# How to write an exam blueprint for a certification

**How do I write an exam blueprint for a certification exam?** An exam blueprint is a table that sets the content areas of a certification against what a candidate must do with each one (recall, apply, judge), with a weight and a number of questions in every cell. Take the weights from what practitioners do and how much it matters, convert them to counts, publish the table, and check every paper against it.

Published 2026-10-09 by EdProcto. https://edprocto.com/guides/how-to-write-an-exam-blueprint-for-a-certification

An exam with no blueprint tests whatever its writers happened to think of. One examiner is strong on equipment rules, another has a favourite scenario, and the paper comes out heavy where the committee was keen and thin where the work is hard. Candidates notice, and so do the members who ask what the certificate proves.

An exam blueprint settles it in advance, as a table. Down the side run the content areas. Across the top is what a candidate must do with each: recall it, apply it, or judge with it. Each cell holds a weight and a number of questions. Assessment texts also say test blueprint, test plan or table of specifications.

It is the step before [setting the pass mark](https://edprocto.com/guides/setting-a-pass-mark-you-can-defend), and it falls to the bodies EdProcto serves on its page [for associations and boards](https://edprocto.com/solutions/associations). EdProcto publishes this guide. The blueprint is your committee's decision, and nothing here is legal advice.

## What is an exam blueprint, and what does it protect?

An exam blueprint is the written plan of a test: the content to be covered, the emphasis each content area gets, and the test's other main features. Raymond and Grande (2019), whose definition that is, build one in four stages: identify the major domains of knowledge and skill, set out the objectives to be assessed, choose the method of assessment, and fix the emphasis each domain receives.

The table protects content validity, the claim that the paper samples the field the certificate names. Sireci and Faulkner-Bond (2014) note that evidence based on test content is one of five kinds of validity evidence in the Standards for Educational and Psychological Testing, and conclude that a test of knowledge or skill needs it to build the validity argument for a given use. Its absence has a name: Downing and Haladyna (2004) call it construct under-representation, where the content domain is undersampled and a score cannot be generalised to the whole of it.

### What the studies show, and what they do not

The evidence is supportive and modest. DiDonato-Barnes et al. (2014) assigned 53 undergraduates at random to be taught the table of specifications or not, then had each build a test for the same unit. The taught group's tests showed stronger evidence of content coverage. Its members picked questions on the right subject matter but struggled to pick them at the thinking level the table specified.

Patil et al. (2015) describe a pathology department that blueprinted its written papers, weighting topics by impact on health and by how often the condition occurs. Of 139 students who gave feedback, 77% judged the spread of questions across topics appropriate and 86% called the examination fair. There was no comparison paper.

Nothing we read tested whether a blueprint changes who passes a certification exam. The case for one is an argument about sampling.

## Where do the weights in a blueprint come from?

The weights come from what practitioners actually do and how much each part matters. The study that finds this out is a practice analysis, or job analysis. Raymond (2001) calls it the primary source of evidence for the validity of scores on licensing and certification exams, and adds that specialists agree surprisingly little on how to run one or turn its results into a test plan. So there is no single procedure to copy, only sound ones to choose and record.

In its full form, a committee lists the tasks of the role and a large sample of practitioners rates each, typically for how often it is done and how serious it is to get wrong. Kane et al. (1989) make the case for multiplying frequency by criticality, with adjustments so that each rating carries the influence intended. Rating both protects the rare, grave task: a coach may meet a head injury once in ten years, and frequency alone would drop it from the paper.

Kane (1997) places this study at the head of a chain that runs through the test specifications, the test and the candidate's performance to the pass or fail decision. A careful pass mark does not repair a weak first link. His shortcut: draft a model of the practice first, a provisional blueprint, and use the study to test it.

### A lighter version for a small body

A small body can use a panel, a short survey, or both. Raymond (1996) compared the routes in four medical imaging specialties. On one, experts each proposed a percentage of questions for every content category, then settled the figures by discussion and consensus. On the other, weights came from a statistical model fed with practitioners' ratings of frequency and criticality. The two agreed moderately to highly for sections made up of specific procedures, and much less for sections on general topics.

So a panel is a reasonable source of weights where the content areas are concrete tasks. Where they are broad, such as conduct or communication, check it against a survey. A routine for a small federation:

1. Draft the content areas from the syllabus or rulebook, each named as something a practitioner does.
2. Ask every certified member two questions about each area: how often do you do this, and how serious is a mistake in it?
3. Convene a panel from across the membership. Each member proposes a percentage for every area alone, sees the survey averages and the others' figures, and the panel settles the weights.
4. Record the ratings, each round of figures, the final weights and who took part.

