What is a quiz answer set?
An answer set is the complete group of choices offered for one question. Quality depends on the set, not only on individual wording. Options can each sound reasonable while the set still contains gaps, overlaps, mixed units, unequal tone, or a hidden scoring bias.
Before release, document the question's purpose, every displayed label and stable value, the scoring or routing consequence, and the cases that do not fit neatly. Remove the question when it does not change a promised result or required workflow.
Related: Trace each question to a result or action · Write qualification questions that earn their place
Run the five-part option audit
Review these dimensions together. A repair that improves one dimension must not create a new failure in another.
| Dimension | Passing condition | Typical failure | Repair |
|---|---|---|---|
| Completeness | Every plausible respondent has an accurate choice | No option for not yet or not applicable | Add a truthful escape or make the item optional |
| Separation | One ordinary case maps clearly to one choice | Numeric ranges overlap | Use explicit non-overlapping boundaries |
| Symmetry | Choices use the same dimension, time frame, and tone | Mixes frequency, confidence, and preference | Rewrite around one construct |
| Consequence | Every stored value has a documented rule | Option exists only to collect detail | Remove, defer, or define its effect |
| Accessibility | Labels, instructions, order, and errors are perceivable and operable | Meaning depends on color or position | Use textual labels and semantic controls |
Check completeness without adding unnecessary data
List representative ordinary, edge, unknown, and non-applicable cases before writing options. Each case must have a truthful response. Other is not automatically the right repair because an unstructured answer may not support the scoring model. Prefer a defined none of these, not sure, or optional skip when that state is meaningful.
Do not make a sensitive or unnecessary question complete by collecting more detail. Data minimisation starts with a clear purpose and limits collection to what that purpose needs. Sometimes the correct repair is to remove the question rather than expand its answer set.
Related: Apply the quiz data-privacy checklist
Sources: ICO, Data minimisation principle
Remove overlap, hidden gaps, and mixed constructs
For numeric ranges, write the exact boundary behavior. One to five followed by five to ten overlaps at five. Under one month, one to three months, and more than three months may still leave ambiguity at exactly three months unless the displayed language and stored rule agree.
All choices should answer the same question. A set that mixes weekly, very confident, and because my team asks combines frequency, confidence, and reason. Split the constructs or keep only the one required for the result. Review tone too: if one option sounds competent and another sounds careless, people may select the socially safer wording.
Verify scoring, routing, storage, and interaction
Give every visible option a stable value ID. Record its displayed label, stored value, score contribution, branch, result effect, and follow-up effect. Test the production payload rather than assuming the saved editor configuration matches the delivered experience.
W3C guidance calls for clear labels and instructions, accessible validation, and user correction. Test native semantics or an equivalent control, visible keyboard focus, text errors, touch targets, long labels, zoom, narrow screens, and changed answers.
| Test case | Input | Expected output | Blocking failure |
|---|---|---|---|
| First option | Select with keyboard | Correct value and visible state | Focus or value mismatch |
| Last option | Select with touch and keyboard | Same rule on both inputs | Device-dependent behavior |
| Boundary option | Exact threshold case | One defined score or branch | Gap, overlap, or silent tie |
| Escape option | Not sure or not applicable | Documented neutral or separate handling | Invented score |
| Back and change | Replace a prior answer | Old value removed everywhere | Stale score or payload |
| Validation | Submit with no required choice | Clear text error and focus | Color-only or trapped error |
Sources: W3C, Form instructions · W3C, Validating input · W3C, Understanding target size minimum
Worked example: make a timing question reproducible
A timing question begins with Immediately, Soon, and Later. The labels lack a shared time frame and cannot support a reproducible qualification rule. The team rewrites them as Within 30 days, 31 to 90 days, More than 90 days, and Timing is not decided. Each choice gets a stable ID, a documented routing effect, and tests at 30 and 90 days.
These ranges are illustrative, not recommended qualification thresholds. A real team must define intervals that match its service, participant understanding, and operational capacity.
Related: Test score boundaries without fake precision · Test every reachable branch after implementation
Use this blocking release checklist
Release only when another reviewer can reproduce the option set, its consequences, and its failure handling from the documented rules.
- Ordinary, edge, unknown, and non-applicable cases all have truthful responses
- Numeric bands and categories do not overlap or leave unintended gaps
- Every choice uses the same construct, time frame, unit, and tone
- Stable visible labels and stored IDs are documented
- Scores, branches, results, and downstream actions are explicit and boundary-tested
- Back navigation and changed answers remove stale values
- No option exists only to collect unnecessary personal detail
- Keyboard, focus, semantics, target size, validation, zoom, and mobile layout pass
- The verification date, sources, limitation, and correction route are visible
Evidence
How to reproduce the method
Reproduce the method with the five-dimension option-quality matrix and six synthetic production-payload tests. Together they cover content quality, deterministic consequences, boundary handling, stale-state recovery, validation, and interaction without requiring a product account.
Limitation
Where this conclusion stops
This checklist can identify ambiguous or inconsistent answer sets, but it does not validate a psychological construct, eliminate self-report bias, or prove that a scoring model predicts business outcomes. High-stakes assessments require qualified domain and measurement review. No benchmark, live campaign, or vendor test is claimed.
Sources and verification
What this guide relies on
- The Lead Quiz Review editorial methodology
- W3C, Form instructions
- W3C, Validating input
- W3C, Understanding target size minimum
- ICO, Data minimisation principle
Sources and method checked September 20, 2026. External standards are linked to their primary publishers. Request a factual correction.