> ## Documentation Index
> Fetch the complete documentation index at: https://docs.uflo.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Question Analytics
> Find the questions and concepts your class struggled with.
Question analytics shows how each question in an assessment performed. Start by comparing questions to find the ones that need attention. Then open a single question to see how students answered it and, for rubric-graded questions, the mistakes they had in common.
The examples on this page use the **Cell structure quiz** from **Introduction to Biology**, taken by 48 students.
## Compare Questions
In the teacher dashboard, select the assessment. The **Analytics** section below the submissions includes **Question Data**, a chart with one bar per question.
Average Score sorts the questions from lowest to highest score.
Choose what the bars show from the dropdown. Select a question's bar or label to open that question in a new tab.
| Option | What each bar shows |
| ------------------------ | ---------------------------------------------------------------------- |
| **Average Score** | The class's average on the question, colored by difficulty. |
| **Regrade Requests** | Requests students sent, split into accepted, open, and rejected. |
| **Tutor Messages** | Messages students sent the Socratic Tutor about the question. |
| **Tutor Effectiveness** | How students who used the tutor did compared with students who didn't. |
| **Discrimination Index** | Whether students who did well overall also did well on this question. |
Average scores follow the assessment's [**How should final grades be calculated?**](/instructors/assessments/assessment-settings) setting. With **Maximum Attempt Score**, for example, each student's best attempt counts. Turn on **Score Gain** above the charts to compare first and final attempts.
**Easy**, **Medium**, and **Hard** compare each question with the others in the same assessment. Roughly the lowest-scoring quarter of questions is Hard, the highest-scoring quarter is Easy, and the rest are Medium.
The legend shows the cutoffs. A question marked Hard in an easy quiz might still have a high average.
The discrimination index compares how students scored on one question with how they scored on the rest of the assessment. It ranges from −1 to 1.
Q6 is Low: students who did well overall did worse on it.
| Color | Value | What it suggests |
| -------- | ------------------ | --------------------------------------------------- |
| **Good** | 0.20 or above | Strong students also did well here. |
| **Weak** | 0.00 to under 0.20 | The question says little about overall performance. |
| **Low** | Below 0.00 | Strong students did worse here than weaker ones. |
A Low question often has a wrong answer key or unclear wording. Uflo colors bars Good or Low only once 30 students are graded. Until then, every bar shows as Weak.
Tutor effectiveness compares students who messaged the Socratic Tutor about a question with students who didn't. Each bar shows the gap between the two groups in percentage points.
Q1 has the largest gap but shows Low Signal, because only 4 students used the tutor.
When scores changed between attempts anywhere in the assessment, the bars compare how much each group improved. Otherwise, they compare final scores.
A gap of 3 points or more shows as **Positive** or **Negative** once each group has at least 20 students. Anything else shows as **Low Signal**. Students who ask for help may differ from those who don't, so treat the result as a pointer, not proof.
## Review One Question
Select a question's bar in Question Data, or open the assessment and go to the question. Then select **Analytics** in the panel beside the question. Only teachers see this tab.
The Analytics tab sits next to Socratic Tutor and Solution.
### Question Overview
The table at the top of the tab summarizes the question. **Rank** places it among the assessment's questions, highest value first. For example, **#6 of 6** on Average Score means the question had the lowest average.
Question 4 has the lowest average and the most regrade requests.
| Metric | What it shows |
| ------------------------ | ---------------------------------------------------------------------------------------- |
| **Average Score** | The class's average, as a percentage of the question's points. |
| **Discrimination Index** | The discrimination index, with the number of students behind it. |
| **Avg Time Spent** | Average time each student spent on the question. |
| **Regrade Requests** | [Regrade requests](/instructors/results/regrade-requests). Rubric-graded questions only. |
| **Tutor Messages Sent** | Messages students sent the Socratic Tutor about this question. |
| **Tutor Usage** | The share of students who messaged the tutor about it. |
Ranks for Average Score and Discrimination Index turn green in the top third and red in the bottom third. Regrade Requests works the other way: many requests turn red.
