How to mark a pile of tests faster

The pile is not the problem. The problem is re-making the same marking decisions thirty times, and handling every paper twice. Craft that fixes both.

Written by the Markpilot team, practising secondary teachers · July 2026

Marking a class set of tests has two hidden costs on top of the reading itself: decision fatigue, because you keep re-litigating what a half mark looks like, and double handling, because most of us mark the pile and then go back through the pile again to enter results, work out what the class got wrong, and plan the re-teach. The practices below are all aimed at paying each cost once.

Mark by question, not by paper

If you take one thing from this page, take this. Mark question 1 on every paper, then question 2 on every paper, and so on. It feels slower because you handle each paper repeatedly. It is reliably faster, and fairer, because:

• You hold one marking decision in your head at a time. After five papers you have seen every way the class approaches that question, and marks become near-instant pattern recognition.
• Consistency is automatic. Paper twenty-eight gets exactly the standard paper two got, because you marked them minutes apart against the same criteria, not days apart with a drifting internal standard.
• Misconceptions surface as you mark. When you see the same wrong method five papers running, you have found next lesson's starter without any extra analysis.

The one exception is very short tests, where per-paper marking is fine because there is nothing to drift on.

Keep a rulings log

The first time you decide "follow-through error after a wrong first step: award the method mark", write it down, on the marking key itself or a sticky note. Every borderline decision goes in the log the moment you make it. This kills the most expensive habit in marking: re-deciding the same edge case on paper nineteen, differently from paper six, then feeling obliged to go back. The log is also what makes the marks defensible when a student appeals, and it is gold if a colleague marks the other half of the cohort.

Decide your annotation policy before you start

Writing on papers is teaching; the question is how much of it the student will actually use. A workable policy: ticks and marks on everything, a written comment only where the paper will be revisited (test corrections, exam preparation), and nothing on questions the whole class aced. If you find yourself writing the same marginal note repeatedly, stop; that note is a whole-class re-teach, not thirty private messages. Deliver it once, to everyone, and save half an hour.

Capture the data while you mark, not after

Enter marks per question as you go, into a spreadsheet or markbook with one column per question, not a single total column. Totals can be computed; per-question marks cannot be reconstructed later without re-handling the whole pile. Per-question data is what turns a marked test into decisions: which questions had the lowest facility, which students fell down where, what to re-teach and to whom. If you only record totals, you have done the marking and thrown away most of what it told you.

Moderate the borderlines, not everything

If you share the cohort with a colleague, full double-marking is rarely affordable. What matters is agreeing on the standard: mark a few papers together first to calibrate, then swap only the borderline and surprising papers. A second pair of eyes on the papers within a mark or two of a grade boundary is where moderation actually changes outcomes; everywhere else it mostly confirms what you had.

Where scanning and AI honestly fit

We build a tool in this space, so read with that in mind. The honest version:

• Reading marks off hand-marked papers is genuinely automatable now. If you mark by hand (many of us always will), a system can scan the pile and read your awarded marks into a per-question markbook, which removes the double handling entirely. Anything it is unsure of should be flagged for you rather than guessed.
• Proposing marks on unmarked papers works when it is done against your marking key and every proposal is reviewable before it counts. Treat it like a fast, tireless colleague doing a first pass: useful precisely because you check it.
• What should stay yours is the standard itself: the rulings log, the borderline calls, the final say on any mark. A tool that silently overrules you is a liability, whatever its accuracy.