Numerical reasoning tests: the practical guide (with examples)

Numerical reasoning tests measure how quickly and accurately you interpret data under time pressure. They rely on percentages, ratios, averages, and chart reading — not advanced maths. Most candidates improve fastest by learning the five repeating question patterns and drilling them in order.

Numerical reasoning tests are timed assessments where you answer questions using facts and figures shown in tables, charts, or short scenarios. The skill being tested is rarely “advanced maths”. It is whether you can interpret data accurately and fast when the clock is uncomfortable. If you want questions first, use the 25 worked practice questions; if your invite names SHL, use the SHL numerical reasoning guide; if it names AON, use the AON numerical reasoning guide.

The biggest mistake candidates make is practising random questions without learning the repeatable patterns. Most numerical tests reuse a small set of data interpretations again and again, and once you recognise them you stop treating every question as new.

If you want reps before the full guide, use the free worked set. Try 25 practice questions

What to focus on first

Common traps

• Calculating before reading the ask. You can get a perfect number for the wrong question.
• Rounding too early. Keep one extra digit until the final option check.
• Overusing the calculator. Estimation often eliminates two options faster.

How to practise from here

Take ten questions from one family only. Review setup before arithmetic: correct row, correct period, correct unit, correct base. Speed comes after that habit is automatic.

What numerical reasoning tests usually measure

• Data interpretation: reading tables, bar charts, line charts, stacked charts.
• Arithmetic under pressure: percentages, ratios, averages, conversions.
• Decision-making accuracy: avoiding “near miss” answers caused by rushed reading.

The most common numerical reasoning question types

1. Percent change and growth/decline

Typical ask: “By what percentage did X increase from 2022 to 2023?”
Fast method: (New − Old) / Old × 100.

2. Proportions and ratios

Typical ask: “If category A is 30% of total and total is 1,200, what is A?”
Fast method: convert % → decimal → multiply.

3. Weighted averages

Typical ask: “What is the average price across products with different quantities?”
Fast method: (sum of (price × quantity)) / (sum of quantity).

4. Rate questions (per hour / per unit)

Typical ask: “If 4 analysts complete 12 tasks in 3 hours, what is the rate per analyst hour?”

5. Multi-step chart questions

Typical ask: “Using the chart, calculate X, then compare to Y.” The skill is sequencing, not maths.

A simple improvement plan that works

This four-week structure outperforms random practice by a wide margin:

• Week 1 — Accuracy first. Aim for clean reading and correct set-up. Do not rush. Wrong answers erase time you will never recover.
• Week 2 — Pattern speed. For each question type above, practise in batches of 10. Track time per question.
• Week 3 — Mixed timed sets. Simulate real conditions. Review every wrong answer and label the failure mode (reading error, formula error, time panic).
• Week 4 — Weakness drilling. Spend 70% of your time targeting the slowest two question types.

How to avoid the classic traps

• Misreading units (thousands vs millions, % vs decimals).
• Ignoring time period (YoY vs quarter vs month).
• Using the wrong base for percent change (always divide by “old” unless the question states otherwise).
• Rounding too early and drifting off the correct option.

When you should practise “like it’s real”

If your target roles use timed online tests, practise with:

• Strict timing (no pausing).
• No calculator unless the real test allows one.
• A quiet environment and single screen (to mimic pressure).

How forge fits in

forge should sit in your workflow as the “practice engine”: repeatable drills, tagged questions by type, and feedback that tells you why you are missing questions — not just that you are missing them. The goal is to leave every session with a specific correction, not a score.

Source notes

forge checks numerical reasoning claims against official provider material, including SHL’s candidate practice tests and cognitive assessment overview. Individual employers may change timing, calculator rules and norm groups.

Frequently asked questions