Practical test

Data entry assessment test

Accuracy transferring information between a source document and a structured form under mild time pressure.

Also searched for as a data entry skills test, data entry accuracy test, alphanumeric data entry test or a data entry clerk test. Same instrument, and this page describes all of it.

Time to complete

About 8 minutes

Format

Twenty invoice lines, one scanned source, field-level scoring

Test family

Practical skill

Included in

All plans

What this test can see, and what an interview cannot

Nobody interviews badly at data entry. The work is invisible in conversation: a candidate says they are fast and accurate, you have no basis on which to disagree, and every CV in the pile says the same thing. The first real evidence usually arrives three weeks in, when somebody downstream finds a payment sitting against the wrong account.

It is worth testing because the spread between candidates is unusually wide and unusually cheap to measure. Transcription accuracy varies enormously between people who describe themselves identically, it is stable enough that eight minutes predicts a working week, and unlike judgement it needs no rubric and no second reviewer. What it does need is a source document with the ordinary defects of a real one, and scoring that keeps apart the two things most tests fold into a single number: how much a candidate entered, and how much of it was right.

What is in the test, and what each part shows

1. Twenty invoice lines

   A scanned supplier invoice into a structured form: supplier, invoice number, date, net, tax, total.

   What it separates The baseline. How much gets entered in the time available and how much of it matches. Everything else in the test is a defect planted in this one.

2. A date in two formats

   The header reads 03/04, the line items read 4 March, and the form asks for one format.

   What it separates Whether the collision is noticed at all, and whether it is then resolved the same way every time or row by row.

3. A total that does not reconcile

   On two rows the net plus the tax does not equal the stated total.

   What it separates Whether the candidate enters what the document says, silently corrects it, or flags it. Entering what it says is the right answer. Flagging it as well is the better one.

4. A supplier named twice

   The same company appears once as Harland and Co and once as Harland and Company.

   What it separates Whether they normalise or transcribe faithfully. Neither is wrong on its own, but doing both inside one document means they are guessing rather than deciding.

5. An illegible digit

   One account number contains a character the scan does not resolve.

   What it separates The most useful item in the test. A candidate who guesses and a candidate who flags produce an identical keystroke count and completely different downstream costs.

How the result is scored, and which number to lead on

Two figures come back and they have to be read together. Field accuracy is the share of fields matching the source exactly, weighted: a wrong digit in an account number or an amount costs more than a misspelled street, because the two are not equally recoverable once they leave the form. Throughput is reported as keystrokes per hour, the unit these jobs are advertised in, extrapolated from the timed run.

Units are quoted inconsistently across the category, so to be explicit: a keystroke is one character, and by the typing convention a word counts as five characters. Forty words per minute is therefore 12,000 keystrokes per hour, and an advert asking for 10,000 KPH is asking for a sustained 33 WPM. That is a lower bar than it sounds, which is worth knowing before you set one.

Underneath the headline figure the errors are split three ways, and the split is where the information is:

• Transcription errors. A value was entered and it is wrong. This is the one that predicts downstream cost, because a wrong value looks exactly like a right one to every system that reads it afterwards.
• Omissions. A field left blank or a row skipped. Usually a pace problem rather than an attention one, and a far cheaper failure: a blank gets caught, a transposition does not.
• Flags. The rows marked as contradicting themselves, against the rows that actually did. False flags count against, so flagging everything does not pay.

Read accuracy first and throughput second, in that order, every time. Sorting a shortlist on keystrokes per hour and checking accuracy afterwards selects the fast and careless, because throughput has the wider spread and dominates any ranking it leads. The candidate worth hiring is usually mid-table on speed and at the top on accuracy, and that person does not surface at all when the columns are read the other way round.

What separates a strong result from a weak one

Roles it suits

Data entry and records clerk

The direct fit, and the only case where the score can carry most of the decision. Weight accuracy hardest here: fidelity is the whole job, and there is no second reader downstream.

Accounts payable and invoice processing

The closest match to the material, which uses invoices for that reason. Weight amounts and account numbers above everything else, because a payment run is where a transposed digit turns into money.

Order processing and fulfilment admin

Good fit with one adjustment: addresses matter more than figures here, so the weighting should follow. Pair it with the attention to detail test when the job is mostly checking somebody else's entry rather than doing your own.

Claims, benefits and medical records

The skill transfers and the compliance context does not. Nothing in these scenarios touches a regulated situation, so add your own items on what may be entered, amended and disclosed before the score decides anything.

Bookkeeping and payroll

Usually the wrong test. When the source is a column of figures worked on a keypad rather than a mixed form, the 10-key numeric entry test measures the actual motion, and the two skills track each other loosely at best.

How long it takes, and where it goes in the process

Eight minutes, and the number is a compromise in both directions. Under about five you are measuring a burst: anyone is accurate for two minutes, and the errors that matter appear once a rhythm sets in and the source stops being read properly. Past twenty the test starts selecting for who was willing to give up an evening for your application rather than for who is good at the work.

Where it sits matters more than how long it runs. Sent before any call it is cheap enough to give to everyone who applied, which is the point: transcription accuracy leaves no trace on a CV at all, so this is one of the few places in hiring where a short test genuinely beats a CV read. Sent after a call it degrades into a formality confirming a decision already taken.

What it will not tell you

Measures care, not domain knowledge. A candidate can score at the top of the range without knowing what an invoice is, or which of the two figures on the page is the one that matters.

It is also a single sitting. Attention on transcription work degrades over hours and eight minutes catches none of that. A high score is evidence a candidate can do the work, not evidence of how long they can do it for, and that distinction is most of what separates a good first month from a good first year in these roles.

And it assumes a supervised run taken once. An open link somebody can retake until the number improves measures persistence instead, which is why the free typing test here is published as practice rather than as screening. No unsupervised online test is proof against a determined attempt; the useful response is to treat the result as a way of deciding who is worth a call, not as the decision.

Every test has a boundary like this. A test used outside it stops predicting anything, which is the usual reason a hiring team loses faith in testing altogether.

Tests that pair with this one

• 10-key numeric entry test

  Numeric keypad speed and accuracy, reported as keystrokes per hour with an error count.

  3 min

• Typing speed test

  Sustained keyboard speed and accuracy on plain prose, reported as net words per minute.

  1 min

• Attention to detail test

  Ability to spot discrepancies between documents and to follow multi step written instructions exactly.

  10 min