Connect questions with an ID, not similar text
In the example bank, three questions describe the Harbor Library garden. All three use passage_id=garden. A selected group prints the passage once, followed by its selected questions. The app never merges passages just because their words look alike.
| Source question | passage_id | What it asks |
|---|---|---|
| 1 | garden | Main idea |
| 2 | garden | Workshop day |
| 3 | garden | Vegetable destination |
Questions without a passage ID remain independent, even if their passage text is exactly the same. A different question bank has its own ID scope; reusing the word garden there does not connect it to an older saved exam.
Two questions still means two questions
Suppose you request two questions and the random selection picks source questions 1 and 3. The paper contains the common passage and those two questions. Question 2 is not automatically added. Within that selected group, original numbers determine the order.
The new paper numbers the selected questions 01 and 02. Their labels still say Source 1 and Source 3. The answer key uses the same new paper numbers, so you can mark the paper and still trace each item back to its source.
Resolve conflicting copies before teaching
If two valid rows use the same passage ID but have different text, the app keeps them in one group and uses the text from the first valid row. The import report identifies the conflicting row and the representative’s line number.
For an accidental edit, make all copies agree and reimport the CSV. For two genuinely different readings, give them different passage IDs. Importing the correction updates the active bank only: an already confirmed exam stays unchanged. Create a new exam version if it needs the corrected passage.
Keep choice order when it carries meaning
Answers such as “None of the above,” references to option numbers, or ordered steps may depend on where choices appear. Set lock_options=true in the CSV or check Lock choices during review. A global shuffle setting respects that per-question lock.
A small phrase detector flags some option references and locks those questions by default. It cannot recognize every wording or language. Inspect each question and its answer key; automated warnings do not replace your review.