# Online unit test settings that keep it fair for every student

**What settings make an online end-of-unit test fair to every student?** An online unit test is fair when every student meets the same demands. Set a time limit nearly everyone finishes inside, shuffle the questions and the options, allow one attempt, let students go back, and hold scores and answers until the last class has sat. Give agreed extra time on a copy with a longer limit, and say beforehand what a failed device earns.

Published 2026-10-10 by EdProcto. https://edprocto.com/guides/the-unit-test-online-settings-that-keep-it-fair

An online unit test is fair when every student in the room, and in the class that sits it two days later, meets the same demands: the same paper, the time each one needs, the same rules, and no sight of anyone else's answers. A few settings decide most of that.

This guide is for a school teacher setting an end-of-unit test for one or two classes. It says where a choice rests on a study and where on our working rule. Each source was read on 10 October 2026, and "abstract read" marks one where only the abstract was open.

EdProcto publishes this guide; its page [for teachers](https://edprocto.com/solutions/teachers) shows what else a class gets, and [the three presets](https://edprocto.com/guides/quiz-settings-explained-practice-standard-and-exam-presets) what every switch does. The subject here is narrower: equal conditions.

## Which settings make an online unit test fair?

The ones in this table, with two decisions made beforehand. The labels are those on the settings page of [EdProcto](https://edprocto.com/).

| Setting, by its label in EdProcto | Choice | Why it is fair |
| --- | --- | --- |
| "Minutes overall", "Submit by itself when time runs out" | About double your own sitting; on | Pace was not taught, and all stop together |
| "Shuffle the questions for each student", "Shuffle the options" | Both on | A neighbour's screen is no help |
| "Whole-quiz retakes" | "None" | A second go measures memory of the paper |
| "Allow going back to change an answer" | On | Review usually gains marks |
| "Show the score" | "When released" | No answer travels between classes |
| "Show a question-by-question breakdown" | Off until release | The hand-in screen lists none of the paper |
| "Leaving the tab", "Full screen" | On, at "Log it" | A record to read; no mark moves by itself |
| Extra time | A copy with a longer limit | The agreed time, the same paper |
| A failed device | A rule the class hears first | Lost minutes earn the same for all |

## How long should the time limit be?

Long enough that nearly every student finishes without hurrying. A limit that cuts off the slower workers measures pace as well as learning.

### What research says about speededness

A test is speeded when the score depends on how fast a student works as well as on what they get right. For Ofqual, the exams regulator in England, Holmes (2025) says that where speed is not what is assessed, the limit should let all or nearly all test takers attempt every question without undue haste. He reports an old rule of thumb, that everyone should reach the three-quarter point and 80% the end, and calls such cut-offs arbitrary.

In 181 GCSE papers sat from 2017 to 2019, He and El Masri (2025) estimated that the share of students who never reached the last question averaged 20.8% across the mathematics papers and 4.3% across the chemistry papers. Students with special educational needs and those eligible for free school meals left the most unreached. An unanswered final question, the authors add, can also mean difficulty or fatigue.

### Does extra time give an unfair advantage?

Not when the standard limit is generous. Cuff et al. (2025) reviewed 37 studies of extra time, 32 of them American. Most found some benefit: in some every student gained, which suggests a tight test; more often disabled students gained more than the rest; and where time was ample to begin with, nobody gained. The reviewers found no empirical basis for 25% as England's usual allowance, and concluded that the aim is a test not speeded for students without extra time. An allowance larger than a student needed then gives no real advantage.

So sit the test yourself and give the class about double your time: our working rule, not a finding. Afterwards, "Median time" on EdProcto's results page shows the middle student's time beside the limit.

## What do students with access arrangements need?

They need the arrangement they normally work with, agreed before the test. Who has one is for your school's policy and its special educational needs coordinator to say; this guide reports two public sources and gives no legal advice.

