrong tab when a metric is split across documents, plus blind end-of-test guessing into the negative-marking penalty.
How to handle it. Do not expect to finish all 37. Aim for high accuracy on 24-28, scanning keywords to the matching tab before calculating.
Verbal reasoning (Scales Verbal)
49 questions · 12 minutes (about 15 seconds per question)
What it tests. High-speed scanning, inference vs explicit fact and comprehension of dense text.
Worked example. Given a passage on Basel IV compressing ROE 'unless offset by operational efficiencies', the statement 'Every bank under Basel IV will see an ROE drop' is False, because the text allows an efficient bank to avoid it.
Common traps. Importing outside market knowledge. If the tab does not explicitly support a statement, 'True' is wrong.
How to handle it. Read the statement first, use it as a search query into the right tab, decide and move on. Never wrestle one statement past 20 seconds.
Deductive logic (SwitchChallenge)
Adaptive, up to ~30 layers · 6 minutes
What it tests. Non-verbal deduction, working memory and algorithmic tracking under stress.
Worked example. With operator 2413, the symbol in position 2 moves to 1, position 4 to 2, position 1 to 3 and position 3 to 4. Stacked operators require deducing the final or the broken code.
Common traps. Losing focus on multi-stage stacked operators by trying to track all four symbols at once.
How to handle it. Track one or two symbols to eliminate code options, and follow the numbers rather than the shapes.
Inductive logic (Scales Ix)
20 tasks · 5 minutes (about 15 seconds per task)
What it tests. Inductive pattern recognition and abstract rule generation.
Worked example. If panels follow a count sequence, the odd one out is the panel with three shapes where the rule expects two; colours and sizes are deliberate distractions.
Common traps. Getting drawn into line thickness or shading when the real rule is count, symmetry, intersection or enclosure.
How to handle it. Run a quick checklist (count, symmetry, intersections, enclosure). If no rule appears within ~7 seconds, make an educated guess and move on.
Situational judgement (chatAssess)
Scenario-based · About 15-20 minutes total
What it tests. Professional judgement, prioritisation, conflict resolution and values alignment.
Worked example. A colleague suggests a faster file-control method set up by your VP. The optimal reply thanks them, asks for detail and offers to review it with your Associate, balancing openness with respect for structure.
Common traps. Answering on personal style or short-term efficiency, and shifting tone between prompts so the consistency check flags you.
How to handle it. Filter every option through the five core values, favouring compliance, client stability and collaboration.