Skip to main content
Back to Blogs
Tips6 min read

Error Checking and Speed Tests: The Section Nobody Practises

Checking tests measure sustained accuracy at speed, not intelligence, and they are the easiest assessment section to improve. How they are scored and the scanning technique that works.

What is measured
Sustained accuracy at speed. Not intelligence, not knowledge, not reasoning. The score is usually built from how many items you cleared and how many you got wrong, together.
Typical length
Short. Saville’s error checking runs around seven minutes, Sova’s checking and accuracy roughly eight to twelve, Revelian’s proofreading task five or six. Long enough to measure sustained attention, short enough that a slow start is unrecoverable.
Who uses it
Finance and accounting, operations and logistics, contact centres, administration, quality and compliance, and increasingly anywhere with a data-entry component. Rare in consulting and graduate management schemes.
Why it is worth practising
It is the most improvable section in a typical battery, because performance depends on a scanning technique rather than on an underlying trait, and almost nobody practises it.

Checking tests are the strangest thing in an assessment battery. They are trivially easy to describe, genuinely hard to do well, and the only section where a candidate can reliably improve their result in an afternoon. They are also, by a wide margin, the section people spend the least time preparing for.

What the task actually is

You are shown two versions of something and asked to find the differences, or shown a record and asked whether it matches a reference. Names, reference codes, addresses, currency amounts, dates, product codes. The individual comparison is easy. Doing three hundred of them in eight minutes without your accuracy collapsing is not.

Vendors implement it differently, and the differences matter for technique.

Two-column comparison
Original against copy
Two lists side by side, mark the rows that differ. The most common form. Technique matters most here because you control the scanning order.
Reference lookup
Record against a key
A reference table at the top and records below, checked against it. Adds a navigation cost on top of the comparison, so the reference should be read into memory once rather than consulted per field.
Proofreading
Text with planted errors
Spelling, grammar, punctuation and consistency errors inside a passage. Revelian’s Cognify includes a version. Reading for meaning is the trap: it is what your brain wants to do and it hides exactly the errors being planted.
Classification
Apply a rule at speed
Sort items into categories according to a stated rule, quickly. Less about comparison and more about resisting the pull of a previous rule, which is why these often include a rule change partway through.
The scoring model, which changes your strategy
Most checking tests score speed and accuracy together, and several deduct for errors. Which means racing is not a strategy: a candidate who clears twice as many items with three times the error rate scores worse than a slower, cleaner one. Find the fastest rate at which your accuracy holds and stay there. That rate is discoverable in one practice run and almost impossible to judge during the real thing.

The technique that works

This is the part that makes checking tests improvable, and it is genuinely mechanical.

Compare in a fixed field order, every time

Do not scan a record as a whole. Decide an order, for example surname, then initial, then reference number, then amount, and apply it identically to every row. A fixed order turns a search into a routine, and routines are much faster and much more reliable under time pressure than judgement is.

Chunk long strings, do not read them

A twelve-character reference code read as characters is slow and error-prone. Broken into three chunks of four and compared chunk by chunk, it is fast. This is the single largest speed gain available on code-comparison items.

Commit on the first difference

Once you find a mismatch, the row is answered. Continuing to check the remaining fields is pure cost, and it is a habit people carry in from proofreading their own work, where the goal is different.

Do not read for meaning

On proofreading tasks especially, comprehension is the enemy. Your brain silently corrects errors in text it understands, which is exactly the mechanism the test is exploiting. Reading slightly against the grain, word by word rather than phrase by phrase, catches far more.

Use a physical anchor

A finger or a cursor on the row you are checking. It sounds trivial and it eliminates the largest source of lost time on these tests, which is losing your place in a dense column and having to re-find it.

The numbers trap

Numeric fields produce a specific failure that is worth knowing about. Transposition errors, where 4,821 becomes 4,281, are near-invisible when you compare by reading the numbers as quantities, because both read as roughly four thousand eight hundred. Compared digit by digit in fixed order, they are obvious.

The same applies to dates in ambiguous formats, to currency amounts differing only in decimal placement, and to codes that differ by a single character in the middle, which is where the eye is least attentive. Test writers know all three and plant accordingly.

Sustaining attention over eight minutes

Checking tests are genuinely fatiguing and accuracy characteristically decays across the sitting. Two things help and neither is glamorous.

Blink and refocus deliberately every thirty seconds or so. Screen-based comparison work reduces blink rate, and dry, unfocused eyes are a real cause of missed differences on the second half of a test.

Do not let a hard row break your rhythm. The correct response to an ambiguous item is to make a call and continue at the same pace. Slowing down after one difficult row is how a good first half becomes a mediocre second half.

Which vendors include one

  • Saville ships an error checking test as part of its Swift and Analysis Aptitude ranges. Our Saville guide covers how it sits within that battery.
  • Sova sells checking and accuracy tests explicitly for fast-paced environments, which in practice means contact centre, operational and back-office roles.
  • Revelian includes a proofreading task in Cognify.
  • Aon and several other publishers sell attention and concentration measures under their own product names, sometimes with the same underlying task.
  • Kenexa Prove It and similar skills-testing suites include data entry and accuracy tests, which are the same construct assessed as a job skill rather than as an aptitude.

Why it is worth twenty minutes

Return on preparation time is not equal across a battery. On numerical reasoning, an hour buys you meaningful but bounded improvement. On a personality questionnaire it buys you nothing. On a checking test, an hour spent establishing a fixed field order, a chunking habit and your own sustainable error rate produces a larger change than almost anything else you could do, because you are learning a technique rather than training a trait.

And nearly every other candidate is spending that hour on numerical reasoning instead.

Our Saville and Sova hubs both include checking sections at their real pace, which is the only way to find the rate at which your accuracy holds.

Practise the section everyone skips

Twenty minutes on a checking test moves your result further than the same twenty minutes anywhere else in the battery.

Try a Checking Test