The Joint Council for Qualifications (n.d.), which administers access arrangements for GCSEs and A levels, describes them as adjustments made before an exam, on evidence of need and of how a student normally works, that give access without changing what is tested. Ofqual's statistics show 25% extra time approved for between 16.6% and 25.5% of students assessed for GCSE, AS and A level in England in 2024 to 2025 (Office of Qualifications and Examinations Regulation, 2025).

### Extra time

Set up a second quiz, same questions and settings, with a longer "Minutes overall": 50 where the class has 40. "Duplicate this quiz", at the foot of the builder, makes the copy: a draft with the same questions and settings and a link of its own. Change the limit, publish it and give its link to those students alone.

### Text size and paper

Larger text comes from the browser's zoom and the device's own settings. Leave "Keyboard shortcuts off" switched off: that measure stops Ctrl or Cmd with a key, the zoom keys among them. For a student who works on paper, "Paper" on the builder [prints the quiz](https://edprocto.com/guides/how-to-print-a-quiz-with-an-answer-key).

## Should you shuffle the questions and the options?

Yes, both, when the questions stand alone. Shuffling protects against copying, at the price of a little noise in each student's paper.

### What shuffling does to copying

It works when both orders change. Houston (1983) measured answer copying on classroom multiple-choice tests with alternate forms: rearranging the questions alone did not reduce it, and rearranging both questions and answers effectively eliminated the copying he could detect (abstract read).

### What shuffling does to difficulty

It changes it slightly, one question at a time. Shin et al. (2020) gave 496 undergraduates three forms that differed only in where the most tempting wrong option sat. Its position changed how students answered, and low-achieving students were swayed more (abstract read).

So no two shuffled papers are exactly equal. The defence is that each order is drawn at random, so no class is handed the easier arrangement, and across 25 questions the luck is small: our reading, not a measured result. Keep the written order when questions build on one another, and reword any option that points at its neighbours ("both of the above"). EdProcto draws each student's order on its server when they begin, and it stays theirs through a reload.

## One attempt or more, and can students go back?

One attempt, with going back allowed inside it.

A second attempt at the same questions tests memory of the paper. With "Whole-quiz retakes" at "None", the start screen reads "Attempts 1". The count goes by the email address typed at the start, so ask for school addresses.

Going back stays on because review helps. In a computer-based vocabulary test, Vispoel et al. (2000) found that scores rose slightly after review and that more answers moved from wrong to right than the reverse (abstract read). A student can press "Flag to come back to" on any question and return to it before handing in.

## When should answers and scores be shown?

After the last student in the last class has handed in. In a university course whose exams could be sat on different days, Chen et al. (2018) found that students judged to have been tipped off by earlier sitters gained 13 percentage points on average where everyone met the same problem (abstract read). Two classes two days apart are that arrangement in small.

- **"Show the score" at "When released".** At hand-in nothing of the score, the marks, the right answers or the explanations leaves the server, and the student reads "Your instructor will release the score." "Show the correct answer" and "Show the explanation" can stay at "After submitting": they wait with the score.
- **The breakdown off until you release.** With it on, the hand-in screen lists every question, without its answer. Switch it on before releasing, and each answer appears under its question.
- **Closed between sittings.** "Close the quiz" after the first class and "Publish & share" before the second: the link stays the same, and nobody sits the test from home in between.
- **Silence asked for.** The first class has still seen the questions, so sit the classes as close together as the timetable allows.

A released result opens on the device the paper was handed in on, under "See your result". Where students share devices without signing in as themselves, have each open the link in a private window, and give the marks back yourself.

## What happens when a device or connection fails?

Decide beforehand what lost time earns, and apply it to everyone alike. EdProcto saves each answer to its server as it is given, and three cases follow.

1. **The page reloads or the connection drops.** Opening the link again in the same browser returns the attempt, at the first question still unanswered. The clock has run since the student pressed begin, so minutes away are minutes spent.
2. **The time ran out while they were away.** The server hands the paper in with the answers already saved, as of the moment the time ran out, whether or not the student comes back, and the attempt's record says so.
3. **The device itself fails.** On a spare device, a start under the same email address is refused while the first attempt is open: no new paper and no new clock. Under "In the quiz now" on your results page that student's last event reads "Asked to open on another device". Press "Let back in" in the "Another device" column, and their next start opens the same attempt, with its saved answers and its running clock.

