Products › MCQs
The one score nobody
can argue with.
Knowledge checks with a declared set of right answers
Four of five correct is four of five correct. An MCQ round is the one part of the pipeline where no model forms a judgement — the answer set is declared before anybody takes it, and the score is arithmetic.
Any device, any connection. Results the moment they finish.
A bank built, a round taken, a score the moment it ends.
No model decides
anything here.
Every other round on this site involves a model forming a view, which is why they all carry a “a human decides” caveat. This one does not. You declare the correct options, the candidate picks, and the result is counted.
The answer set is declared up front. Scoring is arithmetic rather than judgement — which means there is nothing to appeal, and nothing to explain to a candidate who asks how they were marked.
Results are instant. No review queue, no scoring pass, no waiting for a panel. A candidate finishes and the score exists.
Drawn from your own banks. Reusable question sets grouped into sections, so a round can test three areas separately rather than producing one undifferentiated number.
Questions and options can be randomised, so two candidates sitting the same round at the same time are not looking at the same screen in the same order.
Standalone or embedded. Run it as a first filter before anybody spends time on a written round, or mix multiple-choice items into a longer conversational one.
Right for a threshold, not a ranking. Certifications, regulated knowledge, safety procedure, product detail — the places where a floor genuinely exists.
below the 60% floor you set for this section
Scored on submit · no model involved · section floors are yours to set
Build the bank once. Then it just runs.
Five stages, and after the first one nobody touches it. Invitations send themselves and results appear on submit.
What the MCQ round does
A declared answer set, and why that matters
The correct options are recorded before a single candidate sits the round, so the result is a count rather than an assessment. That has a practical consequence worth naming: this is the only round in the suite where a candidate asking “how was I marked” gets an answer you can show them in full, without discussing rubrics, models or reasoning.
Question banks, sections and randomisation
Reusable banks grouped into sections, so a round tests regulatory knowledge, product detail and process separately and reports each one. Questions and their options can be randomised, so two candidates sitting simultaneously are not reading the same screen in the same order — a modest control, honestly described.
Instant results and the floors you set
Scored on submit, with no review queue. You can set a floor per section rather than only an overall pass mark, which is usually what you actually want: somebody strong overall but below the line on the regulated section is a different decision from somebody uniformly average.
Where MCQs are the right instrument — and where they are not
They are the right instrument for a threshold: does this person know the regulated thing, the safety procedure, the product. They are a poor instrument for a ranking, because the gap between 13 and 15 out of 15 tells you very little about who will do the job better. Use it to filter, then use a written or spoken round to choose.
Mixing into a longer round
Multiple-choice items can sit inside a conversational or timed round rather than running separately, which is often the better shape — a few factual checks in among the behavioural questions, in one sitting, without asking a candidate to complete two separate things.
Pipelines and reporting
Results land on the candidate record and the pipeline like any other score, and because they are deterministic they are the cleanest thing in your reporting — useful for showing how a cohort performed against a knowledge requirement over time.
It is a knowledge floor,
not a knowledge test
This round is the most honest and the most limited thing in the suite, and both facts matter equally.
Scoring is arithmetic
The answer set is declared before the round exists. There is no model judgement, no rubric interpretation, and nothing for a candidate to appeal.
The threshold is your decision
The software counts. What counts as a pass, per section and overall, is set by you and recorded — so a decision can be explained months later.
Deterministic by design
Same answers, same score, every time. It is the one part of the pipeline where the result cannot drift between two runs.
And plainly, the big one. There is no proctoring. A candidate sitting an MCQ round at home can look things up, and for most questions a model will answer them correctly in seconds. So treat this as a knowledge floor — evidence somebody could find and apply the right answer — rather than proof they knew it unaided. If unaided recall is genuinely what you need to measure, that is what Timed Q&A is for, and even then a clock is a constraint rather than a guarantee. MCQs also measure recall rather than judgement, so they should filter a pipeline and never choose from one.
Frequently asked questions
How is the multiple-choice round scored, and does an AI mark it?
No AI marks it. You record the correct options before any candidate sits the round, the candidate picks answers, and the score is arithmetic. TalbotIQ describes this as the one part of its pipeline where no model forms a judgement.
How much work is a multiple-choice knowledge test to set up and run?
Setup is five stages, and after the first one nobody touches it. Invitations send themselves and results appear when candidates submit.
Can I set a separate pass mark for each section of the test?
Yes. You can set a floor per section rather than only an overall pass mark, so someone strong overall but below the line on the regulated section is treated as a different decision from someone uniformly average. Scoring happens on submit with no review queue.
Is the test proctored, or can candidates look up answers at home?
There is no proctoring. A candidate at home can look things up, and for most questions a model will answer correctly in seconds, so TalbotIQ tells you to treat the result as a knowledge floor (evidence someone could find and apply the right answer) rather than proof they knew it unaided.
When is a multiple-choice test the wrong way to assess job candidates?
When you need to rank them. The gap between 13 and 15 out of 15 tells you very little about who will do the job better, so use MCQs to filter for a knowledge threshold and then use a written or spoken round to choose between the people who pass.
Can multiple-choice questions be combined with a chat or timed interview instead of running as a separate test?
Yes. Multiple-choice items can sit inside a conversational or timed round, so a few factual checks run alongside the behavioural questions in one sitting without asking the candidate to complete two separate things.
Filter on what is knowable.
Book a 30-minute demo. Bring the knowledge requirement and we will build the round in front of you.
Malaysia-based team · Response within one business day · No obligation