Written and structured response questions

Open text

Open text is for free-form written answers, from a one-line definition or customer reply to a full case analysis or structured recommendation. Short prompts keep review quick, because evaluators read a few lines per response instead of scanning long passages for the relevant point.

For case-based prompts, pair open text with a sidebar layout so the brief stays visible while the candidate writes. Open text can auto-score against exact, contains, or regex rules when there is a narrow set of acceptable answers — terminology, exact values, completions — or go to rubric-based review when you are assessing reasoning and expression.

Numeric and table input

Numeric input captures a single number, which suits calculations, estimates, and quantitative reasoning where you care about the result rather than the working. The answer key takes a correct value with a tolerance, so rounding differences do not cost the candidate the item, and results are easy to compare across candidates at a glance.

Fill-in-a-table captures structured data in a grid. Use it when the answer is naturally tabular — a schedule, a budget breakdown, a categorized list — and you want candidates to fill in cells rather than describe the data in prose. Table items auto-score cell by cell against your answer key, so keyed tables need no manual review at all. Individual cells can also be marked as given: their text is shown to the candidate pre-filled and read-only, so you can hand out a partially completed table — an example row, known values — and ask only for the blanks. Row and column headers can be asked too — turn a header into a blank the candidate must fill, with your text as the answer key — and neighboring cells can be merged into larger regions for realistic table shapes.

Matching and ordering

Matching asks candidates to pair items across two columns — terms to definitions, problems to solutions, tools to use cases. By default the answer is complete when every row is matched, but you can set match limits: a minimum lets candidates stop earlier, and a maximum locks the remaining rows once reached, with the rule shown under the question. Ordering asks them to arrange steps into the correct sequence, which suits process knowledge: an escalation path, a troubleshooting flow, the stages of a procedure. Both types auto-score against their keys, so they combine recall-level difficulty with zero review effort.

How these items are scored

Auto-scoring is driven by an answer key: numeric items take a correct value with a tolerance, table items a per-cell key, matching and ordering their correct arrangements, and open text optional exact, contains, or regex rules. An item that has no key auto-scores as null — and null means "needs manual review", never zero. A blank auto-score does not penalize the candidate; it flags the response for a human.

Plan review capacity accordingly: open written responses without rules will typically be scored by evaluators, often against a rubric, and every recorded score is attributed to a named evaluator with a timestamp. If a task ends up with both a rubric score and an auto score, the rubric score wins.