Scoring

Every question is scored in two steps: the answer is evaluated to a success rate, then that rate is converted to points and, optionally, rolled up into competency dimensions.

Updated 2026/07/21

Every question produces a score in two steps: first the answer is evaluated to a success rate — a percentage that can be positive (partially or fully correct) or negative (partially or fully wrong); then that percentage is converted to points through the question's multipliers. See the panels below for exactly how the positive and negative multiplier apply.

How the success rate itself gets produced, and what happens to the resulting points, depends on the question:

Associated Dimensions — tag a question with one or more competencies so its score also rolls up into a per-dimension breakdown, not just the overall total.
Scored Dimensions — for responses that should be judged differently per competency (an essay strong on structure but weak on vocabulary), score each dimension independently instead of sharing one success rate.
Rubric Scoring — attach a structured set of criteria and levels so reviewers evaluate open-ended responses consistently instead of freehand.
AI Evaluation — let AI compute the success rate automatically from a grading prompt you write, for the question types that support it.
Preview & Scoring Simulation — see the question exactly as a candidate would and simulate different answers to check the scoring behaves the way you intend before it goes live.

Always run a question through Preview & Scoring Simulation after setting up multipliers, dimensions, or a rubric — it's the fastest way to catch a scoring mistake before a candidate does.
Good to know

In this section

Scoring Framework The Scoring Framework is where your organization defines the dimensions and sub-dimensions every question can be scored against — set it up once here, then assign it question by question. Associated Dimensions Associated dimensions are named competencies — Verbal Reasoning, Attention to Detail, Grammar — you attach to a question so results roll up into a per-dimension breakdown alongside the overall score, not just a single number. Scored Dimensions Score one response across several independent competencies: each scored dimension gets its own score from the evaluator or AI, and the question score is derived from them — or judged separately. Rubric Scoring How rubric scoring works in TestInvite — what rubrics are, which question types support them, and an overview of the four rubric types. AI Evaluation Score open-ended answers automatically with a large language model: how AI evaluation works per question type, how to write an effective grading prompt, and how reviewers confirm or adjust the results. Preview & Simulation Preview any question exactly as candidates will see it and simulate its scoring — pick answers and watch the computed score update to verify multipliers, partial credit, and penalties before the question goes live.

The positive multiplier is the question's weight — the points awarded for a fully correct answer. A positive success rate scales against it proportionally: points = success rate × positive multiplier.

100% success rate × multiplier 1 = 1 point
100% success rate × multiplier 3 = 3 points
50% success rate (partial credit) × multiplier 2 = 1 point
0% success rate (wrong or blank) × any multiplier = 0 points

Leave every question at the default multiplier of 1 for equal weighting, or raise the multiplier on questions that matter more — a hard case-analysis question at 5 counts as much as five routine questions at 1.

The negative multiplier works exactly like the positive one, mirrored: it only comes into play once the success rate itself is negative, and points are simply that negative percentage multiplied by the negative multiplier — points = success rate × negative multiplier. A wrong answer isn't a flat penalty; the answer must first evaluate to a negative success rate, and the size of that rate determines how much of the multiplier gets applied.

−50% success rate × negative multiplier 2 = −1 point
−100% success rate (fully wrong) × negative multiplier 1 = −1 point
−100% success rate (fully wrong) × negative multiplier 0.33 = −0.33 points
0% success rate (blank / neutral) × any multiplier = 0 points

A single-answer multiple choice question that's answered incorrectly evaluates to a success rate of −100% (fully wrong), so the full negative multiplier applies. Some question types can produce a partial negative rate instead — for example, a multi-select question where a few wrong picks among several correct ones might evaluate to −50%, so only half the negative multiplier's value is deducted.

In practice, negative multipliers are often set as a fraction rather than a whole number — a common convention on a four-choice single-answer question is a penalty of 1/(number of choices − 1), or about 0.33, so that a random guess (right 25% of the time, fully wrong the other 75%) has an expected value near zero and no longer pays off.

Good to know: the negative multiplier only ever multiplies a negative success rate — setting it to a large value does nothing by itself. If you want a question to actually produce negative marks, structure it so a wrong answer can evaluate to a negative percentage. A question type or configuration that only ever produces 0% for a wrong answer will never subtract points, no matter what the negative multiplier is set to.