How to make flashcards that actually work (and why most decks fail)

Flashcards have the best evidence base of any study tool and the worst hit rate in practice. Almost everyone tries them. Most decks are abandoned inside two weeks, and the ones that survive often produce students who can recite the deck and still cannot answer an exam question.

The technique is not the problem. The cards are. Here is what separates a deck that teaches you something from one that just makes you feel busy.

Rule 1: one card, one fact

The most common failure is the card that hides a list.

Front: What were the causes of the First World War? Back: Militarism, alliances, imperialism, nationalism, the assassination of Franz Ferdinand, the July Crisis, mobilisation timetables...

You will never grade this card honestly. You will remember four of the seven, decide that is close enough, and mark it correct. Do that for a month and you have trained yourself to half-know it.

Split it. One card asking how many causes there are. One card per cause, each asking what it contributed. A card asking which one historians consider proximate rather than structural. Every card should have exactly one right answer, short enough that you either produced it or you didn't — no judgement call.

Rule 2: the front must be answerable without the back

A card that reads "Mitochondria" on the front is not a question. It is a topic, and depending on the day you will answer "the powerhouse of the cell", or "has its own circular DNA", or "site of oxidative phosphorylation", and you will mark all three correct because they are all true things about mitochondria.

Make the front a real question with one target: which organelle carries its own circular DNA? What process happens on the inner mitochondrial membrane? If you cannot tell whether your answer was right, the card is broken.

Rule 3: never test recognition when you will need production

"Is enthalpy a state function? (Y/N)" is a card you will get right forever without ever learning anything, because you have a 50% chance before you know anything at all and 100% once you have seen it twice.

Anything you will have to produce in the exam — a definition, a formula, a mechanism, a date — should be a blank on the card, not a choice. Cloze deletion is the easiest way to enforce this: write the full sentence, then hide the one piece you need to be able to generate.

Rule 4: cards must be in your own words

Copying a definition verbatim off a slide produces a card you can recite and cannot use. Rephrasing it forces you to understand it first — which means the card is doing work at the moment you write it, not only when you review it.

There is one exception worth respecting: technical definitions where the exact wording is what's being marked, and legal or medical terminology where paraphrasing changes the meaning. Those you copy exactly. Everything else, put in your own language.

Rule 5: understand it before you card it

Flashcards are for retention, not comprehension. A card for a mechanism you never understood will be memorised as a string of syllables and will collapse the moment the exam rephrases it.

If you find yourself making a card for something you cannot explain to another person, stop making cards and go read about it. Then come back.

How spacing actually works

The second half of the technique is when you review, and this part is not optional — it is where most of the effect comes from.

The finding, replicated for over a century since Ebbinghaus, is that memory for an item decays predictably, and that a review just as it is about to fade produces far more durable learning than a review while it is still fresh. Reviewing something you already remember perfectly is nearly worthless. So the intervals stretch: a day, then three, then a week, then three weeks, then two months.

Doing this by hand is possible and miserable. This is the entire reason scheduling software exists — Anki, SuperMemo, and the spaced-repetition layer in any decent study app all implement some version of the same idea: cards you miss come back soon, cards you know go away for a long time.

Two practical points people get wrong:

• Do the reviews daily, in short sittings. Twenty minutes every day beats two hours on Sunday by a wide margin, because the schedule assumes you show up.
• Missing days is survivable; abandoning is not. A backlog looks catastrophic and mostly isn't — work through the oldest cards first and the queue collapses faster than you expect.

The failure that has nothing to do with the cards

Here is the honest reason most decks die: making good cards is slow. Getting a fifty-slide lecture into forty well-formed, one-fact, own-words cards takes somewhere north of an hour. Nobody has that hour every week, so the deck stops being updated around week three, and a deck that covers a third of the course is worse than useless — it gives you confidence in exactly the third you already knew.

That gap is what StudyPass is for. You upload the lecture slides or the reading, and the cards are generated from that file — your lecturer's terminology, your course's framing, not a generic textbook's. The rules above are still yours to enforce: read the generated cards, split the ones that hide a list, delete the ones testing something you don't yet understand, and rewrite anything that sounds like the slide rather than like you.

Generated cards you edit in ten minutes are a real deck. Perfect cards you never found time to write are not.

The checklist

Before a card enters your deck:

• One fact, one answer, no lists.
• The front is a question with a single target.
• It asks you to produce, not to recognise.
• It is in your words, unless the exact wording is the point.
• You could explain it to someone else right now.