AbilityBenchFree score reader — raw counts, pace and accuracy
What a raw score on this assessment is, and is not
A raw score here is one number — the count of questions answered correctly — and every other figure an employer sees is built from it by a conversion that is not public. This page does the half that is public: it takes your counts, gives you the accuracy behind them with an interval, the pace that produced them against the 18 seconds an item that reaching question 50 needs, and the arithmetic of what more items would have been worth. It is free, it stores nothing, and it deliberately produces no percentile, no cutoff and no range for any job.
- 100% free
- No signup
- No percentile, by design
- Raw-count arithmetic
- Reads the drill's token
Put a raw count in and this works out everything a raw count supports, and stops where the publisher’s conversion begins. Numbers arriving from the fifteen-minute drill fill the fields on their own.
Right answers per item you actually reached
16 / 24
On the 22 you answered, 73% were right — a 95% Wilson interval of 52% to 87%. That interval is the honest width of a count this size: 22 items is not many, and two people with the same raw score can sit a long way apart inside it.
Source of these figures: typed in by hand.
Your pace
37.5 s
per item reached
Needed for 50
18.0 s
19.5 s an item away
Throughput
1.60
items a minute
| If you had reached | Raw count becomes | Pace that needs |
|---|---|---|
| 30 items | 16 + 4 = 20 | 30.0 s an item |
| 40 items | 16 + 12 = 28 | 22.5 s an item |
| 50 items | 16 + 19 = 35 | 18.0 s an item |
This is arithmetic on your own two numbers and nothing more. It assumes your accuracy would have held over items you never saw, which is the assumption most likely to be wrong: the items you did not reach are the ones you would have met while going faster. Read the rows as the size of the prize rather than as a forecast.
Where the arithmetic stops and the publisher’s begins
Everything above is a ratio of numbers you supplied. What an employer receives is a raw count plus a percentile against the publisher’s own norm group, and that second number cannot be reconstructed from the first without the norm table — which is sold with the assessment and is not published. There is no curve to fit, no leaked mean worth trusting, and no way to tell from the raw count alone whether it was normed against everyone who has taken the test or against a comparison group chosen for the role.
So no percentile appears on this page, and neither does a passing mark, a cutoff, or a range for any job family. Those belong to the publisher and to the employer, they differ by employer for the same assessment, and a number invented for that slot would be the one thing here that could cost somebody an application.
A raw count is not a fixed property of you either. It moves with the form you were given, with how much of the clock the first five items took, and with whether you had ever seen the format before — which is the only defensible reason to practice at all, and a different claim from the one a percentile makes. The note further down the page says what this exercise is and is not; it is not repeated here.
Nothing typed into the four fields is stored or sent anywhere; reloading the page clears them unless a token is still in the address bar.
How to read your own counts
Three counts and a duration, and every ratio they can honestly produce.
Give it three counts that nest
Items reached, of those how many you answered, and of those how many were right. The three have to nest, because they answer different questions: reached is about the clock, answered is about whether you skipped, and correct is the raw score in the publisher's sense. If they do not nest, the panel says so rather than quietly picking one.
Read the interval, not just the percentage
The accuracy figure comes with a 95% Wilson interval, and on a count of twenty-odd items that interval is wide — often twenty points from end to end. That width is the single most useful thing on the page: it is why two candidates with the same raw score are not distinguishable by it, and why one more lucky guess moves a percentage a long way.
Use the pace row to see which half your run was limited by
Seconds an item against the 18.0 seconds reaching all fifty needs tells you whether the clock or the questions stopped you. If you were inside the budget and still short of fifty, something else did — a couple of items that ate a minute each. If you were well outside it, accuracy on the ones you did reach is beside the point.
Technical specifications
| Input | Four numbers you type or bring over: items reached out of 50, how many of those were answered, how many of those were right, and the minutes the sitting took |
|---|---|
| Computed | Accuracy on the answered items with a 95% Wilson interval; seconds per item reached; items per minute; the seconds an item that separate you from reaching question 50; and what your own accuracy over 30, 40 or 50 items would have added to the count |
| The form assumed | 50 questions in 15 minutes, which is 18.0 seconds an item — the format Criteria publishes for the assessment. Change the minutes field if your sitting was a different length and every ratio follows |
| What an employer receives | Criteria's published candidate guide states it plainly: the raw score, meaning the number of questions answered correctly, together with a percentile comparing that count with everyone else who has taken the test |
| Handover from the drill | The practice page on this site links here with a token in the URL fragment carrying five numbers and the form number. A fragment is never sent to a server by any browser, and the token holds no answers and no per-item times |
| A malformed token | Refused outright with a message, never repaired and never partially read. A fragment that decoded into the wrong numbers would render as a confident score, which is a worse failure than an unreadable link |
| Two names, one assessment | The abbreviation and the full name are the same product. This page is the scoring half; the timed drill at the real tempo is the sibling page, and neither repeats the other |
| Never computed here | A percentile, a passing score, a cutoff, a score range for a job family, or a comparison against anybody. Those are the publisher's numbers and the employer's, they differ between employers using the same test, and inventing one costs somebody an application |
Frequently asked questions
What counts as a good score?