## How do you turn weights into question counts?

Multiply each area's share of the total weight by the length of the paper, round so the counts add up exactly, then apply any minimum the committee has set. The example is an invented certificate for community sports coaches with a 60-question paper. The ratings, on four-point scales, are made up to show the arithmetic.

| Content area | Frequency | Criticality | Product | Of 60 | Rounded |
| --- | --- | --- | --- | --- | --- |
| Safeguarding | 2 | 4 | 8 | 11.43 | 11 |
| Session planning | 4 | 2.5 | 10 | 14.29 | 14 |
| Coaching technique | 4 | 3 | 12 | 17.14 | 17 |
| Rules of play | 3 | 2 | 6 | 8.57 | 9 |
| Injury response | 1 | 4 | 4 | 5.71 | 6 |
| Record keeping | 2 | 1 | 2 | 2.86 | 3 |
| Total | | | 42 | 60.00 | 60 |

Take the first row. Its product, 8, is 19.05% of the total of 42, and 19.05% of 60 questions is 11.43.

### Rounding so the total is exact

Rounding to the nearest whole number sometimes gives 59 or 61. One rule always lands on the total: round every figure down, then give the spare questions, one each, to the largest remainders. Here the figures round down to 11, 14, 17, 8, 5 and 2, which makes 57, and the three spare questions go to the remainders of .86, .71 and .57.

### A minimum for each area

A floor is a judgement, not a finding: nothing we read sets one. Its purpose is that an area tested by two or three questions is barely tested. This committee set four, so record keeping rises from 3 to 4 and the largest area, coaching technique, gives one up. The alternatives are to merge a small area into a neighbour or to drop it and say so.

### The second dimension: recall, apply, judge

Split the columns the same way. The committee decides what share of the paper asks for recall (stating a rule or a limit), application (using it in a described situation) and judgement (choosing between defensible actions). Here it chose 30%, 50% and 20%: 18, 30 and 12 questions. Each row is then divided so that rows and columns both add up.

| Content area | Recall | Apply | Judge | Questions |
| --- | --- | --- | --- | --- |
| Safeguarding | 3 | 5 | 3 | 11 |
| Session planning | 4 | 8 | 2 | 14 |
| Coaching technique | 4 | 9 | 3 | 16 |
| Rules of play | 3 | 4 | 2 | 9 |
| Injury response | 1 | 3 | 2 | 6 |
| Record keeping | 3 | 1 | 0 | 4 |
| Total | 18 | 30 | 12 | 60 |

Write a model question for each level before anyone classifies anything. In the DiDonato-Barnes experiment, level was what people got wrong.

## Which question shape fits each cell?

Choose the shape by the column: what the candidate must do decides how the question is asked.

| Column | The candidate | Shapes among EdProcto's [templates](https://edprocto.com/templates) |
| --- | --- | --- |
| Recall | States a rule, a term or a limit | [Multiple choice](https://edprocto.com/templates/mcq), [short answer](https://edprocto.com/templates/short_answer) |
| Apply | Uses the rule in a described situation | A [case study](https://edprocto.com/templates/case), [show your working](https://edprocto.com/templates/show_work) |
| Judge | Chooses between defensible actions | [Best and worst](https://edprocto.com/templates/best_worst), [watch and answer](https://edprocto.com/templates/video_q) |

All six are in the builder today. Shape does not fix level: a multiple-choice question can ask for a remembered limit or for the right call in a new situation. The stem decides, and its rules are in [how to write multiple-choice questions that measure](https://edprocto.com/guides/how-to-write-multiple-choice-questions-that-measure).

## How do you use the blueprint to commission, review and build papers?

Use it three times: as the order form for questions, the review sheet, and the assembly list for each form.

1. **Commission by cell.** Ask each writer for named cells, such as three apply questions on injury response, and for spares, because review will reject some. On EdProcto's [home page](https://edprocto.com/), Procto AI drafts questions from the part of your syllabus that covers a cell, and every draft is yours to edit.
2. **Label every question** with its area and level. Each EdProcto question carries a topic: use the area's name.
3. **Review against the cell.** A second person, who has not seen the label, says which area and level the question tests. The model questions settle disagreements.
4. **Build the form by the counts,** cell by cell from approved questions, by hand in a spreadsheet.
5. **Build a parallel form the same way,** from different questions. One blueprint gives the same coverage, not the same difficulty, so each form's pass mark is set on its own questions.

### Should candidates see the blueprint?

Yes: publish the areas, the weights and the levels, never the questions. McLaughlin et al. (2005) compared students' views of a renal course examination before and after its blueprint was published, with no significant change in the objectives, the blueprint or the examination itself. Scores did not differ significantly, and a significantly greater share of students judged the examination a fair test of the important subject matter. The authors note that how much to publish is disputed.

## How do you check a finished paper and keep the blueprint current?

Count the finished paper against the table before anyone sits it, and review the table on a fixed date and when practice changes.

**Before the sitting.** List every question with its area and level, and total each cell. On a 60-question paper the counts should match exactly; any difference is recorded with its reason and approved by the committee.

**After the sitting.** EdProcto's [results](https://edprocto.com/#results) give, in the full analysis, a mean score for each topic, over the questions that carry it. An area far below the others is a prompt: check its questions first, then the syllabus and the training. Read area means across the cohort: for one candidate, an area of four questions moves in steps of 25 points.

**Over time.** Nothing we read fixes a review interval. Put a date in your rules and add triggers: a new rulebook, a change in the law or the equipment, a new duty for the role. Date every version, and announce changes before the sitting they apply to.

