SHL Numerical Reasoning Interactive: the complete 2026 guide

The SHL Numerical Reasoning Interactive test is not a chart-reading test. It is a build-the-answer test — you drag, rank, classify, and plot rather than select A, B, C, or D. The underlying maths is GCSE-level, but the interaction model is completely different from the standard multiple-choice format. If you are not sure which format you have, read which SHL test you are taking first. For the multiple-choice version, see our SHL Numerical Reasoning Test guide.

SHL runs three interactive tests — numerical, inductive, and deductive — each with a completely different task type. If you are facing multiple tests in the same assessment battery, the other two guides are: SHL Inductive Reasoning Interactive (node sequences and fill-in shapes) and SHL Deductive Reasoning Interactive (constraint satisfaction drag-and-drop).

What makes the interactive format different

Most candidates who have done the standard MCQ numerical test underestimate how different the interactive format is. Three things change fundamentally:

• You cannot eliminate from options. On MCQ, you can often rule out two distractors without calculating. On interactive, there are no options — you have to compute the correct value from scratch.
• Some answers accept a tolerance; others require exactness. Pie, line, and stacked-bar answers score within ±2 of the correct value. Classification and ranking must be exact. Knowing which is which determines how much time you spend on precision.
• The animated setup takes time. Each question opens with a short animated scenario. That animation counts against your 108 seconds. Your effective calculation window is closer to 70–90 seconds.

The three rules that apply to every question

Forge’s analysis of 100+ SHL interactive numerical questions shows three habits that separate candidates who score in the top 20% from those who run out of time:

The five question types

1. Table Classification

You are given a policy with one or more conditions and a table of rows (often sales reps, products, or employees). For each row you drag it into the correct category bucket — typically "meets criteria", "does not meet criteria", or a tiered category system.

The method is pure logic, almost no arithmetic. Convert the policy into a set of AND/OR conditions written out. Then test each row against the cheapest disqualifying condition first — the one that’s fastest to check. If any condition fails, the row goes in the reject bucket; you do not need to check the rest.

For tiered classification (multiple tiers like Bronze / Silver / Gold), test the top tier’s conditions first. If they fail, drop to the next tier. Do not assume a rep who fails Silver automatically gets Bronze — they still need to pass Bronze’s conditions.

2. Drag Ranking

You are given a team of 4–6 people with word relationships describing their values — "Cara earned 60% more than Ben", "Ana earned 10 less than half of Femi’s total" — and one person with an absolute number. You derive every person’s value, then drag them into rank order.

The method Forge calls Anchor → Dependents → Remainder:

• Anchor: find the one person given an absolute number. They are your starting point.
• Dependents: resolve anyone whose value you can compute now that you have the anchor. Continue until everyone is resolved.
• Remainder: one person is always "the rest" — total minus everyone else. Compute them last.
• Then rank. Do not sort until every value is final.

3. Pie Chart

You set each slice of a pie chart using sliders so that the segments satisfy stated equations or ratios and sum to 100%. There are two sub-types:

• Anchor given: one segment’s percentage is stated. Derive each dependent slice using the relationships, and set the final slice to 100 minus the rest.
• No anchor, only ratios: set the smallest part = x, express every other segment as a multiple of x, sum the multiples, solve for x so they total 100.

Pie questions use tolerance scoring (roughly ±2 percentage points). Set your slider to the nearest integer and move on — do not chase the decimal.

4. Line Graph

You plot points on a line graph. Two sub-types appear in Forge’s interactive database:

• Sequential rule: a starting value plus a step that may have a conditional adjustment (e.g. "each month, add 3; if the value strictly exceeds 50, subtract 7 instead"). Walk through period by period, apply the step, test the condition after stepping.
• Derive-from-data: each point is computed from a transaction or rate (e.g. a foreign exchange rate applied to a set of transaction values). Identify the formula once, apply it to every period.

5. Stacked Bar Chart

You set both the total height of a bar and the split between its segments (typically two segments: a base and a top layer). The most common scenario: one segment grows by a stated percentage, the total grows by a different percentage, and you have to work out both the new total and the new segment split.

The critical rule: tag every percentage to its correct base. A 40% increase in the premium segment is not a 40% increase in the total bar. Apply each percentage to the quantity it acts on, then recompute the total from the new segments.

Timing and pacing strategy

108 seconds per question sounds generous. It is not. The animated setup on each question takes 10–20 seconds you cannot skip. Your real working window is 70–90 seconds.

• Classification: 30–45 seconds. There is no computation — this is pure condition-checking. If you spend more than 60 seconds, you are re-reading instead of deciding.
• Ranking: 60–90 seconds. The chain of derived values takes time. Pre-write the anchor-dependents-remainder order before computing.
• Pie (with anchor): 40–60 seconds. With an anchor, this is arithmetic. Tolerance scoring means you do not need exactness.
• Pie (ratios only): 60–90 seconds. The algebra step (name x, multiply, sum to 100) takes an extra 20–30 seconds — do not skip it in favour of guessing.
• Line graph: 45–75 seconds. Sequential rules are fast once you have the step written out. Rate-derived graphs take longer on the formula identification step.
• Stacked bar: 60–90 seconds. Multi-step percentage problems with segment tagging. The reverse-engineering sub-type adds 15–20 seconds.

How scoring works

SHL Interactive Numerical has two scoring modes, and knowing which applies changes your time allocation:

The five cross-cutting traps

Forge’s database tags every wrong answer with its primary trap type. These five account for the majority of errors across all question types:

• Wrong qualifier: using "gross" when the question says "net", or applying a monthly rule to annual data. Fix: read the qualifier word before touching the data.
• Percentage-of-an-increase: computing 60% instead of 160% (60% more than = × 1.6). Fix: write out "X% more than = × (1 + X/100)" before computing.
• Wrong base: applying a percentage to the wrong quantity — segment when it should be total, or new value when it should be original. Fix: tag every percentage to its base before calculating.
• Boundary confusion: "strictly exceeds 50" triggering at 50 instead of 51. Fix: treat the operator like a legal clause — highlight it and state the exact boundary condition out loud.
• Remainder forgotten: failing to compute the last person or slice from the total, and leaving them as a guess. Fix: always identify the remainder item first (even though you compute them last) so you do not forget them.

If you are sitting the full SHL interactive battery, the same "read the rule before you act" discipline runs through all three tests. The SHL Inductive Reasoning Interactive guide covers how to apply it to node sequences and pattern families; the SHL Deductive Reasoning Interactive guide covers constraint chains and IF-THEN scheduling.

Related SHL guides

• SHL Tests: the complete provider hub
• SHL Numerical Reasoning Test (all formats including MCQ)
• SHL Inductive Reasoning Interactive
• SHL Deductive Reasoning Interactive
• Which SHL test am I taking?

Frequently asked questions