How to Build a Question Bank from Past Exam Papers

Every teacher sits on the same asset: years of past papers, each one containing dozens of tested, tuned questions — locked inside page-oriented PDFs where they are nearly impossible to reuse.

A question bank unlocks that asset. Instead of recycling whole papers (and letting students find them online), you assemble fresh tests from individual questions that have already proven they work.

This guide is the extraction-and-organization system, not just the theory.

Why extract at all?

You could keep whole papers and copy-paste pages into new documents. It fails in practice:

• Numbers leak. A recycled question keeps its old number and wording, and students recognize it from the 2023 paper circulating in group chats.
• You can’t reshuffle. Want five algebra + three geometry questions? That means manual editing of two source PDFs every single time.
• No visibility. “Which questions on this topic have we already used?” is unanswerable when the questions are trapped in pages.

The fix is a one-time extraction per paper, after which every question is a reusable object.

Step 1 — Extract: one file per question

Load the paper into a browser-based question cropper, draw a box around each question, and export the ZIP. The batch export means a full paper takes minutes, not an hour.

Details that matter for the bank:

• Include the question number and marks inside the box. They are part of the question.
• Stitch page-break questions. A question spanning pages must come out as one file — drag the box across the boundary and the cropper joins the fragments. Two half-questions in your bank is worse than none. Full technique: how to crop questions from a PDF test book.
• Leave answers out. If the paper has an answer key, do not box it. Answers go in your index, not in the student-facing file.
• Nothing is uploaded. Past papers with answers on them are exactly the files you should not be putting on a third-party server — see is it safe to upload a PDF to an online tool?.

Step 2 — Name so that search works

Your future self will find questions by typing a word into a search box. Design for that:

Algebra2023-Q7.pdf
Trig2022-Q3.pdf
Geometry2024-Q12.pdf
WordProblem2023-Q5.pdf

Topic-Year-QuestionNumber is the minimum viable scheme. Two rules:

• Topic first, because that is what you search by.
• Consistent spelling. Trig, not a mix of Trig / Trigonometry / Trigs — mixed naming destroys search, which is your only index.

If a question spans topics, pick the primary skill it tests. One question, one folder.

Step 3 — Index in a spreadsheet

One row per question, one column per attribute:

This is the part teachers skip and regret. Without it, “how many hard quadratic questions have we not used?” takes an afternoon to answer. With it, it is a filter.

Marking Used in is what stops the recycled-question problem at the source.

Step 4 — Assemble tests

Now new papers are a selection exercise:

1. Filter the index for topic + difficulty + unused.
2. Drop the chosen files into a new document (or import into an LMS / Google Forms quiz).
3. Reorder and renumber.
4. Log the result in Used in.

A term test that used to take an evening of page-shuffling becomes a 20-minute selection task — and every student gets a paper no one has seen before.

Common mistakes

• Banking whole papers instead of questions. The unit must be the question, or nothing above works.
• No index. A folder of 400 unnamed crops is not a bank, it is a pile.
• Including answers in student files. Crop questions only; answers live in the index.
• Never marking usage. Recycle detection dies here.
• Banking without tagging difficulty. In two years you will not remember which questions were hard.

Do it now

Pick this year’s most recent paper, crop it in the browser cropper (~5 minutes for a full paper), rename the files Topic-Year-Qn, and create the index spreadsheet with those rows. One paper is a proof of concept; by paper three, the system runs itself.