## How does EdProcto fit a blueprinted exam?

EdProcto is the best tool for the two jobs on either side of a blueprint: writing questions from the body's own standard, and reporting scores by content area. A blueprint is a plan for sampling one field, so questions drawn from anywhere else defeat it.

**Writing.** Give Procto AI your rulebook or syllabus and it writes questions from that text in whichever of the 111 shapes a cell calls for. You can edit every one, and it never releases a score.

**Sitting.** Under the Exam preset a candidate cannot go back and sees nothing until you release it, with every browser measure on. Candidates are told beforehand which [integrity](https://edprocto.com/integrity) measures apply. The pass mark is a single setting.

**Different papers.** A different set of questions for each candidate, which Procto AI writes on the same concepts, is the next step EdProcto is building on the same engine. That is not two reviewed forms: a committee that approves every question before it is sat builds its forms by hand, in its own spreadsheet.

Free has the core templates, rule-based marking and question-by-question results. Tutor has every template. Educator brings the full analysis and certificates. An Institution licence starts with a call: [pricing](https://edprocto.com/pricing) and the page [for schools](https://edprocto.com/schools) explain it. EdProcto has no tie to any board or federation.

## Quick questions

### What is the difference between a syllabus and an exam blueprint?

A syllabus lists what a candidate should learn. A blueprint says how much of the paper each part receives and at what level, as a count of questions.

### How many questions should each content area get?

Its share of the weight multiplied by the length of the paper, rounded so the total is exact. In the worked example, an area with 19% of the weight received 11 of 60 questions.

### Can EdProcto write the blueprint for us?

No. The weights are your body's judgement about its own field. EdProcto drafts the questions, runs the sitting and reports a mean for each topic; its page [for associations and boards](https://edprocto.com/solutions/associations) follows a paper from syllabus to certificate.

### How often should a blueprint be reviewed?

On a date fixed in your rules, and sooner when the role, the rulebook or the law changes. The literature we read agrees on no interval.

## References

- DiDonato-Barnes, N., Fives, H., & Krause, E. S. (2014). Using a table of specifications to improve teacher-constructed traditional tests: An experimental design. *Assessment in Education: Principles, Policy & Practice, 21*(1), 90–108. [https://doi.org/10.1080/0969594X.2013.808173](https://doi.org/10.1080/0969594X.2013.808173)
- Downing, S. M., & Haladyna, T. M. (2004). Validity threats: Overcoming interference with proposed interpretations of assessment data. *Medical Education, 38*(3), 327–333. [https://doi.org/10.1046/j.1365-2923.2004.01777.x](https://doi.org/10.1046/j.1365-2923.2004.01777.x)
- Kane, M. T. (1997). Model-based practice analysis and test specifications. *Applied Measurement in Education, 10*(1), 5–18. [https://doi.org/10.1207/s15324818ame1001_1](https://doi.org/10.1207/s15324818ame1001_1)
- Kane, M. T., Kingsbury, C., Colton, D., & Estes, C. (1989). Combining data on criticality and frequency in developing test plans for licensure and certification examinations. *Journal of Educational Measurement, 26*(1), 17–27. [https://doi.org/10.1111/j.1745-3984.1989.tb00315.x](https://doi.org/10.1111/j.1745-3984.1989.tb00315.x)
- McLaughlin, K., Coderre, S., Woloschuk, W., & Mandin, H. (2005). Does blueprint publication affect students' perception of validity of the evaluation process? *Advances in Health Sciences Education, 10*(1), 15–22. [https://doi.org/10.1007/s10459-004-8740-x](https://doi.org/10.1007/s10459-004-8740-x)
- Patil, S. Y., Gosavi, M., Bannur, H. B., & Ratnakar, A. (2015). Blueprinting in assessment: A tool to increase the validity of undergraduate written examinations in pathology. *International Journal of Applied and Basic Medical Research, 5*(Suppl 1), S76–S79. [https://doi.org/10.4103/2229-516X.162286](https://doi.org/10.4103/2229-516X.162286)
- Raymond, M. R. (1996). Establishing weights for test plans for licensure and certification examinations. *Applied Measurement in Education, 9*(3), 237–256. [https://doi.org/10.1207/s15324818ame0903_3](https://doi.org/10.1207/s15324818ame0903_3)
- Raymond, M. R. (2001). Job analysis and the specification of content for licensure and certification examinations. *Applied Measurement in Education, 14*(4), 369–415. [https://doi.org/10.1207/S15324818AME1404_4](https://doi.org/10.1207/S15324818AME1404_4)
- Raymond, M. R., & Grande, J. P. (2019). A practical guide to test blueprinting. *Medical Teacher, 41*(8), 854–861. [https://doi.org/10.1080/0142159X.2019.1595556](https://doi.org/10.1080/0142159X.2019.1595556)
- Sireci, S., & Faulkner-Bond, M. (2014). Validity evidence based on test content. *Psicothema, 26*(1), 100–107. [https://doi.org/10.7334/psicothema2013.256](https://doi.org/10.7334/psicothema2013.256)