The objection to oral assessment is never that it does not work. It is that the arithmetic is impossible. So let us do the arithmetic properly, because the impossible version usually comes from a plan nobody would actually run.
Stop planning the maximal version
The default mental image is a fifteen minute exam for every student on the whole syllabus, conducted by you, at the end of term. That is seventy five hours for two hundred students and it will not happen. Almost every workable design cuts one of those four variables.
The most useful cut is scope. An oral check does not have to cover the course. It has to cover the thing the student handed in. Five focused minutes on one submission answers the question you actually have, which is whether this work belongs to this person and whether they understand it.
A design that survives contact
- Attach the check to a submission, not to the syllabus. The student defends what they wrote last week, not everything since September.
- Five to seven minutes. Long enough for two follow-ups, short enough that the tail is bearable.
- Everyone, not the suspicious. Universal checks are cheaper politically and produce a comparison baseline. A targeted check is an accusation and will be argued about.
- Two or three assessed submissions per term, not eight. The signal saturates faster than people expect.
- Write the questions from the submission itself. Generic questions produce generic answers and teach students to prepare a speech.
The point of the check is not to catch anyone. It is to make understanding the thing being graded, which changes how students work weeks before the check happens.
The scheduling mistakes
Three failures show up over and over.
The first is booking everything in the final fortnight. Assessment stacked at the end of term collides with every other course doing the same, and a missed slot has nowhere to go. Spread checks across the term and attach them to the work as it lands.
The second is treating a missed slot as a failure. Students get sick, lose wifi, and mix up time zones. Build a rebooking path from the start or you will spend the term adjudicating individual cases by email.
The third is making the first one high stakes. Whatever format you use, students should meet it once with nothing riding on it. The first oral exam of someone's life should not be worth twenty percent.
Accommodations are not an afterthought
Decide in advance what happens for a student with an anxiety diagnosis, a speech difference, a hearing impairment, or no private room. Extra time and reduced time pressure are the usual first answers. An alternative format, run by you directly, should exist and should not be a downgrade in the record. Announce this in week one. Students who need it should not have to ask you for permission to exist.
Where the time actually goes
If you run these yourself, budget the exam time plus roughly a third again for setup, notes and the students who need rescheduling. If you use software to run the conversation, the time moves from conducting to reviewing, which is faster but not free. Reviewing a recorded five minute check with a transcript and a summary runs about a minute per student once you have a routine, plus real attention on the handful that look unusual.
That is the honest number. Two hundred students at a minute each is three and a half hours per assessed submission. Compared with marking two hundred essays properly, most instructors find that a trade they will take.
If you want to see the review side of it before deciding, there is a real interview and report at the demo.