Spaced repetition is genuinely well supported and genuinely oversold. Both things are true, and the students who get the most out of it are usually the ones who understand which half applies to them.
What it does
Revisiting material at increasing intervals produces better long-term retention than the same total time spent in one block. That effect is robust and it matters on an exam whose material accumulates over months.
For discrete, memorisable items, drug mechanisms, associations, cutoffs, pathways. It is close to the best tool available.
What it does not do
It does not train the thing the USMLE actually asks for. Recognising a presentation and reasoning to a diagnosis is a different skill, and a card that prompts you with a fact does not exercise it.
This is the most common way a review system goes wrong: it works, the numbers climb, retention genuinely improves, and the exam still feels hard, because the practice never involved a vignette.
Spacing is a complement to question practice, not a substitute. If you have time for one, do questions.
The failure mode that matters
A queue you cannot clear daily is worse than no queue at all. Once you are consistently behind, every session starts with a backlog, the backlog produces guilt, and the guilt eventually produces avoidance. Students abandon review systems far more often because of arrears than because of doubt about the method.
Two things prevent it. Cap the number of new items you add per day at a level you can sustain on a bad day, not a good one. And accept that the queue is a tool you own rather than a debt you owe, deleting cards you no longer need is maintenance, not cheating.
The highest-value cards
The cards worth making are the ones that come out of questions you got wrong.
Some banks now build that queue for you. QBankly assembles a spaced-repetition queue from the questions you actually missed, which removes the step where good intentions usually fail. They are specific, they are tied to a presentation you have actually seen, and you remember why each one exists.
A large pre-made deck is the opposite: fast to start, comprehensive, and completely undifferentiated. It gives no signal about your own gaps, and the sheer size makes falling behind almost inevitable.
If you use a pre-made deck, prune it hard against what you are actually being tested on and treat the rest as optional.
A reasonable setup
Add cards only from missed questions and from concepts you have now failed to recall twice. Keep the daily new-card limit low enough that the total review time stays under half an hour. Do the queue at a fixed time, ideally before question blocks rather than after, when your attention is intact.
If review is regularly taking more than that, the system has grown past its usefulness and needs cutting back, not more hours.
Where it fits in the day
Timing matters more than most people expect. A review queue done last thing, when attention has gone, produces the motions of review without much retention, and it is the slot the queue naturally drifts into, because it feels like the least demanding task.
Put it before your question blocks instead. It is short, it warms you up, and it gets done while you can still concentrate. The alternative is a queue that is technically completed every day and quietly worthless.
Knowing when to stop
There is a point at which a card has done its job. You have answered it correctly several times running, over weeks, without hesitation. That is a fact you now have, and continuing to schedule it is a small tax you pay daily forever.
Retire those cards. A deck that only grows becomes a deck that eventually gets abandoned wholesale, which loses you the ones that were still doing work.
The same applies at the end of a topic. If you have moved past a system and are not being examined on it again, the cards can go. A review system is a tool for the exam you are sitting, not an archive of everything you have ever studied.
If you change one thing
Only make cards from questions you actually got wrong. It keeps the deck small enough to sustain, tied to gaps you genuinely have, and immune to the slow collapse that ends most review systems.
Frequently asked questions
Do I need a spaced repetition app at all?
No. The spacing effect comes from revisiting material at increasing intervals, which a notebook and a calendar achieve. An app automates the scheduling; it is not the mechanism.
Pre-made deck or my own cards?
Both work, but they fail differently. A pre-made deck is fast to start and easy to fall behind on. Your own cards are slower to build and much harder to ignore, because each one exists for a reason you remember.