How to Set a Passing Score (Cut Score Methods)

To set a passing score, do not pick a round number like 70%. Convene a small panel of subject-matter experts and use a standard-setting method such as the Angoff method: for each question, the experts estimate the share of minimally qualified candidates who would answer it correctly, then you average and sum those estimates to get a defensible cut score tied to the actual difficulty of your exam.

A passing score decides who is qualified and who is not, so it carries real weight, especially on a certification, licensure, or compliance exam. Set it too low and you pass people who are not ready. Set it too high and you fail competent candidates and invite appeals. This guide explains how exam professionals actually set a cut score, what the common methods are, and why a fixed 70% is usually the weakest choice. It is written for certification bodies, training and L&D teams, and instructors who need a passing score they can defend.

How do you set a passing score?

You set a passing score by linking it to the level of knowledge a competent candidate must show, not to a convenient percentage. The standard approach is to gather subject-matter experts, define what a minimally qualified candidate looks like, and have those experts judge each question against that definition. Their combined judgments produce a cut score that reflects the specific content and difficulty of your exam form rather than an arbitrary number.

The process has a few repeatable steps. First, write a clear description of the borderline candidate: someone who barely has enough competence to pass. Second, assemble three to eight experts who know the role. Third, walk through every question and collect each expert's estimate. Fourth, average the estimates and sum them into a recommended cut score. Fifth, review the result against how real candidates perform and document the rationale so the score holds up if anyone challenges it.

What is the Angoff method?

The Angoff method is the most widely used way to set a cut score on certification and licensure exams. A panel of experts looks at each question and estimates the probability that a minimally qualified candidate would answer it correctly, on a scale from 0 to 1. You average each item's estimates across the panel, then add those averages across all items. The total is the recommended passing score.

An example makes it concrete. Say a 100-question exam has expert estimates that average out to 0.68 per item. Summed across 100 items, that gives a recommended cut score of 68 correct answers, or 68%. Notice that the number came from the exam's real difficulty, not from someone deciding 70% sounded right. The yes/no variant of Angoff simplifies the judgment to a single call per item (would the borderline candidate get this right, yes or no), which speeds up the panel while keeping the logic intact.

What is a good passing score for a certification exam?

There is no single good passing score, because the right cut score depends on the difficulty of the questions and the competence the role demands. A defensible passing score is one produced by a standard-setting study, and it varies by exam. Many real certifications also report scaled scores rather than raw percentages, so the underlying pass mark is not a simple 70%.

For perspective, several well-known certifications use scaled passing scores: CompTIA A+ Core 1 requires 675 on a 100 to 900 scale, Network+ requires 720, and Microsoft certification exams pass at 700 or above. Those numbers come from standard-setting work, not from a fixed percentage. The lesson for anyone building their own exam is to set the bar against competence and document how you got there.

How many questions do you need to pass?

The number of questions you need to pass depends on the cut score and how the exam is scored. On an exam with a raw cut score, you pass by answering at least that many items correctly, for example 68 of 100. On a scaled-score exam, the exact number varies because harder questions can carry different weight, so there is no fixed count you can miss and still pass.

This is exactly why a flat percentage is shaky. If two versions of an exam differ in difficulty, the same 70% rule passes more people on the easy form and fewer on the hard form, even though the candidates are equally competent. A cut score set per form keeps the bar fair across versions. If you build multiple versions of an exam, run a quick standard-setting check on each form rather than reusing one percentage.

Should the passing score be a fixed percentage like 70%?

A fixed percentage like 70% is the most common choice and usually the weakest one. It ignores the difficulty of individual questions and the competence the profession actually requires. On an easy exam, 70% can pass candidates who are not ready; on a hard exam, it can fail people who are. A round number feels objective but has no link to what the score is supposed to measure.

That said, a fixed percentage is not always wrong. For low-stakes classroom quizzes or internal knowledge checks, the cost of a standard-setting study outweighs the benefit, and a sensible default like 70 or 80% is fine. Reserve formal cut-score methods for exams where the result matters: certifications, licenses, compliance sign-offs, and any test that gates a job or a credential.

What is the difference between a cut score and a passing score?

A cut score and a passing score are essentially the same thing: the point that separates passing from failing. The phrase cut score is the technical term used in psychometrics and standard setting, while passing score is the everyday term candidates see. The small distinction is that cut score usually refers to the threshold on the score scale (raw or scaled), and passing score is how that threshold is communicated.

In a criterion-referenced exam, the cut score is fixed in advance and does not move based on how the group performs. It does not matter whether everyone passes or only a few do; the cut score is the cut score. That is the opposite of grading on a curve, where the bar shifts with the cohort. Certification and licensure exams are criterion-referenced for exactly this reason.

Can AI help create the exam you set a passing score on?

Yes. AI handles the part of the work that comes before standard setting: writing a deep, well-aligned pool of questions and a clean answer key. With PDFQuiz, you upload your source material (a study guide, training deck, policy manual, or PDF) and the tool drafts multiple-choice, true/false, short-answer, and other question types at the difficulty level you choose, then exports a printable PDF or editable Word file with the answer key.

What AI does not do is set the cut score for you. The passing score still needs human judgment from people who know the role, because it is a decision about competence, not a calculation a model should make on its own. Use PDFQuiz to build and refine the question bank fast, then run your Angoff panel on that exam form to set a passing score you can defend. To plan how many items each topic should carry before you set the bar, start with a test blueprint, and to build the exam itself see how to create a certification exam.

Set a defensible passing score on a stronger exam

A passing score is only as good as the exam under it. If the questions are vague or poorly aligned, no cut-score method will save the result. Build a solid question pool first with an AI certification exam generator or the exam creator, generate practice forms with the practice test generator, and draft questions from any document with the AI test generator. Then convene your experts, run an Angoff study on the finished form, and document the cut score. That sequence (strong questions, then a standard-setting study) gives you a passing score that holds up under scrutiny.