Error checking tests and a scan that holds up

The skill

What is an error checking test?

An error checking test gives you two versions of the same information and asks whether they match. The information is deliberately meaningless — reference codes, account numbers, names, dates, amounts — because the moment content becomes interesting, candidates start reading it instead of comparing it.

SHL describes its checking format directly: each question shows a piece of data on the left, and you match it to the correct data from a list on the right, which contains similar but not identical samples. Korn Ferry describes its checking test as measuring the ability to quickly and accurately detect errors in data.

The intellectual content is close to zero and the discipline required is high. That combination is the point: employers use these tests where a single mistyped digit has a real cost, and what predicts performance is consistency rather than cleverness.

Try it before anything else

Free error checking practice questions

Six original questions covering the formats above, from counting errors to certifying an exact copy. Run your scan and commit to a verdict before opening the solution — each one walks the comparison chunk by chunk and names why every wrong option is on the list.

What employers are looking at

What do error checking tests measure?

Employers use checking items for roles with high transaction volume and low tolerance for error: finance and payments, data entry, claims, compliance, logistics, administration and customer operations. Saville groups error checking with verbal and numerical comprehension in the battery it describes for administrative, customer service, operational and commercial roles.

Two things are being measured here, and neither is attention to detail as a personality trait. Method stability: does the scan you use on question twenty match the one you used on question two. Error sensitivity: are you equally likely to catch a transposition, which reuses the same characters, as a substitution, which introduces a new one.

Reading your result

How error checking tests are scored

The tempting metric is how many questions you got through. The providers reviewed here weigh accuracy against the time available and leave the benchmark to the employer, which changes both what a good result means and what is worth practising.

• Completed questions are not a score. Checking is normally scored on accuracy within the time available rather than on volume alone. The number of questions you answered is not a result you can interpret on its own, and it is the wrong number to chase in practice.
• No provider reviewed publishes a pass mark. Korn Ferry states directly that it sets no pass or fail marks and that employers choose their own benchmarks. If a recruiter has disclosed a target, that is the only target worth working to.
• Inside a battery, checking may arrive as a sub-score. Saville reports one total with individual sub-scores on its combination tests, and places error checking alongside verbal and numerical comprehension in its Swift Comprehension range. Your checking result can reach the employer as part of a combined figure rather than as a standalone number.
• On adaptive delivery, difficulty is not feedback. Korn Ferry states that all of its cognitive ability tests are adaptive, with the next question selected according to whether the previous answer was correct and each question carrying its own time limit. A harder question often follows a correct answer, so you cannot read your score from how hard the test feels.
• The numbers that move in practice are error rate and error type. Track how often you miss and which kind of error you miss — omission, substitution or transposition — because those are the two numbers that respond to training. Speed added before they are stable multiplies errors rather than score.

Question formats

What do the questions look like?

Checking assessments come in three broad shapes, and the response action matters as much as the comparison.

• Match one item against a list. SHL's published format: a piece of data on the left, and a list on the right containing similar but not identical samples, from which you select the exact match.
• Same or different. Two strings side by side and a binary verdict. There is the least to read per item here, and a drifting scan is hardest to notice, because there is no third option to slow you down.
• Multi-field record comparison. A source record and an entered record, each with several fields — reference, date, amount, name — and a question about how many fields differ or which one does.
• Counting the differences. Rather than reporting whether an error exists, you report how many there are. Stopping at the first mismatch is fatal here, because a count needs the whole record read even after a difference has been found.
• Classification rather than comparison. Some formats ask which kind of error occurred rather than whether one did, which requires the vocabulary of omission, substitution and transposition.
• Checking inside a wider battery. Saville places error checking alongside verbal and numerical comprehension in its Swift Comprehension range and inside Swift Apprentice Aptitude, so it often arrives as one section of a short combined test.

Launch coverage

Which providers use error checking tests?

Three launch families publish a checking assessment, and their delivery models are genuinely different.