### Charts
The charts below the table depend on how the question is graded. Uflo uses one attempt per student: the highest-scoring one with **Maximum Attempt Score** grading, and the most recent one otherwise.
Multiple choice, multiple correct choices, number, and math questions show a **Response Breakdown**. Each bar is one answer, and the correct answer is green. **No response** counts students who left it blank.
The correct answer, B, is green. Most students who missed it chose A.
* Number responses within the question's margin of error share a bar.
* A math response shows green only when its text matches the answer key. An equivalent form, such as 2x + 2 for 2(x + 1), is graded correct but isn't green.
* Fill-in-the-blank questions with more than one blank show a **Correct Response Rate** instead, with one bar per blank. Select a blank to see every answer given for it.
Written, drawing, and image questions show three charts.
* **Score Distribution** groups students by their score on the question.
* **Per-Rubric Breakdown** shows the average score for each rubric criterion. Switch the dropdown to **Avg Confidence** to see how sure the AI grader was on each one.
* **Regrade Requests** splits requests into accepted, rejected, and open.
Explaining why cells stay small has the lowest average of the three criteria.
Select a bar in **Score Distribution** or **Response Breakdown** to list the students behind it, with their score and number of attempts. Select a student to open their latest submission.
## Find Patterns in Responses
For rubric-graded questions, Uflo can group the mistakes students made into patterns. When it has analyzed the class's graded responses, the patterns appear under **Response Summaries**, below the charts. Each pattern sits under the rubric criterion where students lost points.
Students who earned full marks on every attempt aren't included, and neither are very short responses.
Each pattern shows its type, how many students it affects, and how many fixed it.
Select a criterion card's header, or its bar in **Per-Rubric Breakdown**, to see that criterion's pattern breakdown and score distribution.
### Pattern Types
| Label | What students did |
| ----------------- | ----------------------------------------------------------------------------- |
| **Misconception** | Showed a wrong idea, such as thinking larger cells exchange materials faster. |
| **Incomplete** | Left out required content or answered too vaguely to earn the points. |
| **Calc Error** | Understood the concept but made a calculation mistake. |
Select a pattern to read its description. Select its **students** count to list the students in it.
### Resolution
Uflo compares each student's lowest-scoring attempt with their highest-scoring one. Each student in a pattern gets a status, and the small bar on the pattern row shows the mix.
The Status column shows whether each student fixed the mistake in their highest-scoring attempt.
| Status | Meaning |
| ------------------ | ------------------------------------------------------------------ |
| **Resolved** | The mistake is gone from their highest-scoring attempt. |
| **Partial** | Part of the mistake remains in their highest-scoring attempt. |
| **Still Stuck** | The same mistake remains in their highest-scoring attempt. |
| **Single Attempt** | The student made one attempt, or scored the same on every attempt. |
Patterns with many students still stuck are usually the best place to start. Select a student to open their submission and read the response yourself.
### Insights
The **Insights** card suggests likely causes behind the patterns and what you could do in class. It gives more weight to mistakes students haven't fixed. AI writes these suggestions, so read a few responses behind a pattern before you change your teaching or the rubric.
## What Each Path Supports
Scores, difficulty, and the discrimination index work for every assessment, including imported paper and LMS scores. Some data needs more:
* **Response Breakdown** and **Response Summaries** need typed responses. Scanned paper assessments save each response as a page of the scan, so they don't get either.
* **Avg Time Spent** needs students to answer in Uflo.
* **Tutor Messages**, **Tutor Usage**, and **Tutor Effectiveness** need students to use the [Socratic Tutor](/instructors/assessments/assessment-settings) on the question.
* **Score Gain** and pattern resolution need more than one attempt in Uflo.
## Common Questions
Question groups don't have one. Open one of the questions inside the group. The tab is also hidden when you switch to student view.
Response Summaries cover only written, drawing, and image questions worth points, and not scanned paper responses. They also need graded responses with mistakes to group, so a question most students answered fully may have none.
Every question in the assessment has zero for that metric. For example, Tutor Messages Sent shows N/A until a student messages the tutor about any question.