On Monday morning the teacher says the weekend went into marking the batch test. Everyone nods, because that is what weekends are for. The question nobody asks is what the weekend bought.
It is worth asking, because most of a marking weekend is not spent marking. It is spent on four other things that look like marking from the outside, only one of which needs the judgment of the person you are paying for their judgment.
The five parts of the job
Take one weekly test for a batch of forty: twenty objective questions and five written ones. That is 800 objective answers and 200 written answers in front of somebody on a Sunday.
| Step | What it involves | Needs a teacher? |
|---|---|---|
| 1. Collecting and ordering | Gathering papers, finding the three that are missing, putting them in roll-number order | No |
| 2. Marking the objective section | 800 answers against a key, one tick at a time | No — a key does this |
| 3. Marking the written section | 200 answers, each needing a decision about partial credit | Yes |
| 4. Totalling and transcribing | Adding each paper up, writing it in the register, copying it into a sheet | No |
| 5. Working out what the batch got wrong | Which questions, which topics, which students | Yes — and it is almost never reached |
One of those five requires a trained teacher. One more would change what happens in Tuesday's class. The other three are clerical, and they are where the Sunday goes.
Step five is the one worth being annoyed about. It is the only step that changes a student's result rather than recording it, and it is last in the queue, which means it happens when there is time left over. There is never time left over.
The expensive part is switching, not marking
Here is the part that is invisible while you are doing it.
Marking paper by paper means marking question one, then two, three, four and five — then starting again at question one on the next paper. Every time you move to a different question you have to reload the rubric: what was I giving two marks for here, where did I draw the line on a half-right answer, what did I do for the last student who wrote almost this.
Across forty papers with five written questions, that reload happens 200 times.
Marking question by question — all forty answers to question one, then all forty to question two — reloads the rubric five times.
If 20 seconds sounds high or low for your subject, change it. The structure of the arithmetic is the point, and it does not depend on the number.
Nothing in that requires software. It is a change in the order you pick papers up, and on paper it is done by sorting into piles by question rather than by student.
It is also fairer, which is the better argument
Holding one rubric across forty answers does something a per-paper pass cannot: it puts the forty answers next to each other. You are no longer comparing each answer against a memory of the rubric — you are comparing answers against each other, which is what consistency actually means.
There is a second effect worth knowing about. Once answers are read side by side, the student's name stops being information and starts being bias. Marking a strong student's vague answer generously, and the same answer harshly from a weak one, is not a character flaw — it is what happens when a name sits at the top of every page. Covering the names while marking a single question removes the input entirely.
On paper, fold the corner over. Any tool that offers a question-by-question queue should let you hide names in it.
Transcription is where marks silently go wrong
Step four looks like the safest part of the job and is the least safe.
A mark typically gets written three times: on the paper, in the register, and in whatever sheet produces the report. Each copy is an opportunity, and the failure is invisible — a total of 34 that should have been 43 looks exactly like a total that is right. Nobody audits a plausible number.
The cheap defence is to record marks question by question rather than as a total, and let the total be computed. A transposed digit then has to survive as an impossible per-question mark, which is visible, instead of hiding inside a plausible total, which is not. It also means that when a parent queries a result in March, the answer is a per-question breakdown rather than a number with nothing behind it.
This is the same argument as the mark sheet format for importing paper marks: the shape you record in decides what you can say later.
What step five would have told you
Suppose 31 of the 40 students got question seven wrong.
Read as marks, that is 31 students who lost two marks. Read as teaching, it is one hour of class that needs doing again — and it is the single most valuable thing the test produced. The marks tell you who is behind. The question-level pattern tells you what to do on Tuesday.
Most institutions never see it, not because they do not care but because it sits behind three clerical steps that ate the weekend. That is the real cost of steps one, two and four: not the hours themselves, but the step they crowd out. More on reading that pattern in finding questions a batch keeps getting wrong.
What software changes, and what it does not
Being plain about both halves, because the second half is usually left out.
What it removes:
- Step two, but only for tests taken on screen. Objective answers submitted online are marked against the key with no human involvement. On paper they are not: somebody still ticks 800 boxes against a printed key. Be suspicious of any claim otherwise — a platform cannot see a sheet of paper.
- Most of step four. Marks go in per question, for the whole batch at once, and the total is computed from the parts instead of being added up by hand and copied into two more places.
- The switching cost in step three, by presenting one question across the whole batch instead of one paper at a time, with names hidden if you want them hidden.
- The project nature of step five. The per-question and per-topic breakdown falls out of marks already recorded, rather than being an afternoon with a spreadsheet.
What it does not touch: somebody still has to read 200 written answers and decide what each one is worth. No tool reads a student's explanation and judges it. Anything claiming otherwise is either marking something simpler than you think, or guessing. Step three is the job, and it stays.
That is also the honest answer to whether this is worth paying for: it does not make marking disappear. It removes the work around marking.
The arithmetic worth doing once
For an institution owner, the question is not whether teachers are busy. It is which hours produce teaching.
Count the hours one teacher spends per batch per year on steps one, two and four. Collecting and ordering, ticking an objective section against a key, adding up and copying across. For a weekly test across a term, that is usually somewhere between 30 and 50 hours.
That is a working week per teacher per batch, and it is the week that produces nothing a student or a parent ever sees. It is also the week that step five was supposed to come out of.
You do not have to buy anything to recover most of it. Mark by question, cover the names, record marks per question rather than as totals, and look at the per-question pattern before you hand the papers back. If you then want the clerical steps to stop existing rather than shrink, that is what an assessment platform for coaching institutes is for — and the test of whether it is working is whether anyone finally reaches step five.
Related: what absent does to a class average, and the challenges schools face during assessments.