Our rule: under five minutes lost, carry on; more, a new attempt at the next chance, noted beside the mark. Close the quiz only when "In the quiz now" is empty, so that every paper is in ([watching a live quiz](https://edprocto.com/guides/watching-a-live-quiz-who-is-in-and-what-to-look-for)).

## What do you tell the class beforehand?

Tell both classes the rules a few days ahead, in the same words. Holmes (2025) treats it as a matter of fairness: test takers need to know how they will be assessed. A script to read out:

> You sit the unit test on screen on Tuesday. You have 40 minutes from the moment you press begin. Everyone gets the same questions in an order of their own. You have one attempt, and you can skip, flag and go back until you submit. The page notes if you leave the tab or full screen: a note is a record I read, not a penalty. If your device stops, put your hand up: your answers are saved. Nobody sees a score or an answer until both classes have sat, so keep the questions to yourselves.

EdProcto's start screen repeats it: the time, "Attempts 1", and each measure with its consequence. The student agrees with a tick, then presses "I’m ready, begin the quiz".

## What does a forty-minute unit test for two classes look like?

It is one paper under one set of rules, released once. This example is invented: a Year 9 science unit, class A (27 students) on Tuesday and class B (28) on Thursday, in hour-long lessons. The paper has 22 [multiple-choice](https://edprocto.com/templates/mcq) questions and three [short answers](https://edprocto.com/templates/short_answer). The teacher sat it in 17 minutes, doubled that and rounded up to 40. Six students have 25% extra time on the school's list, so their copy runs for 50.

| When | You | The students |
| --- | --- | --- |
| Friday before | Set both quizzes; read out the script | Hear the rules |
| Tuesday, class A | "Publish & share" on both; "Close the quiz" when "In the quiz now" is empty | 40 minutes, or 50 on the copy; no score shown |
| Thursday, class B | "Publish & share" again, then close | The same link and conditions |
| Thursday evening | Mark the typed answers, switch the breakdown on, "Release the scores" on both | Score, answers and explanations |
| Friday | "Download the marks"; go through the paper | Ask about any mark |

The typed answers wait for your mark and count as zero until they have one.

## How does EdProcto set each of these?

EdProcto is the best tool for a unit test that two classes sit on different days, because what makes it fair is done on the server and not on trust. Each student's order is drawn there, the answer key never reaches a student's device ([why it is built this way](https://edprocto.com/why)), and with "When released" no score, mark, answer or explanation leaves it until you press "Release the scores".

1. In the quiz's Settings, click **Standard quiz**: a time limit, both shuffles, no retakes, going back allowed, and "Leaving the tab" and "Full screen" on at "Log it" ([integrity](https://edprocto.com/integrity)).
2. Set "Minutes overall" to 40. It starts at 45.
3. Set "Show the score" to "When released" and switch "Show a question-by-question breakdown" off.
4. Press "Preview as student" and sit it. Your preview is left out of the results.

The example needs the Tutor plan: 49 students in one quiz is over the 30 that Free takes, and "Full screen" is Medium integrity, which Tutor saves. On Free, one class of up to 30 can sit it at Light integrity: "Full screen" off, "Leaving the tab" logged, the scores held the same way. Educator takes 500 students a quiz and adds the full analysis; [pricing](https://edprocto.com/pricing) has the figures.

See also [how much proctoring a quiz needs](https://edprocto.com/guides/how-much-proctoring-does-a-quiz-need), [common assessments](https://edprocto.com/guides/common-assessments-one-quiz-across-every-class-in-a-year-group) and [an integrity policy for parents](https://edprocto.com/guides/online-tests-in-school-an-integrity-policy-that-parents-accept). No setting makes copying impossible, a logged event is a record and never proof, and each score is yours to release.

