AbilityBenchData interpretation
Numerical reasoning test that separates finding the numbers from using them
Twelve tables and charts with five options each, a four-function calculator beside every question, and no time limit on any single item — free, with no account and no partial result. Each item opens with the data and the question and no answer options, so the seconds before you ask for them are counted as search and the seconds after as arithmetic; the report gives both, because a graduate-scheme paper is usually lost in one of the two and candidates cannot tell which. Every wrong option is the output of one named mistake, so the result says what you did rather than only that you were wrong.
- 100% free
- No signup
- 12 items
- Calculator allowed
- Search vs solve split
Twelve items, every figure you need printed on the screen, and a four-function calculator beside the question because the employer version allows one. The run is scored on what you answered and on where the seconds went: an item opens with the data and the question and no options, and the control that reveals them marks the boundary between hunting for the right two numbers and doing something with them.
Nothing is timed out. The 75-second marker beside each item is a pace guide, not a cutoff, and an item you spend four minutes on is still scored.
How to take the numerical reasoning test
One table, one question, two timings and a named mistake behind every wrong option.
Read the question before you read the table
The data appears with the question underneath it and no options in sight. Work out which two or three of the sixteen to twenty printed figures the question is about — the row, the column, and whether the heading carries a multiplier. The item names the figures it depends on at the foot of the panel, so you can check afterwards whether you looked in the right place.
Press the space bar the moment you know which figures you need
That press is the boundary the report is built on: everything before it counts as search, everything after it as calculation. It is your own judgment of when you had the figures, which makes the split as honest as you are with it and is why the page says so rather than presenting it as measured. Tick the box on the start screen to keep all five options on screen from the beginning, the way the employer version does, and the split is dropped instead of guessed.
Answer with A to E, then read what each wrong choice was
Options are lettered so the number keys stay free for the calculator: digits, the four operators, Enter for equals and Escape to clear all work without clicking into it. At the end the result lists every item you missed with the specific error that produces the option you chose — dividing by the later figure, reading the row below, treating a percentage-point gap as a percentage change, or ignoring the multiplier in the column heading.
Technical specifications
| Items | 12 scored — 8 tables and 4 charts, with the four table kinds cycled rather than drawn at random so no breakdown cell ends up holding one item — plus 2 optional worked items that appear in no figure |
|---|---|
| Item kinds | Percentage change, share of a column total, a per-site rate across two columns, a one-period projection, and two chart kinds: a percentage change and an absolute difference read against gridlines |
| Options | Five per item, lettered A to E, each a figure produced by a specific misstep rather than a number near the answer. An option whose arithmetic lands on the correct value is dropped and replaced, so no two options ever render the same string |
| Calculator | Four functions, a 12-digit entry limit, and full keyboard control while an item is up. It is present because the employer format allows one — a numerical reasoning paper is not testing whether you can divide |
| The units trap | Two of the twelve items state the multiplier in the column heading and then ask for the answer in whole dollars or whole units, which is the single most common careless loss on a real paper |
| Pace marker | 75 seconds an item, which puts the twelve inside a quarter of an hour. That figure is ours and not a publisher's, and nothing times out: an item you spend four minutes on is still scored |
| What voids an item | Three things void one: a letter pressed inside 150 ms of the options appearing, an answer landing before the item has painted, and a visit to another tab mid-item. Whatever is voided comes back as a fresh item of the same kind at the end, so the run still delivers twelve |
| What leaves the page | Nothing. Every table, chart and option is assembled here in the tab out of one integer, the answer key never exists anywhere but this page's memory, and the copy button writes plain text straight to your own clipboard |
Frequently asked questions
Am I allowed a calculator on a real numerical reasoning test?
Almost always, and that fact is the key to the whole format. The employer version prints every figure you need and lets you compute with it, which means the difficulty was never in the arithmetic — it is in deciding which two of twenty numbers the question is about, under a clock. Test publishers who withhold a calculator are usually running a separate, shorter numerical-computation section and will say so in the invitation email. If yours does not say, assume one is allowed and check the instruction screen before the timer starts.
Why are the answer options hidden until I press a button?
Because with five options visible you can back-solve, and back-solving destroys the measurement this page is for. Given the options, a candidate who cannot find the right cells can still eliminate three figures by size and guess between the rest, which is a real exam skill and a useless diagnostic: the run then produces one number that mixes searching, calculating and eliminating. Hiding them until you say you have the figures splits the first two apart. The trade is honest and it is stated on the start screen, where you can turn the split off and get the faithful format instead.
How do I tell whether a question wants percentage change or percentage points?
Percentage points are the arithmetic difference between two percentages; percentage change is that difference divided by the earlier percentage. A share that moves from 20% to 25% has risen 5 percentage points and 25%, and both are correct answers to different questions. The wording that decides it is the preposition: a rise of 5% is a change, a rise of 5 percentage points is a subtraction. Employer papers put both figures in the options deliberately, so an item where two options are 5 and 25 is telling you which distinction it is testing.
The column heading says thousands. Do I put the zeros back?