• Verify Checking. SHL's published example format shows a piece of data on the left to be matched against a list on the right containing similar but not identical samples. SHL states that the test measures the ability to compare information quickly and accurately, which it describes as critical to ensuring the quality of work. SHL does not publish a time limit for it in its candidate example material.
• Talent Q Aspects Ability Checking. Korn Ferry lists Aspects Ability Checking at 6 minutes and describes checking tests as measuring the ability to quickly and accurately detect errors in data. It states that all of its cognitive ability tests are adaptive, with each question carrying its own time limit and the next item selected from whether the previous one was correct.
• Error Checking. Saville publishes Error Checking as a single aptitude test with its own free practice test, and inside two batteries: the Swift Comprehension Aptitude range at 9.5 to 16 minutes with verbal and numerical comprehension, and Swift Apprentice Aptitude at under 20 minutes across six areas. Saville states that its fixed-length format gives a gradual increase in difficulty.

Provider names identify the assessment format. Cognivy is independent and is not affiliated with or endorsed by these providers.

What varies, and what does not

Timing, delivery and scoring

These are short assessments, and the difference between a per-question limit and a whole-paper limit changes the right strategy completely.

• Korn Ferry Aspects Ability Checking: 6 minutes, adaptive. Korn Ferry states that each question has its own time limit and advises using the full time given rather than rushing into an unconsidered answer. On a per-question clock there is no time to bank, so care costs nothing.
• Saville places checking inside short batteries. The Swift Comprehension Aptitude range runs 9.5 to 16 minutes for two to three aptitudes including error checking, so the checking section is a few minutes rather than a paper of its own.
• Fixed-length means the hard items come last. Saville states that its fixed-length format gives a gradual increase in difficulty. On a fixed paper, spending your reserve early costs you the items that were designed to discriminate.
• SHL does not publish a checking time limit for candidates. SHL's candidate example material describes the task and what it measures but not its length. Take the timing from your own invitation rather than from a figure quoted on a practice site.

A repeatable approach

A scan that survives the clock

One scan, used identically every time. The whole assessment is a test of whether you can do that, so it is worth building the scan deliberately.

1. 1

   Chunk before you compare. Break a code into groups of three or four characters and compare group against group. Whole-string comparison exceeds what most people hold accurately, and it is where the middle of a long reference gets skipped.

2. 2

   Go field by field in a fixed order. On a multi-field record, always start at the same field and move the same way. A consistent route means a field is never checked twice and never missed once.

3. 3

   Say the character class as you go. Letter O against digit zero, letter I against digit one, letter Z against digit two. Naming the class catches the substitutions that are invisible at a glance and are deliberately planted.

4. 4

   Never stop at the first mismatch. Finding one error feels like finishing the question, and on a counting item it is exactly half an answer. Re-anchor at the next chunk and complete the scan.

5. 5

   Answer and move without re-reading. A second pass on a checking item almost never changes the verdict and always costs the next question. Trust a scan you have practised rather than checking your check.

What slows progress

Common mistakes, and the fix for each

• Letting the scan drift. Fix: rehearse one route until it is automatic, and use it identically on the last item and the first. Accuracy in this format collapses through inconsistency rather than through any single hard question.
• Comparing whole strings. Fix: chunk into groups of three or four. A long reference is more than one comparison can hold at once, and the middle is where the planted error usually sits.
• Stopping at the first difference. Fix: finish the scan every time, including after you have found an error. On any item that asks how many, half an answer scores the same as none.
• Speeding up before accuracy is stable. Fix: hold the pace until your error rate is flat across a full set, then increase it. Speed added to an unstable method multiplies the errors instead of the score.
• Missing character-class swaps. Fix: name the class when you meet an ambiguous character. Letter O against digit zero, letter I against digit one, letter S against digit five. These are planted deliberately and they are invisible to a fast visual pass.

In practice

How do you compare two reference codes without reading them as words?

Break both codes into the same chunks of three or four characters and compare group against group, in the same order every time. Reading happens when the eye takes a string whole and recognises it instead of checking it — sample question 5 above shows the cost, because a reader silently normalises Harriss to Harris and certifies the error as a match. Chunking keeps the comparison at the level of characters, which is where the errors live.

Name the character class wherever a code mixes letters and digits: letter O against digit zero, letter I against digit one, letter S against digit five. These substitutions are planted precisely because a visual pass slides over them, and naming the class — even silently — forces the distinction a glance never makes.

