Job knowledge test
Excel assessment test
Practical spreadsheet work at three levels: lookups across sheets, pivot tables, conditional logic and cleaning a messy sheet before anyone can use it.
- Time to complete
- About 20 minutes
- Format
- Three levels, you send one. Browser based, open book, no install
- Test family
- Job knowledge
- Included in
- All plans
What this test can see, and what an interview cannot
Almost everyone applying for a job that touches a spreadsheet says they are proficient in Excel, and the word carries no information. An HR manager hiring temporary staff for basic formula work put the gap plainly: everyone says they are proficient and then surprise, some have trouble just doing copy and paste. Proficient is a self-assessment against a standard the candidate picked, and an interview does not fix it: asking someone to describe a lookup tests whether they can describe a lookup.
A pre-employment Excel assessment test replaces the claim with a sample of the work. This one runs in the browser, takes about twenty minutes, and asks for specific outputs from sheets shaped like the ones your team actually receives: exports whose keys do not quite match, numbers stored as text, and a few thousand rows nobody has summarised yet.
It is deliberately not a multiple-choice quiz about which ribbon a feature lives under. Menu knowledge is the part that moves between versions and the part anyone can look up in seconds. What does not move is whether a person can get the right number out of a messy sheet, which is the thing the job is actually asking for.
The three levels, and what each one asks
Foundation: a sheet somebody else built
Sort and filter a supplied list to answer one specific question about it, copy a formula down a column so the references still point where they should, and fix a column of numbers stored as text so that it totals correctly.
What it separates The applicants who cannot do the day-to-day at all. Most people clear this level, and the ones who do not are the reason the test exists.
Working: two exports that do not agree
Join two sheets on a shared but inconsistently formatted key, build a monthly summary from four thousand transaction rows, and write a conditional that flags the rows meeting two conditions at once.
What it separates Whether strong Excel on a CV means the level most operations, admin and bookkeeping jobs actually run on. This is the level to send if you are only sending one.
Advanced: a model other people will rely on
Find and explain the three errors in a supplied budget model, restructure a sheet so next month can be added without rewriting formulas, and say which figure in a summary you would not trust.
What it separates Judgement rather than recall. Send it only when somebody else will depend on the files this person builds, because it rejects candidates who would have done a lighter job well.
How the result is scored, and how to read it
Per task, on the output, and never as a single percentage. One number hides the part you needed: it cannot tell you whether a candidate cleared the basics and stalled on the join, or scrambled the basics and then landed the pivot table by luck. Results come back task by task and grouped by level, so what you read is where the ceiling sits.
- Task output. Scored, and it is effectively the whole of the score. Whether the joined sheet, the total or the flagged rows are actually right.
- Level reached. Reported as the level the candidate stopped clearing, not as an average smeared across all three.
- The explanation on the advanced task. Scored. Finding an error in a budget model counts for little if the candidate cannot say what the error was.
- Time taken. Recorded, not scored. It is context rather than a verdict: on most of these roles a slow correct answer is worth more than a fast wrong one.
- Method. Not scored. Formula, pivot table or query, because the file your team receives does not care which was used.
What separates a strong result from a weak one
| Reading the report for | A strong candidate | A weak one |
|---|---|---|
| Lookups across sheets | Normalises the key first, then joins, and finishes with one row per record. | Joins straight away and leaves a column of errors on every row whose key had a trailing space. |
| Summarising volume | Produces the monthly view in one pass, and it still recalculates when a row is added. | Arrives at the right figure by hand for the sample shown, and could not repeat it next month. |
| Messy data | Notices the numbers stored as text before totalling the column. | Totals the column, gets a plausible-looking figure, and reports it. |
| Conditional logic | Flags exactly the rows meeting both conditions. | Flags the rows meeting either one, or hard-codes the ones visible on screen. |
| Judgement, advanced level only | Names the figure they would not trust, and gives the reason. | Reports every number with the same confidence. |
Roles it suits
- Financial analyst
- Advanced level. The error-finding task predicts more here than speed does, because much of the job is reviewing a model somebody else built.
- Operations coordinator
- Working level. The join and the monthly summary are the daily work, and foundation tasks alone will not separate this field.
- Bookkeeper
- Working level, read with the messy-data tasks weighted. Numbers stored as text and keys that do not match are the friction of the actual job.
- Sales operations
- Working level. The join across two exports is the task that matters, because CRM and billing exports rarely agree on formatting.
- Administrative assistant
- Foundation level is usually enough. Sending the advanced set here rejects people who would have done the job well.
An Excel test and a Microsoft certification answer different questions
A Microsoft Office Specialist certification is a candidate-paid exam against a published syllabus for a specific version of the software. Its value is that it travels: an independent body watched this person meet a defined standard on a defined date, and you can verify that without taking anyone's word for it.
It is not, though, the question in front of a hiring manager, which is narrower. Can this person do what this job needs, on data shaped like yours, now. A certificate cannot answer that because it has never seen your data, and its absence answers nothing either, since most competent applicants never sat one. Requiring the certificate mostly filters for who had a training budget.
The two are not in competition. If a candidate holds a current certification, treat it as a real signal and send the working level anyway: the certificate covers the features, the test covers a sheet that arrived broken.
How long it takes, and where it goes in the process
Twenty minutes at one level. Length is a real cost rather than a design preference: a candidate who is also interviewing elsewhere weighs an unpaid hour against the odds of getting the job, and the people with options decline first. Twenty minutes at the right level separates a field better than an hour at the wrong one.
Send it at the same point in the process to everyone applying for the same role, and say in the invitation how long it takes and that it is open book. The consistency is what makes two results comparable if a rejection is ever questioned; the disclosure is what keeps the strong candidates in.
What it will not tell you
Scores reflect the version and locale used. Formula names differ across localised Excel installations, so tasks are phrased to avoid version specific syntax.
It also says nothing about whether the person will use a spreadsheet appropriately. Somebody can clear the advanced level and still put a company's forecasting in a workbook that should have been a database. That judgement surfaces in the interview, in the answer to why they structured the sheet the way they did.
And it is unsupervised, so somebody else can sit it. The cheapest defence is not a webcam, it is one question at interview: ask the candidate to walk through how they got one specific answer. A person who did the work handles that in thirty seconds.
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.
Frequently asked questions
Is an online Excel assessment test fair if the candidate can look things up?
How long should a pre-employment Excel test be?
Which level should I send?
Can a candidate get someone else to take it?
Does it work across Excel versions and languages?
Tests that pair with this one
Accounting and bookkeeping skills test
Working accounting judgement: coding a transaction to the right account, reconciling a month that does not tie, and knowing which discrepancies to fix and which to flag.
Data entry assessment test
Accuracy transferring information between a source document and a structured form under mild time pressure.
Workplace numeracy test
Everyday work arithmetic: percentages, unit conversion, reading a table and checking a total.