Only when the question names the unit, and then always. A heading reading $ thousand means the cell showing 1,240 is $1,240,000, so a question asking for revenue per site in dollars needs the multiplier applied before the division and a question asking for a percentage does not need it at all — a ratio is unaffected by a multiplier both figures share. Two of the twelve items here are built on exactly that asymmetry, because it is where a candidate who did the hard part right loses the mark anyway.
There is no clock on the items. Isn't a real test timed?
It is, and this page reports your pace instead of imposing one, for a reason worth stating: a cutoff tells you that you ran out of time and nothing about where the time went. The report gives your median item, your median search and your median calculation, against a 75-second marker that puts twelve items inside a quarter of an hour. If searching is the larger half — and for most people it is — practicing arithmetic will not move it, and a per-item clock would have hidden that.
Why is there no percentile, grade or band on my 9 out of 12?
Because the phrase describes a shape of question, not one measured test, and a shape of question has no population behind it to be a percentile against. The score depends entirely on how hard this page made its own items: make the tables wider and the answer changes without anything about you changing. Commercial tests in this format do have norms, and those belong to the publishers who collected them and are not published. What transfers instead is the breakdown — which item kind you missed, and which specific error produced the option you picked.
Is this the same thing as a mental math test?
No, and the two are worth keeping apart if you are preparing for something specific. This page hands you the figures and asks you to select and combine them with a calculator in reach; the mental arithmetic drill gives you two numbers and takes the calculator away, and reports a median time per operation instead. An employer numerical reasoning paper is this page. A trading-floor or teaching-assessment numeracy screen is often the other one, and a few graduate schemes run both.
What an employer numerical reasoning paper is actually testing
The format is data interpretation, and it is not a maths test in any useful sense. Every figure needed is printed on the screen, a calculator is normally allowed, and the arithmetic in a typical item is one subtraction and one division. What makes a twenty-minute paper hard is the selection step: a four-by-four table carries sixteen numbers, a stub column, a header row and often a multiplier stated once at the top, and the question names two of those figures in prose. Candidates who fail these papers rarely fail the division. They spend seventy seconds locating a cell, answer nine items instead of eighteen, and conclude they are bad at maths. That is why this page times the two halves apart rather than reporting one number for the item.
The other thing worth knowing is how the wrong options are built, because it is the same everywhere and it is not random. A publisher writing a numerical item does not scatter plausible numbers around the answer; it computes the figures that predictable errors produce and offers those. Divide the change by the later value instead of the earlier one and there is an option waiting. Read the row below the one named, and there is another — arithmetically flawless, and wrong. Treat a gap between two percentages as a percentage change and the option is there too, which is why an item whose choices include both 5 and 25 is telling you what it is testing. Ignore the multiplier and your answer is out by exactly a thousand, sitting in the list. This page builds its distractors the same way and then does the thing a real paper never does: it tells you afterwards which error you made, because on twelve items that is the only output with any information in it. Three wrong answers of one kind is a finding; three of different kinds is noise.
There is no population norm for 'a verbal reasoning test' or 'a pattern recognition test', because those name a question format rather than an instrument. The score depends entirely on how hard the page made its own items, so any average quoted would describe this page's difficulty setting and nothing about people. The commercial tests in these formats do have norms, and those are restricted — see the hiring-assessment entry. So nothing here converts into a rank, and the useful comparison is between item kinds inside your own run. If the selection step is what is costing you, the neighboring pages take it apart from other directions: the numeracy test puts the same skill into a price tag and a utility bill, where the irrelevant figures are the whole difficulty, and the math aptitude test removes the calculator and the clock together to ask whether the quantities make sense to you at all. If the arithmetic is the slow half instead, the mental arithmetic drill reports where it breaks down by operation, and the sixty-second sprint gives you one throughput number to beat. Graduate assessments pair a numerical paper with a verbal one almost without exception, so the verbal reasoning test and, for law and consulting, the Watson-Glaser format practice are the other half of the same afternoon.
A reaction time here is the interval between the frame that painted the stimulus and the timestamp the browser attached to your key, both read from the same monotonic clock. What neither can see is the display pipeline behind it, so on a 60 Hz screen roughly 16 ms of every figure below is the machine rather than you. That is the timing floor: two numbers closer together than that are the same number, and this page reports no precision it cannot support.
This page therefore reports item times in tenths of a second and never in milliseconds. A search that took 41.3 seconds and one that took 41.4 are the same search, and the difference between the two halves — typically tens of seconds, not tens of milliseconds — is thousands of times larger than anything the display can add.
This is a measurement exercise, not a clinical assessment. It reports what you did on this page against a stated reference and nothing more — it cannot establish dyscalculia, a numeracy difficulty, or whether you will pass a particular employer's paper. Only a qualified professional, working with more than a browser, can make that judgment.
Where these tables come from and where your answers stay
Every number on this page is worked out by JavaScript running in the tab you are reading it in. Your answers, your reaction times and your score are never uploaded, logged or kept — which is also why the test carries on working after you disconnect from the network, and why nothing here can be held back behind an email address.
The tables and charts are not fetched. One integer decides all twelve of them, and that integer appears beside your score, which is what lets an identical set be reproduced on another machine with nothing having been kept anywhere. That is the whole of the round trip: no request leaves the page during a run, including the one that would tell us you started it.