Keep the question-level evidence
The answer ledger records filename, question, detected answer, expected answer, mark state and ink scores. The summary reports confirmed correct answers and unresolved counts. An unresolved answer should be checked before a final grade is entered elsewhere.
Check the batch setup
Read a small known sheet first, verify the answer key length, and confirm that option letters match the printed form. A wrong answer key can produce internally consistent but meaningless grades.
Keep original scans according to your school’s data-handling rules. This tool has no server archive or account history; download the ledger before closing the tab.
Build a useful batch naming convention
Use filenames that identify the sheet within your own records before selecting the images. The CSV keeps those filenames as the main link back to the source. Avoid ambiguous names such as scan.jpg repeated across folders. If student identifiers are sensitive, use an internal code and keep the code-to-student mapping separately under your normal access controls.
Run the first batch without manual corrections. Download the answer ledger and summary, then filter the ledger for ambiguous or multiple-mark states in a spreadsheet. Reprocess each affected sheet individually when adding corrections, because the same question=answer correction otherwise applies to every image in the current batch.
Check totals before transferring results
For a 10-question sheet, compare confirmed correct, incorrect and unresolved rows across all 10 questions. Do not copy only the correct total when unresolved marks remain. After correction, retain the updated ledger with a clear filename so it cannot be confused with the first pass. The export does not submit grades to a school system, store a server history or identify students automatically; those administrative steps remain separate from optical mark reading.