Signals Numerical Reasoning
Can a candidate read figures accurately, interpret data and work through a multi-step analysis? Charts and tables, delivered so AI assistance is disadvantageous to use. About 14 minutes, 36 questions.
Powering hiring decisions at Lenovo, EDF and more
How the test works
Chart or table first, then four questions one at a time with the data still on screen.
Review the data
A chart, table or short business scenario appears on its own, with time to read it before any question is shown.
Answer one question at a time
The data stays on screen. Each question has its own short time limit and there is no going back.
Move to the next set
After four questions the next set is chosen from the bank to match the candidate's ability so far.
The full guide, to share with colleagues
The three subtests, what candidates see and how the timing works. Written with financial services teams in mind; easy to forward to whoever signs off your process.
-
Short - About 14 minutes, timed per question.
-
Phone or desktop - Instructions built in.
-
Standard format - Read the data, choose the answer. No new format to learn.
-
Reasoning, not knowledge - Everyday business data. No finance or sector knowledge needed.
Three subtests, from quick reading to multi-step analysis
36 questions in about 14 minutes. The time per question changes with the subtest, so quick reading is tested quickly and analysis gets the time it needs.
Use it on its own or with Signals Verbal Reasoning in one sitting, with one report. Both are part of the Signals Series.
Numerical Fluency
16 True, False or Cannot Say judgements, eight seconds each. Can the candidate read a figure correctly, fast?
What do the numbers actually say?
Numerical Interpretation
12 questions, around 20–25 seconds each. Can they work out what the data shows?
Is the calculation right?
Numerical Analysis
Eight questions, around 45 seconds each. Can they reason through a multi-step problem?
Is that conclusion justified?
Numerical assessment without an easy AI shortcut
A chart with four questions underneath is easy to hand to an AI tool. One timed question at a time is slow and unreliable to cheat on. No remote assessment is completely immune, and we keep testing Signals as AI tools change.
One question at a time
One screenshot cannot capture a whole set, and there is no previewing what comes next.
Short answer windows
Little time to capture a question, ask an outside tool and enter the answer.
Adaptive selection
Candidates see different questions, so a shared answer set is of little use.
Built and maintained by our business psychologists, with every question calibrated and monitored for fairness. Technical documentation on request.
-
A percentile score against the comparison group you choose
-
Results in the platform, with PDF reports to share
-
Scores exported to CSV or Excel
What clients say
On Test Partnership and our wider assessment service.
Since they upgraded these to gamified tests, it has only enhanced their effectiveness. Add to that their ability to identify candidates' true potential rather than that of ChatGPT, we consider it an invaluable part of our skills assessment process.
Test Partnership has helped us bring more structure and objectivity into our hiring process. The assessments give us useful insight into candidate potential and help us identify stronger talent more efficiently.
These tests help us quickly narrow down large applicant pools and focus on candidates most likely to succeed. The platform is easy to use, and their team provides excellent support whenever needed.
Signals Numerical Reasoning FAQs
Graduate, apprentice, intern and analyst intakes in finance, professional services and any role where candidates work with data. No finance degree or specialist knowledge needed.
At the stage where you assess numerical ability, before interviews or an assessment centre.
Yes. It is designed for the candidate’s own device, including a phone.
Talk to our team before your campaign and we will configure the assessment for candidates who need adjustments or assistive technology.
Data first, one timed question at a time, no going back, and adaptive selection make an AI tool slow and unreliable to use mid-test. For our highest level of AI resilience, see MindmetriQ.