Markscanner
Back to writing
/ Markscanner team / feedback

Giving better per-question feedback without starting every paper from scratch

Practical advice on writing math feedback that actually helps the student — and how to reuse comment patterns across a class set without the feedback feeling canned.

Good feedback on a math quiz has three parts. It names what the student did wrong, names what a correct approach would look like, and ideally points at what to practise next. That takes about forty-five seconds per question to write well. On a class set of thirty, with five questions each, that’s nearly two hours of feedback writing — after you’ve already marked the papers.

Which is why most of us don’t do it. We write “check your signs” on one paper, “see me” on another, put a mark at the top, and move on. The students who needed the feedback the most are the ones who got the least of it.

There’s a better version of this, and it doesn’t require another two hours. It requires the observation that on a given question, most wrong answers are wrong the same way.

The same five mistakes, over and over

Pick a question your class just wrote. Something multi-step, say factoring a quadratic of the form ax² + bx + c with a ≠ 1. Mentally sort the wrong answers into buckets. You won’t need many buckets.

Almost every wrong answer will fall into one of these:

  • Didn’t factor out the common factor first.
  • Got the signs wrong on the middle term.
  • Factored as if a = 1.
  • Got the arithmetic wrong in the last step.
  • Left it in an unfinished form.

Five buckets, thirty papers. If you write one good feedback comment per bucket — concrete, kind, actionable — you’ve covered twenty-eight of the thirty students on that question. The other two get a bespoke comment because something genuinely unusual happened.

What a good feedback comment actually looks like

Short, specific, and pointing forward. Three sentences is usually plenty.

Bad: “Check your signs.”

Better: “The middle term here is -7x, not +7x — that flips which pair of factors of -18 you need. Worth re-doing this one from the factor list.”

Also better: “Setup’s correct, and the factoring is right up to the second-last line. The arithmetic slip at the end cost you the final mark — try writing each multiplication on its own line next time.”

Both of those name what happened, point at what should have happened instead, and leave the student something concrete to practise. Neither took long to write.

Reuse comment patterns — just personalise the specifics

This is the step where people get nervous. “Reusing comments” sounds like “mass-produced feedback”, which sounds impersonal. It isn’t, if you do it right.

A comment pattern is the shape of the feedback. The specifics change per student.

Pattern: Your [specific correct thing] is right. The issue is at [specific incorrect step]. To practise this, try [concrete suggestion].

Student A: “Your setup of the quadratic is right. The issue is at the factoring step — you used (2x - 3)(x + 6), but -3 × 6 = -18, not +18. Try a few of the a ≠ 1 examples from section 5.3 to build the pair-finding habit.”

Student B: “Your setup of the quadratic is right. The issue is at the last arithmetic line — 3 × 4 = 12, not 16. The method is solid; try writing each multiplication out separately under time pressure.”

Same pattern, genuinely different comments, and each one is useful to that student. You wrote one thing; they got something personal.

The “strengths and next steps” bit

On a one-page feedback sheet with per-question comments, most of the comments are already about what went wrong. That’s fine — it’s literally the information that’s missing from a bare number at the top of a paper.

But it’s worth saving the last couple of lines for two things a student can’t reconstruct from per-question notes:

  • One thing they did well, across the assessment. Not on a specific question — across. “Your setup of word problems is consistently clean” is worth more than another check-your-signs note.
  • One thing to practise next. If there’s a pattern across the errors — sign errors, skipping the common factor, rushing the arithmetic — name it and suggest one specific thing.

Two lines. A student who reads nothing else on the page will read those two.

On using a tool to draft the comments

If you’re using AI to draft per-question comments, the same rules apply: the tool can produce the pattern, but it shouldn’t be the last word. Before you print, read the draft comments as if a colleague wrote them. If one sounds like a chatbot, rewrite it. If one is kinder or harsher than you’d be, change the tone. If one is generic — “keep practising!” — replace it with something specific.

The rule of thumb: the tool’s job is to save you the first forty-five seconds per comment. Your job is to spend the last ten making sure the comment sounds like you and actually helps that student. Done well, it still comes out to way less than two hours, and the feedback is noticeably better than the pile usually got.