A candidate needs four things from an exam interface. They need to know where they are, move freely between questions, park a hard one, and see what is unanswered before submitting. The last one is a setting somebody has to switch on.
None of that is decoration. Each one changes what the final mark means.
What does an exam interface actually decide?
It decides how much of the score is about the subject.
Every exam measures two things at once. It measures what the candidate knows. It also measures how well they coped with the way the material was presented. Only the first one is meant to be there. Psychometricians call the second construct-irrelevant variance: a score that changes for reasons unrelated to what is being tested.
An interface is one of the larger sources of it. Time spent working out where you are is time not spent answering. A candidate who cannot find question 14 again loses minutes. A candidate who cannot park a question spends the rest of the paper half-thinking about it.
None of this shows up in the result. The mark sheet says the candidate got it wrong. It does not say why.
That is the cost, and it arrives at appeal. A candidate who ran out of time will say the system slowed them down. You need to know whether they are right. If the interface made the paper harder for everyone, the exam is still comparable, but the pass mark is wrong. If it made the paper harder for some candidates, comparability is gone.
Most of these decisions are made on a settings screen by somebody who will never sit the paper. Often they get made by leaving a default alone.
Can candidates move freely between questions?
This is the first decision, and it is bigger than it looks.
A linear exam shows one question after another, with no way back. It is simple to build and easy to reason about. It also punishes a specific kind of candidate: the one who reads ahead, answers the easy questions first, and returns to the hard ones. That is a taught exam technique. A linear interface makes it impossible.
Free navigation is the alternative. Every question is reachable at any point, from a list that stays on screen.
That list does more work than it appears to. It answers three questions the candidate would otherwise carry in their head:
- How many questions are there?
- Which ones have I answered?
- Which ones did I want to return to?
Holding those three facts in working memory costs something. Putting them on the screen gives it back.
There is one detail worth asking any vendor about directly. When a candidate jumps to another question, does the answer they were typing finish saving first? StudyDrome Exam Manager waits for that save before moving them to the list. A system that does not wait can drop the last thing a candidate wrote, and the result will look exactly like a question they never attempted.
What does flagging a question actually do?
First, a disambiguation, because the word does two jobs in this industry. An invigilator tool flags a candidate for a possible integrity violation. That is not this. Here, flagging is something a candidate does to their own paper, and nobody else needs to see it.
It is the digital version of folding a corner. The candidate is unsure, wants to keep moving, and wants to find it again.
The reason it matters more than it looks is that an unresolved question does not leave the mind. The candidate keeps a low-level process running on it while answering the next four. Marking it externally closes that loop. They can stop holding it.
For a flag to be worth anything, it has to survive:
- It should appear in the question list. A candidate needs to see every parked question at a glance, not hunt for them.
- It should be counted at the end. "Three flagged for review" is a usable instruction. "Go back through and check" is not.
- It should come off as easily as it went on. A flag a candidate hesitates to remove stops being a working tool.
Better items reduce how often it gets used. Ambiguity is what makes a candidate hesitate, and flawed items generate far more second-guessing than clear ones. No bank is perfect, so the flag still earns its place.
Where should the reading sit?
Some questions arrive with something to read: a case, a set of results, a scenario. Often several questions share it.
If the material and the question sit in different places, the candidate scrolls. They scroll up to read, back down to answer, then up again to check a number. Every trip costs attention. The cost repeats for every question that shares the material.
Keeping the material beside the question removes the trip. Two consequences follow, and both are easy to miss.
Questions that share material have to stay together. If the exam shuffles question order, the shared case must travel with its group. Otherwise the candidate meets the same case three times, in three places, and reads it three times.
Images have to be readable at exam resolution. A scan or a diagram that is legible on the author's monitor may not be legible on a laptop in a hall. Whether a candidate can enlarge it is usually a setting, and it is worth checking rather than assuming.
The interface also has to suit the item. A matching question, an essay and a fill-in-the-blank are three different interactions, and each needs its own affordances to be obvious without instructions. It is worth looking at the question types an exam can be built from before deciding an interface handles them all equally well.
What should a candidate see before they submit?
Submitting is irreversible, and it happens at the moment the candidate has the least information.
A review step closes that gap. Before the final confirmation, the candidate sees a summary of how many questions were answered, which were left blank, and which are still flagged.
Two decisions sit inside that step.
Blank answers: warn, or block? Blocking guarantees a complete paper. It also removes a legitimate choice because a candidate may leave one blank. Warning respects the choice but lets genuine oversights through. Both are defensible. Not deciding is not.
Whether the step exists at all. It is usually a setting rather than a given. In our own system, it is off until somebody turns it on, and that is worth knowing before exam day rather than after.
