The spacing effect — that reviewing material at increasing intervals beats massed repetition — has been replicated since Ebbinghaus described the forgetting curve in the 1880s. It is about as solid as findings in memory research get, and the software implementing it is free.
Which makes it worth asking why so many learners abandon it after a few months. The failures are consistent, and almost none of them are about the algorithm.
Failure one: too many new cards
Every new card creates a stream of future reviews. Adding twenty a day feels manageable in week one and produces a daily review load of several hundred by month three.
| New cards per day | Steady-state daily reviews (approx.) |
|---|---|
| 5 | 50 |
| 10 | 100 |
| 20 | 200 |
| 30 | 300 |
The number to choose is not the one you can do today. It is the one you will still do on a bad day in three months, because the system only works if it is never abandoned.
A missed week is not a disaster; quitting is. Returning to 800 overdue reviews is demoralising enough that most people delete the deck instead. The fix is to reduce the daily new-card limit to zero for a while and simply clear the backlog — the reviews are the value, not the additions.
Failure two: cards that test the wrong thing
A card should test one item, recalled in the direction you need it.
| Poor card | Better card |
|---|---|
| A word with six listed meanings | One meaning, in one sentence |
| An isolated word | The word in a sentence you met it in |
| A long phrase to recall wholesale | A sentence with one word blanked out |
| Two-way cards for every item | Recognition for most, production for the useful ones |
Two-way cards double the workload for a benefit that is often not needed. Most vocabulary only needs to be recognised when reading or heard when listening; a smaller active set needs to be produced.
Failure three: cards made from lists
Cards built from a frequency list carry no context and no personal hook. Cards built from something you were reading arrive with a sentence, a situation and a reason you wanted the word.
The practical rule: make a card when a word blocked you. If you were reading and had to stop, that word has earned a card. If it came from a list and you have never encountered it, it has not.
Failure four: expecting it to teach
Spaced repetition consolidates. It does not introduce. A card for a word you have never met and do not understand becomes a meaningless string you learn to recognise on a card and cannot use anywhere else — the well-known problem of items that are known only inside the app.
| Good at | Bad at |
|---|---|
| retaining vocabulary you have met | teaching new concepts |
| fixing a writing system | grammar in use |
| holding a language you are not using | listening comprehension |
| exam-specific content | real-time production |
What a good routine looks like
| Step | Detail |
|---|---|
| Read or listen first | encounter words in context |
| Note what blocked you | with its sentence |
| Make a small number of cards | ten or so from a session |
| Review daily, briefly | fifteen minutes is usually enough |
| Delete cards that keep failing | remake them differently, or drop them |
That last step is the one people resist. A card you have failed fifteen times is a badly made card, not a difficult word. Deleting it and rebuilding it around a clearer context costs less than the reviews it will otherwise consume indefinitely.
Where it stops
Vocabulary retention is a real bottleneck and spaced repetition addresses it well. It does not address the others. A learner with 5,000 mature cards who has never held a conversation has built exactly one of the several abilities that make up knowing a language — and the review time that produced it eventually competes with the reading and speaking that would build the rest.
Look words up fast, then review
Good cards start with a good lookup. NDT Studio's free offline dictionaries cover 45 pairs with no connection required.
Open the dictionary appFrequently asked questions
How many new cards a day should I add?
Fewer than most people start with. Each new card generates reviews for months, so twenty new cards daily settles into a substantial permanent review load. Ten to fifteen is sustainable for most learners; the number that matters is the one you can still face in three months.
Should I make my own cards or use shared decks?
Your own, in most cases. The act of deciding what goes on a card is itself learning, and cards drawn from things you actually read carry context a shared deck cannot. Shared decks are reasonable for standardised content such as a writing system or a fixed exam list.
Is spaced repetition enough on its own?
No. It builds recall of items you have already met and understood. It does not teach grammar in use, listening, or the ability to produce language in real time. It is a maintenance and retention tool that works best alongside reading, listening and speaking.