Setting a test that is worth checking.
A paper that produces a ranked list has done the easy half of its job. The useful half is telling you why the ranking came out that way, and most papers are not built to do that.
The problem with a well-set paper
A good paper, in the conventional sense, spreads students out. It has easy questions, medium ones, and two that only the top few will finish. It produces a clean distribution.
And it can still leave you unable to answer the only question that matters on Monday morning: what does this batch not understand? A student who scores 58 might have lost marks on one concept or on eight, and the total cannot tell you which.
Five changes that make a paper diagnostic
1. Make questions test one thing each
A question that requires three concepts gives you no information when it is wrong. Two questions testing one concept each tell you exactly where the failure is. Compound questions are appropriate at board level; they are poor instruments mid-chapter.
2. Repeat a concept in two forms
Ask the same underlying idea as a numerical in one place and as a reasoning question in another. A student who gets one and not the other has a specific, nameable gap — usually that they can execute a procedure without understanding what it means.
3. Award the method visibly
Write the marks per step into your own scheme before the test, not while marking it. It forces you to decide what you are actually rewarding, and it is what makes error carried forward applicable at all.
4. Include one question you expect to go badly
Not to be cruel — to get a reading. A paper where everything is answerable tells you only that the class can do what you already taught. One question at the edge tells you where the edge is.
5. Ask for reasoning where you can mark reasoning
“Explain why” is only worth asking if you have the time to read the explanations. This is the honest constraint behind most paper design: teachers set questions they can afford to mark, and short-answer papers win by default. If the checking cost falls, the constraint moves, and better questions become affordable.
What this has to do with checking
A diagnostic paper is a more expensive paper to mark. That is why most centres do not set them, and it is the actual reason marking cost shapes teaching quality rather than just faculty workload.
When every mark is recorded question by question rather than as a total, the batch-level pattern falls out for free — which two questions collapsed, and which section they collapsed in. That is what makes the effort of setting a better paper pay for itself.
See what your last paper actually says.
Send one you have already marked. You will get the question-by-question picture for the whole batch, not just the totals. The first one is free.