Updated 15 November 2026 · About 9 minutes to read
“80% to pass” is a habit, not a method. A pass mark should describe the score a person who is just good enough would get onthis test. Here are three practical ways to set one for workplace and course tests, and how to check it afterwards.
Start with the purpose
Ask what a pass is supposed to mean:
- “Has read and understood the essentials” — common for onboarding and annual refreshers. A high mark on a short, focused test fits; people can usually retake it.
- “Can be trusted to do the task” — safety procedures, equipment use. Some questions may be must-pass items; consider a separate short test for those.
- “Is among the stronger candidates” — hiring screens. Here a fixed pass mark matters less than ranking; you may not need one at all.
Method 1: the borderline person (a simple Angoff)
This is the most widely used approach for setting a cut score from content, and it works with a small group.
- Get two to five people who know the job and the material.
- Together, describe the borderline person: someone who is just barely good enough. Write two or three sentences — “knows where the extinguishers are and which to use; may hesitate on rarer cases”.
- For each question, each judge estimates the probability (0–100%) that a borderline person gets it right.
- Average the estimates for each question, then average across questions. That average is your pass mark.
Example. Three judges rate a 10-question test. Their per-question averages are 90, 85, 60, 75, 80, 95, 70, 65, 85, 75. The mean is 78%, so the pass mark is 78% — not the 80% you might have picked by habit.
If questions carry different points, weight each question's estimate by its points.
Method 2: must-know items plus a total
Some knowledge is non-negotiable (“never block a fire exit”). Averaging lets someone miss it and still pass. Options:
- Put the must-know questions in their own short test with a 100% pass mark and unlimited attempts, and the broader material in a second test with a lower mark.
- Give must-know questions more points so missing one costs more.
Method 3: compare groups (contrasting groups)
If you already know who is competent — for example experienced staff versus new starters — give both groups the test. Plot their scores; the pass mark goes where the two distributions cross. This needs real data (20+ people per group is a sensible minimum) but is very convincing.
Adjust for guessing
Guessing raises scores on choice questions. On a test of four-option single-choice questions, someone who knows nothing averages 25%. A pass mark of 50% on such a test means less than it sounds. Ways to reduce the effect:
- Use fewer true/false questions (a guess is right half the time).
- Use multiple select with all-or-nothing scoring, or short text and numeric questions where guessing barely helps.
- Use more questions: random luck evens out over 20 questions far more than over 5.
Account for measurement error
On a short test, a person's score could easily differ by several points on a different day or a different draw from a pool. Two consequences:
- Don't treat 79% and 80% as meaningfully different. If the decision matters, look at people within a few points of the line more closely, or let them retake.
- With pools, make sure each pool's questions are of similar difficulty, so one person's draw isn't harder than another's (see Question pools explained).
Check it after the first run
Once 20–30 people have taken the test, open the results:
- Pass rate. If almost everyone passes, either training worked well or the test is too easy. If most fail, check the questions with the lowest % correct before blaming the people.
- Questions with negative discrimination. These are questions strong performers get wrong. They pull good people below the line unfairly — rewrite or drop them, then regrade.
- Borderline cases. Read a few attempts just below the mark. Would you be comfortable saying these people aren't ready? If not, your pass mark is probably too high.
Write it down
Record how you set the pass mark, who was involved and the date. If anyone disputes a result, “three supervisors rated each question for a borderline employee and averaged the ratings” is a much better answer than “we always use 80%”.
Setting it in ExamKit
Each test has its own pass mark (0–100%). Publishing freezes it with the questions, so changing it later creates a new version and doesn't change results people already have. If essays are involved, pass/fail is decided once a grader has marked them; certificates are issued only to attempts whose final score passes.
ExamKit is meant for training, course and skills tests. Don't use it for licensing or regulated exams, which need formal standard-setting and supervision.