A flawed item costs more after a window than before it: a candidate appeal, a rescore, a form pulled. Run each draft through these checks while it is still a draft.
Does the item belong here?
Content checks, done by a subject matter expert who did not write the item.
Matches the blueprint
Tagged to one domain and task, at the thinking level the blueprint sets.
One defensible key
Exactly one best answer, confirmed by a second expert against a cited reference.
Important, not trivial
Tests what a competent practitioner needs on the job, not a footnote or a date.
Current and sourced
The reference and its edition are recorded, and no rule behind the item has changed since.
Stands alone
Neither gives away nor depends on the answer to another item on the form.
Can every candidate read it?
Clarity checks, so the item measures the skill and not the wording.
Stem asks the question
Cover the options: a candidate who knows the content can answer from the stem alone.
Positive wording
Avoid NOT and EXCEPT; if one is essential, set it in bold capitals.
Parallel options
Options match in length, grammar and form, and the key is not the longest.
Plausible distractors
Each wrong option reflects a real mistake; three good options beat four with filler.
No catch-all options
Drop “all of the above” and “none of the above”; they reward test-taking tricks.
Fewest words needed
Cut background that does not change the answer; keep the reading load to the task.
Is it fair to every group?
Bias checks, done by a reviewer trained for it, with a second look after the window.
No insider references
No regional slang, sports, holidays or cultural knowledge from outside the field.
Varied, respectful people
Names and roles in scenarios drawn from varied backgrounds, with no stereotyped roles.
Sensitive topics justified
Distressing content appears only when the job itself requires handling it.
Statistics checked later
After the window, items flagged for differential functioning go back to reviewers for a content look.
Can every candidate reach it?
Accessibility checks on the item itself, before it reaches the interface.
Images described
Alt text carries the information needed to answer, without revealing the key.
Not color alone
Charts and hotspots label their parts in words or patterns as well as color.
Keyboard tested
Drag, hotspot and ordering items tried by keyboard and screen reader before approval.
Readable math
Equations entered as MathML, never as pictures of equations.
One flawed item, then fixed
The draft stem read “Which of the following is not true about grounding?” Its key was the only option with a qualifier, and one distractor was partly correct. The fix: a positive stem asking the purpose of an equipment ground, one key checked against the code reference, and three distractors drawn from mistakes apprentices actually make.
Train your experts to this standard
An Item Writing Workshop from $5,500 trains your experts to write and review to these checks.