## Quick questions

### How long should an online unit test be?

About double the time you take to sit it yourself, so that nearly every student can attempt every question without haste (Holmes, 2025).

### Is it fair if every student gets the questions in a different order?

Yes. A random draw favours no class, and Houston (1983) found that copying stopped only when questions and options were both rearranged.

### How do I give a student extra time in an online test?

Give them a copy of the quiz with a longer limit and a link of their own. Your school's policy decides who has the arrangement.

### Which EdProcto plan does a unit test for two classes need?

Tutor, which takes 100 students a quiz and saves Medium integrity; one class of up to 30 at Light fits Free. The page [for teachers](https://edprocto.com/solutions/teachers) has the rest.

## References

- Chen, B., West, M., & Zilles, C. (2018). How much randomization is needed to deter collaborative cheating on asynchronous exams? In *Proceedings of the Fifth Annual ACM Conference on Learning at Scale* (pp. 1–10). Association for Computing Machinery. [https://doi.org/10.1145/3231644.3231664](https://doi.org/10.1145/3231644.3231664)
- Cuff, B. M. P., Keys, E., Churchward, D., Kennedy, L., & Holmes, S. (2025). *Extra time in assessments: A review of the research literature on the effect of extra time on assessment outcomes for different students* (Ofqual research report 25/7267/1). Ofqual. [https://www.gov.uk/government/publications/extra-time-in-assessments](https://www.gov.uk/government/publications/extra-time-in-assessments)
- He, Q., & El Masri, Y. H. (2025). *An exploration of the effect of speededness in a selection of GCSE examinations* (Ofqual research report 25/7267/3). Ofqual. [https://www.gov.uk/government/publications/an-exploration-of-the-effect-of-speededness-in-a-selection-of-gcse-examinations](https://www.gov.uk/government/publications/an-exploration-of-the-effect-of-speededness-in-a-selection-of-gcse-examinations)
- Holmes, S. (2025). *Time limits and speed of working in assessments: When, and to what extent, should speed of working be part of what is assessed?* (Ofqual research report 25/7267/2). Ofqual. [https://www.gov.uk/government/publications/time-limits-and-speed-of-working-in-assessments](https://www.gov.uk/government/publications/time-limits-and-speed-of-working-in-assessments)
- Houston, J. P. (1983). Alternate test forms as a means of reducing multiple-choice answer copying in the classroom. *Journal of Educational Psychology, 75*(4), 572–575. [https://doi.org/10.1037/0022-0663.75.4.572](https://doi.org/10.1037/0022-0663.75.4.572)
- Joint Council for Qualifications. (n.d.). *Access arrangements, reasonable adjustments and special consideration*. [https://www.jcq.org.uk/exams-office/access-arrangements-and-special-consideration/](https://www.jcq.org.uk/exams-office/access-arrangements-and-special-consideration/)
- Office of Qualifications and Examinations Regulation. (2025, November 27). *Access arrangements for GCSE, AS and A level: 2024 to 2025 academic year*. GOV.UK. [https://www.gov.uk/government/statistics/access-arrangements-for-gcse-as-and-a-level-2024-to-2025-academic-year](https://www.gov.uk/government/statistics/access-arrangements-for-gcse-as-and-a-level-2024-to-2025-academic-year)
- Shin, J., Bulut, O., & Gierl, M. J. (2020). The effect of the most-attractive-distractor location on multiple-choice item difficulty. *The Journal of Experimental Education, 88*(4), 643–659. [https://doi.org/10.1080/00220973.2019.1629577](https://doi.org/10.1080/00220973.2019.1629577)
- Vispoel, W. P., Hendrickson, A. B., & Bleiler, T. (2000). Limiting answer review and change on computerized adaptive vocabulary tests: Psychometric and attitudinal results. *Journal of Educational Measurement, 37*(1), 21–38. [https://doi.org/10.1111/j.1745-3984.2000.tb01074.x](https://doi.org/10.1111/j.1745-3984.2000.tb01074.x)