What a candidate sees after submitting is a separate decision with separate consequences. Showing the full paper back helps candidates learn and undermines item security. The exam office view works through the levels and who gets to change them.
What about candidates who need the interface to adapt?
Some candidates need it to behave differently, and this is a procurement question rather than a design one. Four things are worth asking any vendor.
Ask for the accessibility statement and the conformance level it claims to meet. WCAG 2.2 AA is the level public bodies across the UK and EU are generally measured against. A statement that names its own gaps is more useful than one that claims none.
Ask which screens the statement covers. A statement may cover the candidate-facing exam and stop there, leaving the staff-facing reports unassessed. The scope of the claim matters as much as the level.
Ask how extra time is granted. There is a real difference between an allowance declared in advance and attached to a candidate, and minutes added by a member of staff on the day. Both can be legitimate. Only one of them works when that person is unreachable at 09:15.
Ask what a candidate does when something fails mid-exam. The answer should name a person and a route, not a feature.
The settings that decide this for you
Most of what a candidate experiences is fixed before the exam opens. These decisions are usually made once and then inherited by every exam afterward. That is fine when somebody chose them. It is not fine when they are just the defaults.
Setting | What it decides for the candidate | Ask before you publish |
|---|---|---|
Review before submitting | Whether they see what is blank and flagged, or submit blind | Is it on, and do we block on blanks or warn? |
Marks shown per question | Whether they can budget time against what a question is worth | Do we want time spent in proportion to marks? |
Results after submitting | What they see the moment they finish: nothing, a score, or the paper | Does this exam reuse its items? |
Image enlargement | Whether a diagram can be read at all on a small screen | Do any of the images here contain the answer? |
Private notes | Whether working out a calculation has to happen in their head | Does this subject involve multi-step reasoning? |
Question and answer order | Whether two candidates see the same paper, and whether shared material stays with its group | Are we shuffling for integrity, and have we checked what it does to grouped items? |
All questions required | Whether a blank is a choice or an error | Is leaving one blank ever legitimate here? |
Integrity controls | How much of the screen is policing them rather than helping them | What does each one cost the honest candidate? |
That last row deserves its own sentence. Every integrity control you switch on takes something from both the honest and the dishonest candidate. Blocking copy stops a question being pasted into a chat window. It also stops a candidate from copying a figure from a table into their own work. Neither effect is hypothetical, and only one of them was the point. A vendor who says that out loud is being straight with you.
Frequently asked questions
Should candidates be able to go back to a previous question?
In most exams, yes. Reading ahead, answering the easy questions first and returning to the hard ones is a taught technique, and a linear interface makes it impossible. Some systems can lock the order, which is defensible when a later question would reveal an earlier answer. That is a property of the paper, so decide it on an exam-by-exam basis rather than once for everything.
What does "flag for review" do in an online exam?
It lets a candidate mark a question to come back to without answering it or telling anyone. The flag is private to them. A useful version shows in the question list, so parked questions are visible at a glance, and gets counted before submission so none is forgotten. It is a different thing from an integrity flag raised against a candidate.
How many questions should be shown on one page?
One, in most cases. A single question with the navigation beside it keeps attention on one decision at a time. Long scrolling pages make it easy to accidentally skip an item and hard to tell what is left. The exception is a group of short items that share one piece of reading, where splitting them across pages forces the candidate to reread it.
Is there parity between the mobile and desktop experience?
Ask, because it varies. A browser-based exam will usually run on a phone or tablet, but running and being usable are not the same. The question list and any shared reading need somewhere to go on a narrow screen, normally by collapsing rather than disappearing. Sit a full exam on the smallest screen you intend to support before you tell candidates it works.
Can you set time limits and accommodations for candidates who need them?
Time limits are standard. Accommodations need a closer look. Ask whether an allowance can be declared in advance and attached to the candidate, or whether somebody has to grant it on the day. Ask for the accessibility statement, the conformance level it claims, and which screens that claim covers. Get those answers in writing during procurement.
Should candidates see their marks while they are taking the exam?
It depends on whether you want time budgeted by value. Showing marks helps a candidate spend longer on a question worth ten than one worth one. It also puts a number on every screen, which some candidates find stressful. If the paper weights every question equally, showing marks adds noise and no information.
Where to go next
Reviewing a platform rather than fixing a policy? Three questions separate a considered interface from an assembled one. Can a candidate move freely between questions? Is there a review step before submission, and is it on? And what happens to a half-typed answer when they navigate away from it?
Those three questions have concrete answers on one system, screen by screen, in the candidate experience.
The mechanics behind a written exam, from authoring through to what an invigilator sees, are in written exams. What happens when the connection drops is a separate problem with its own answers, and it is covered in losing internet mid-exam.