The unit test is supposed to tell you whether the unit worked. In practice, by the time it arrives, it's too late to do anything with what it tells you. A misunderstanding from lesson two has had three weeks to calcify into a habit, and now you're reteaching a whole concept during test corrections instead of catching it with a five-minute fix back when it was cheap.
A mid-unit checkpoint quiz exists to move that diagnosis earlier, while there's still unit left to adjust.
What a checkpoint quiz is not
It's not a shrunk-down version of the unit test. If you just cut the test in half and give it at the midpoint, you get the same problem at a smaller scale - a grade, not a diagnosis. A real checkpoint is built to answer one question: *which specific misconception is still live in this room, and in whom?* That means it should be short (8-10 questions, 15 minutes), ungraded or low-stakes, and built entirely around the one or two concepts most likely to break the rest of the unit if left unfixed.
Build it backward from "what would break the ending"
Before writing a single question, ask: if a student still doesn't get X by day 10, what will they get wrong on everything after that? That's your checkpoint's real content. Everything else in the unit so far is probably fine to skip.
- Unit on cellular respiration: if students still think mitochondria "make" energy from nothing rather than converting it from glucose, every later question about ATP yield will be wrong for the wrong reason. Checkpoint targets that specifically.
- Unit on the American Revolution's causes: if "taxation without representation" is memorized as a slogan but nobody can explain *why* it mattered to colonists, every later essay prompt about grievances will be surface-level. Checkpoint targets the *why*.
Use AI to generate the wrong-answer options, not just the right ones
The diagnostic power of a checkpoint quiz lives almost entirely in the distractors - the wrong multiple-choice options. A generic wrong answer tells you nothing. A wrong answer that matches a specific, common misconception tells you exactly who has that misconception and lets you group them for a five-minute reteach.
Try a prompt like this:
- "I'm teaching [concept] to [grade level]. Students commonly confuse it with [related idea]. Write 5 multiple-choice questions where one wrong answer specifically reflects that confusion, one reflects a surface-level partial understanding, and only one is fully correct. Flag which distractor maps to which misconception."
Read the draft against your actual class before using it - AI doesn't know which specific misconception showed up in third period yesterday, but it's a fast way to generate options that are diagnostic instead of just plausible.
What to do with the results the same day
The whole value of a checkpoint collapses if it sits in a folder until the weekend. Score it fast (or have students self-score against a projected answer key with the reasoning behind each option), sort into three rough groups - solid, partial, still confused - and give the "still confused" group five minutes of targeted reteach before the next lesson builds on top of the same shaky foundation.
Where our tool fits
Our [quiz generator](/quiz-generator) builds a question set from your unit's learning target directly, and our [learning objectives](/learning-objectives) tool helps you nail down the specific, gradable target a checkpoint should be built around before you write a single question - useful when "do they get photosynthesis" is really three or four separate sub-skills that each need their own check.
A checkpoint that catches a misconception on day four costs you five minutes. The same misconception caught on the unit test costs a full reteach unit and a grade nobody's happy with.
