Blog
5 Common Mistakes When Preparing Question Banks for Online Test Platforms
The recurring errors that break a bulk-import — mismatched option counts, wrong answer format, missing marks — and how to avoid each one.
1. Inconsistent option counts
Most bulk-import templates assume a fixed number of options per MCQ — usually four. A question with only three visible options because one was accidentally dropped during manual entry, or five because a "none of the above" got tacked on inconsistently, is one of the most common reasons an otherwise-fine file gets rejected on import. Matching-type questions that legitimately need a fifth option are the one common exception most templates do account for — but only when it's expected, not by accident.
2. Answer format mismatches
Different platforms expect the correct answer in different shapes — a letter (A/B/C/D), a 1-based option number, or the option marked correct/incorrect directly in its own row. Mixing conventions within one file, or using the wrong one for the target platform entirely, is a fast way to end up with a bank where every question imports but half the answer keys are wrong.
3. Losing the answer during manual transcription
When questions and their answer key live in two separate source documents, manually cross-referencing them for even a hundred questions invites mismatches — an answer copied against the wrong question number is a silent error that only surfaces when a student reports a wrong "correct" answer. Matching by question number automatically, rather than by hand, removes this specific failure mode entirely.
4. Missing positive/negative marks
A bulk-import file with no marks specified per question, or with marks that don't match what the test's marking scheme is supposed to be, either gets rejected outright or silently defaults to whatever the platform assumes — rarely what you actually wanted. Setting positive and negative marks explicitly for the whole set before exporting avoids finding this out after students have already taken the test.
5. Skipping questions with no stated answer
A source PDF that doesn't explicitly state a question's answer shouldn't mean that question gets dropped from the bank entirely — but it also shouldn't mean a guessed answer gets silently inserted either. The better middle ground is having those questions solved and clearly marked as AI-derived (or left blank for manual review), so you know exactly which answers came from the source material and which didn't.