A 600-student course can produce more than 3,000 written responses from a single exam. Spread across a teaching team in the last week of term, that is where grades go late, scores drift, and regrade requests pile up into the next semester.

Grading on time is mostly a planning problem. This article walks through the math, the exam design choices that make grading manageable, and a day-by-day plan for posting grades by Friday.

Do the math first

Before the exam is written, estimate how long it will take to grade. Most teams never do, which is how they end up surprised.

Example: say an introductory biology course has 600 students and a final exam on Monday evening. The exam has:

  • 40 multiple-choice and numeric items, scored automatically
  • 4 short-answer questions, about 1.5 minutes each to score
  • 1 data interpretation question, about 4 minutes to score

The written work adds up quickly. Four short answers for 600 students is 2,400 responses. At 1.5 minutes each, that is 3,600 minutes, or 60 hours. The data question adds 600 responses at 4 minutes, or 40 hours. Total: about 100 hours of scoring.

With 8 teaching assistants, that is about 12.5 hours each. Add double scoring on a 10 percent sample for consistency checks, and it rises to roughly 14 hours each. Spread over Tuesday to Thursday, that is under 5 hours a day per person, which is demanding but possible.

These estimates are illustrative. Time a few real responses from a past exam to get your own numbers, then plan with them.

Design the exam for grading

The exam itself decides most of the grading load. A few choices made in week 10 save days in week 15.

  • Score recall automatically. Facts, definitions and calculations can usually be multiple-choice, numeric entry or matching.
  • Save writing for judgment. Use constructed response for explanation, interpretation and argument, where it adds information.
  • Bound the response. “Explain in two to three sentences” is faster to score and fairer than an open box.
  • Write the rubric with the question. If you cannot write a rubric for it, the question is not ready.
  • Avoid chained parts. A part (b) that depends on part (a) multiplies grading decisions and regrade disputes.

Our item review checklist helps the teaching team check items for clarity, cues and accessibility before the exam is set.

Grade by question, not by student

The fastest consistent method is to split the team by question, not by stack of exams. Each TA, or pair of TAs, scores one question for all 600 students.

Grading by question has three advantages:

  • A scorer holds one rubric in mind, so consistency improves.
  • Speed rises after the first 50 or so responses, as the scorer learns the patterns.
  • Disagreements are easy to trace, because one question has one owner.

Assign the hardest question to the most experienced scorers, and pair new TAs on a shorter question for the first session.

Calibrate before anyone scores

Spend 60 to 90 minutes on Tuesday morning calibrating. For each question:

  1. The instructor walks through the rubric and the expected answer.
  2. Everyone scores the same 10 responses independently.
  3. The group compares scores and discusses every disagreement.
  4. The instructor picks anchor responses for each level and writes a line on why.
  5. Scorers start live grading only once they match the agreed scores on a short check set.

An hour of calibration on Tuesday is far cheaper than rescoring 600 responses on Saturday. Our article on writing rubrics that score consistently covers how to write descriptors that make calibration faster.

A day-by-day plan

Here is an illustrative week for the example course.

Day What happens Who
Monday Exam 7 to 9 p.m.; automatic items scored on submission Instructor, proctors
Tuesday Calibration 9 to 10:30 a.m.; scoring begins by question Instructor, all TAs
Wednesday Scoring continues; double-scored sample checked at 4 p.m. TAs, head TA
Thursday Scoring finishes by 3 p.m.; flagged responses reviewed TAs, instructor
Friday Item statistics checked; grades released by 5 p.m. Instructor

Two checkpoints carry the plan. On Wednesday afternoon, the head TA compares double-scored responses and stops any scorer who is drifting. On Friday morning, the instructor checks multiple-choice item statistics for any item with a negative discrimination, which may signal a miskey, before grades go out.

Where automation fits

Automated scoring can take on part of the load, with people still accountable for every grade.

  • Multiple-choice and numeric items are scored on submission, with item statistics ready to review.
  • Short answers can get an automated first pass that groups similar responses and proposes scores. TAs review the groups and any response the system is unsure about, rather than starting from a blank page.
  • Every proposed score can be changed by a TA, and the change is logged.

How much time a first pass saves depends on the question, the rubric and the routing rules. It is worth measuring in a pilot on one exam rather than assuming.

Plan for students who test later

In a 600-student course, some students will not sit the exam on Monday evening. Students with approved accommodations may test with extra time or in a separate session. Others will have documented conflicts or illness.

Plan for them in advance rather than on Thursday night:

  • Schedule make-up sessions early. Tuesday or Wednesday keeps their responses inside the grading week.
  • Use an equivalent form. If make-up students see a different version, check that it matches the main form in content and length.
  • Keep extra time configured. Set accommodations on the exam before Monday, per student, so nothing is set up by hand on the night.
  • Route late responses to the same owners. The question owner scores make-up responses with the same rubric and anchors.

A student who tests on Wednesday should receive a grade on Friday like everyone else.

Handle regrades on a schedule

Regrade requests are fair and expected. They become a problem when they arrive by email for weeks.

  • Open a regrade window of five to seven days after grades post.
  • Require a short written reason tied to the rubric.
  • Route each request to the question owner, not the student’s section TA.
  • Allow the whole response to be rescored, so scores can go up or down.
  • Log every decision with a reason.

A clear window and a logged decision protect students and the teaching team alike.

Fewer big exams, more small ones

One structural change reduces the Friday crunch: spreading assessment through the term. Frequent low-stakes quizzes scored automatically shift weight away from a single high-stakes final.

There is also a learning reason. Roediger and Karpicke (Psychological Science, 2006) found that students who took recall tests remembered more two days and one week later than students who spent the same time restudying, although restudying did better on a test five minutes later. Short weekly quizzes may support learning as well as lighten the final grading load.

What to do this week

  • Time five real responses for each written question on your next exam.
  • Multiply by enrollment and divide by your team to get hours per person.
  • Move any pure recall question to an automatically scored format.
  • Assign one owner per written question.
  • Book a 90-minute calibration session for the morning after the exam.
  • Write the regrade policy and post it before the exam.

Where we can help

Our Grading Pilot runs automated first-pass scoring with TA review on one course or exam for one term, so you can compare its scores with your TAs’ scores and the hours the team spent. It is indicative from $2,400. The full automated grading service then scales that across courses, with LTI 1.3 grade passback to Canvas, Moodle, Blackboard and D2L Brightspace.

For the wider picture of how we work with colleges and universities, including large-enrollment courses and accessibility, see our page for higher education teams.