Let the punctuation set your chunk boundaries where the string provides it. Hyphens and spaces are free anchors — 8QK-0937-LN2 in sample question 2 is three comparisons rather than twelve characters — and a fixed set of stopping points is what stops the middle of a long reference being skipped.

In practice

Why are transpositions the hardest errors to catch?

Because a transposition leaves the character set intact. An omission changes a string's length and a substitution introduces a character that was never there, but a transposition merely swaps two neighbours — 40712 against 40172 in sample question 1 contains exactly the same five digits. Any check that asks whether the right characters are present passes it; only a check that asks whether they are in the right order catches it.

The fix is comparison at chunk level, with order held inside the chunk. Three or four characters are few enough to hold as a sequence rather than as a bag of characters, which is exactly what whole-string comparison cannot do. It is also why the middle of a long reference is the classic planting site: the ends of a string get compared as sequences, while the middle gets recognised as familiar.

Transpositions are also the reason to log error types during review rather than a bare right or wrong. If your practice record cannot show that omissions and substitutions are already at floor while transpositions persist, it cannot show you the repair either — which is slowing the within-chunk comparison, not slowing the whole scan.

In practice

How should you pace an error checking test?

It depends on which clock you are under, and the two reward opposite habits. Korn Ferry gives each question its own time limit and advises using the full time given rather than rushing into an unconsidered answer. On a per-question clock there is no time to bank: seconds saved on this question are not available for the next one, so care costs nothing.

Saville's fixed-length tests run the other way, with one limit for the paper and what Saville describes as a gradual increase in difficulty. There, the reserve matters: the questions designed to discriminate arrive late, and time spent double-checking easy verdicts early is spent on exactly the wrong questions.

Under either clock, answer from the scan and move without re-reading. A second pass on a checking question almost never changes the verdict, costs the next question on a fixed paper, and buys nothing on a per-question one. The pace you rehearsed at is the pace to trust — raised one increment at a time in practice, only after the error rate held.

A clear route

How to prepare in the days you have

What is being trained here is a routine rather than an insight, which means preparation has something concrete to rehearse. Work in this order.

1. 1

   First, choose your scan and write it down. Chunk size, direction, field order. Committing to one specific routine is more valuable than any amount of unstructured practice, because it is the thing that has to survive the clock.

2. 2

   Next, practise untimed for accuracy only. Work slowly enough to be right every time and pay attention to whether you are following the routine. If accuracy is not near perfect untimed, speed will not fix it.

3. 3

   Then learn the three error types by name. Omission removes a character, substitution replaces one, transposition swaps two. Naming them out loud during review is what makes transpositions visible, because a transposition leaves the character set intact and only the order changes.

4. 4

   Add time in small increments. Raise the pace one step and check whether the error rate held. If it rose, drop back. This is the only reliable way to find your real ceiling rather than an optimistic one.

5. 5

   Rehearse the provider's response action. Selecting a match from a list, giving a same-or-different verdict, and counting incorrect fields are three different actions. Korn Ferry's per-question clock also rewards using the full time, which is the opposite of what a whole-paper limit rewards.

Cognivy uses your assessment date to choose the route rather than asking you to predict a study schedule. When you sit down to practise, you choose the session length that fits that day.

Pace

Getting faster without losing accuracy

Speed on a checking test is a by-product of a stable method. Four things add it without costing accuracy.

• Fix your chunk size. Three or four characters, the same every time. A consistent chunk size means your eye stops in the same places and nothing between the stops gets skipped.
• Reject on the first chunk where you can. On a match-from-a-list item, an option that fails at the first group is finished. There is no reason to read the rest of it.
• Check the ambiguous characters first. Where a code mixes letters and digits, scan the confusable pairs before the obvious ones. That is where the planted errors concentrate.
• Use the full per-question time when there is one. Korn Ferry advises using the full time given rather than rushing into an unconsidered answer. On a per-question clock, time saved on one item is not available for the next.

Keep exploring

Related test types and how they differ

Error checking borders three test types that look similar and measure something else.

Direct answers

Error checking test FAQs