Question Analytics

Question Analytics reports how individual questions perform across every place they've been asked — useful for finding questions that are too easy, too hard, or not actually separating strong candidates from weak ones.

Where to find it

Open Analytics → Question Analytics for an organization-wide view across every question. It's also embedded scoped to context: the Analytics tab in the Test Editor shows only questions used in that test, and the Analytics tab in the Task Editor shows only questions used in that task.

Filters

Narrow by folder, test, task, and date range, and choose a page size. Folder filtering is hidden when you're already scoped to a single test. Selecting a task narrows which tests you can pick (and vice versa) — the two lists stay mutually consistent, since only tests actually delivered through the selected task(s) remain selectable. When opened from inside a test or task, that filter is locked to the current context.

Reading the table

Each row is one question, with a header strip totaling question count, asked count, and answer rate/average score across your current filters. The columns:

• Asked — how many times the question was presented to a candidate.
• Answered — how many of those were actually answered, plus the answer rate.
• Average score — the mean score candidates received on this question.
• P-index — the percentage of candidates who answered correctly. Shown red under 30% (likely too hard, or a possible error in the question/key), amber over 85% (likely too easy to discriminate between candidates), green in between.

D-index: only meaningful for one test at a time

A D-index column appears only when your filter narrows to exactly one test. D-index compares how the top 27% of scorers on that test answered a question against how the bottom 27% did — a well-discriminating question should be answered correctly far more often by top performers. That comparison only makes sense within one shared, ranked population, which is why D-index can't be computed across multiple tests or a folder at once: candidates from different tests aren't ranked against each other.

Exporting to CSV includes P-index always, and D-index with its interpretation label whenever the single-test view is active.

A question with low P-index (hard) can still be a good question if its D-index is high — it's correctly separating strong candidates from weak ones. Treat P-index and D-index together, not P-index alone, before deciding a question needs rewriting.

Tip