Guidelines for creating/designing an exam?

In my opinion the most important question is what should the exam measure? Depending on the institution, the type of exam, and the course this can vary a lot. Basically, exams are to learning what thermometers are to temperature. You want to have it as precise as possible, and not to build a barometer instead.

Once you know what you want to measure, check it once more: are these expectations realistic given the course and training given to the students? Are they enough to ensure that passing students will benefit from the follow-up courses? Here you will usually need some coordination with other courses.

I advise not to adapt the exam to the level of effort actually produced by the students, but to the expected level of effort. Since my students seem to produce much less effort than I expect, I usually give quite low grades (I always warn them of this fact). This may seem harsh, but in fact I think I am doing them a favor not to blind them with artificially boosted grades that let them believe they will do great with little effort in the following years, which is usually a lie. I have seen too many students stuck in third year with no hope to go forward, after hardly passing the first two years thanks to very generous grades. Of course, this point of view is hugely dependent on the environment.

An exam should be too short rather than too long given the allowed time. If it is too long, you will have to boost the grade to be fair, but that will probably give too high grades to students who master only a small portion of the course. If the exam is too short, you can still measure the level of mastery they achieved. The only exception is when you really want to measure their speed, but it is rarely relevant. My personal guideline is that I let my students three to four times as much time than I take to write the complete solution to the exam (I have been trained to take exams quickly all along my studies, your preferred ratio may differ but it is useful to establish a ratio in a way that ensures that time is not too much of an issue).

I advise against always giving more points to more difficult questions, especially for numerical grades (a note is in order : in France we grade on 20 points and the passing grade is always 10, and in higher education one can almost always compensate a grade below 10 by better grades in other courses, even in minors). The point is to have a grading scheme which actually matches what you want to measure. I thus usually assign about 8 points to very basic or classical questions, 8 points to questions that are easy but necessitate some understanding of the material (as opposed to reproducing a method seen ten time without the need to actually think), and only 4 points to more sophisticated questions, even if they are long to solve. I try to make this clear to students, so that they first solve the easier questions. The more sophisticated questions thus only serve to do the difference between student with a correct mastery on the course, from truly bright ones. I also want to prevent students who don't really understand what is going on in the course but manage to reproduce standard exercises to get a passing grade. This grading scheme is not necessarily appropriate in other systems, but the important point is that your grading must be tailored to your goal, without letting other principles getting in your way.

Another point to think about is whether you want to have your exam very close to the course itself (which is what student prefer and what gives the highest grades) or on the contrary to have parts which are significantly different, but rely strongly on the course material. This later choice is in order if you want to measure how much your students are able to use what they learned in another context. Be warned that results are often disappointing, and choose this option with care.

On the long run, try to adjust your exam depending not on the percentage of passing grades (unless you are forced into it, which often happens), but depending on the correlation between passing grades and future success of students. The information is often difficult to obtain, but if you realize that your students struggle to get a passing grade with you but have good success in follow-up courses nonetheless, you should be more gentle in your questions or grading. On the contrary, if many of your passing students struggle a lot with standard follow-up courses, you should probably be harsher.

In education, it is normally advisable to develop assessment prior to teaching. This allows the teacher to be sure that he or she is teaching to the assessment that they designed. When the assessment is made after teaching there is a risk that you will assess things you did not cover thoroughly or even assess things you never taught. Many scoff at this but it is common.

In terms of how much weight to give to various sections, this again is base on what you consider most important. Skills that are more important for students to show mastery should be worth more and lesser skills should be worth less. This is a modified application of bull's eye curriculum.

Lastly,all students should face the same conditions during the assessment. This means everyone has a cheat sheet and their formulas are marked or no one. If the conditions are different you cannot compare the results of one student to another, unless you are conducting an experiment and want to see if a cheat sheet makes a difference. If you can't compare results it is difficult to know who learned something and who didn't