Nobody outside the hiring process can tell you, and a site that gives you a number is guessing. The published percentile depends on the norm group the raw count is compared against, the range an employer treats as acceptable is set by that employer for that role, and the two together are what decide the outcome — neither of them is a property of your count. Every figure circulating online for a target score is either somebody's recollection of one employer's threshold or a straight invention, and both are wrong for almost everyone who reads them.
Why does the page put an interval on my accuracy?
Because a proportion measured on twenty-odd items is much less precise than it looks. Sixteen right out of twenty-two reads as 73%, and the 95% Wilson interval around it runs from roughly 52% to 87% — one lucky guess or one careless slip moves it several points. The interval is the width the count itself supports, and printing the percentage without it is the commonest way a practice site makes a small sample sound like a measurement.
My employer talked about a percentile. Where is mine?
It does not exist outside the publisher's system, and this page will not approximate it. A percentile is a rank against a specific comparison group, and Criteria sells the norm table that defines that group along with the assessment. Without it there is no way to turn a count of right answers into a rank, and every shortcut — fitting a normal curve to a remembered average, stretching the 0-to-50 range linearly — produces a confident number that is wrong by an unknown amount in an unknown direction.
If I sat it again, would I get the same raw score?
Almost certainly not, and how much it would move is not something this page can tell you from one count. A raw score carries the sampling noise the interval above describes, plus whatever the particular form gave you, plus how much of the clock the first few items took. That is the honest argument for practice: it does not raise an underlying ability, it removes the part of the variation that comes from meeting the format for the first time.
Does answering everything always beat skipping?
On this assessment a blank and a wrong answer are worth the same, because Criteria's guide states there is no penalty for a wrong answer — so an answered item can only help the count. What it can cost is the seconds it took, which is the trade the pace row is for. The projection table shows the size of the prize: your accuracy applied to the items you never reached. It also names its own weakness, which is that the items you did not reach are exactly the ones you would have met while hurrying.
Why is there no timer on this page?
Because nothing here is measured — every number on the page was typed in or arrived from the page that did the measuring. That is also why this is the one page in this group with no note about display timing: a duration you supplied carries your stopwatch's error, not the browser's, and pretending otherwise would be theater.
What exactly travels in the link from the practice page?
Five numbers and a form number, encoded into one opaque token in the part of the address after the hash. Browsers never send that part to a server, so following the link moves your counts between two pages of this site without them leaving your machine. The token contains no answers, no per-item times and nothing about you or your device, and this page refuses it outright if a single character has changed.
Raw score, percentile, and the gap between them
The scoring of this assessment is unusually simple at one end and entirely closed at the other. At the open end, the raw score is a count: how many of the fifty questions you answered correctly, with nothing subtracted for a wrong answer and nothing added for speed. At the closed end, the report an employer opens carries that count and a percentile beside it, and the percentile is the product. It is a rank against a norm group the publisher assembled and licenses, and there is no public route from one to the other. That gap is why this page exists as something other than a second drill: the arithmetic on the near side of it is genuinely useful and almost never shown, and the far side cannot be crossed honestly at all.
Two consequences follow that are worth carrying into the room. The first is that a score range attached to a role is the employer’s setting rather than the test’s. The same assessment is used for a warehouse supervisor and for a quantitative analyst, and what counts as sufficient is configured per role by whoever is hiring — so a target number quoted in a forum is a fact about one company at one moment, not about the instrument. The second is that a raw count of the low twenties is not the anomaly people assume, because reaching all fifty questions is rare by the publisher’s own account. A count is bounded above by how far you got, which makes pace and accuracy two different levers with two different fixes, and the panel above separates them for that reason.
The figure this site will not print, and why
Commercial hiring assessment norms — Wonderlic, CCAT, Predictive Index, Hogan, Caliper, Korn Ferry, SHL, Watson-Glaser
Every one of these publishes its norms only to licensed customers, and several norm by role and by employer rather than against a general population, so a single 'average score' would be wrong even if it were public. The item banks are copyrighted as well: reproducing them would breach the spec's rule that these pages are practice and preparation, never the instrument. The same applies to the mechanical comprehension tests these pages sit next to, which are sold as hiring instruments with their own restricted norms.
If you came here for the practice rather than the arithmetic, the drill under the abbreviation runs fifty generated items against the real fifteen minutes and links straight back here with its counts. The twelve-minute paper is the same species at a tighter tempo, and the Predictive Index page measures the thing this page can only describe, by running you over two equivalent forms and showing how far your own count moved between them. Where a process pairs an aptitude test with a questionnaire — and it usually does — the Hogan practice page covers how scale-based inventories are reported, and the word association test is a reminder of what a measurement without a scoring key looks like from the inside.
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 a learning difficulty behind a low count, or anything about how you will do the job. Only a qualified professional, working with more than a browser, can make that judgment.
Not affiliated with, endorsed by, or connected to Criteria Corp. Criteria Cognitive Aptitude Test is a trademark of its owner and is used here only to name the assessment this page prepares for. The questions on this page are our own: no part of the published test is reproduced, and a score here is not comparable to a score from the real instrument.
Where these four numbers live
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 four fields are ordinary state in this tab. Nothing is written to storage, and the only place a count can persist is the fragment of the address bar if you arrived through a link — which means a copied URL carries those five numbers, and a URL you did not copy carries nothing.