Most people who try Anki give up within a few weeks. Spaced repetition itself is not the problem: it is one of the best-supported ideas in learning research. What usually goes wrong is the setup. Too many new cards, cards that are really paragraphs, and a review pile that grows faster than anyone can clear it.
This guide covers what the method does, what the evidence says, and the small number of settings and habits that decide whether Anki becomes a ten-minute daily routine or a constant source of guilt.
The idea in one paragraph
You forget things along a fairly predictable curve, and every time you successfully pull a memory back out, that curve gets flatter. Spaced repetition schedules each review close to the point where you would otherwise forget: soon at first, then after a few days, then weeks, then months. Anki does the scheduling for you. You see a prompt, try to recall the answer, flip the card, and tell Anki how it went. Anki decides when you will see that card again.
What the research actually shows
Hermann Ebbinghaus started all of this in the 1880s by memorizing lists of nonsense syllables and measuring how much faster he could relearn them later. His “savings” fell to roughly 58% after twenty minutes, about a third after a day, and around a fifth after a month. When Jaap Murre and Joeri Dros repeated the experiment in 2015, they got a very similar curve. Meaningful material fades more slowly than nonsense syllables, so the exact numbers will not match your biology notes, but the shape holds: a steep drop at the start, then a long, slow tail.
Two well-documented effects explain why Anki works.
The spacing effect. Spreading study over time beats cramming the same total hours into one sitting. A 2006 review by Nicholas Cepeda and colleagues pooled 317 experiments and found the advantage across ages and types of material. Their follow-up in 2008 taught facts to more than 1,350 people and varied both the gap between sessions and the date of the final test. The best gap depended on how long the information had to last. For a test one week away, the ideal gap was about 20 to 40% of that week, so a day or two. For a test a year away, it was about 5 to 10% of the year, which works out to roughly three to five weeks. Growing intervals are exactly what Anki’s scheduler produces.
The testing effect. Retrieving information strengthens memory more than reading it again. In a 2008 study published in Science, Jeffrey Karpicke and Henry Roediger had students learn Swahili–English word pairs. Students who kept being tested on every pair remembered about 80% of them a week later. Students who stopped being tested on a pair once they had recalled it, even though they kept studying it, remembered only about a third. Extra studying added almost nothing. Extra retrieval made the difference. The students’ own predictions of how well they would do had no relationship to how well they actually did, which is a good reason not to trust the feeling that rereading “worked.”
When John Dunlosky and colleagues reviewed ten common study techniques in 2013, only two received their highest rating: practice testing and distributed practice. Highlighting and rereading, the two most popular habits among students, were rated low. Anki is essentially those two high-rated techniques packaged into one app.
How Anki decides when you see a card
After you flip a card, you press one of four buttons:
- Again: you did not remember it, or you got it wrong.
- Hard: you remembered, but it took real effort.
- Good: you remembered it normally.
- Easy: you knew it instantly and it felt trivial.
Most of your answers should be Again or Good. A common mistake is using Hard as a softer version of Again when you actually forgot. That feeds the scheduler wrong information, and the card comes back later than it should.
Older versions of Anki used a scheduler based on SM-2, an algorithm Piotr Woźniak designed for SuperMemo in the late 1980s. Since version 23.10, Anki also includes FSRS, an open-source scheduler that learns from your own review history to predict when you are likely to forget each card. You can turn it on in the deck options. Its main setting is desired retention, the share of due cards you want to remember when they come up. The default of 0.90 is a sensible place to stay. Pushing it to 0.95 or higher makes the daily workload climb steeply for a small gain.
Where the time goes: new cards decide everything
A new card is not one card. It is a stream of future reviews spread over the following months. As a rough rule of thumb, once you have been using a deck for a couple of months, expect about five to ten reviews a day for every new card you add per day.
That is why Anki’s default of 20 new cards a day catches so many people out. Twenty new cards today means somewhere around 100 to 200 reviews a day a few months from now. At eight to ten seconds a card, that is 15 to 30 minutes, every single day, before you have learned anything new.
Some starting points that hold up:
- Casual learning (a language on the side, general knowledge): 5 to 10 new cards a day.
- Serious exam preparation: 10 to 20 new cards a day.
- Very heavy workloads (some medical and law students): more is possible, but only with a clear plan for the review time it will cost.
Pick the number by working backwards from the time you would still be willing to spend on a bad day. If the daily total starts creeping past that, lower the new-card limit. Do not plan on pushing harder for a week until it settles down. It will not settle down on its own.
Writing cards that actually stick
Bad cards cause more Anki failures than anything else. A few principles do most of the work.
One question, one answer. Woźniak called this the minimum information principle. Compare:
Q: What is photosynthesis?
A: A paragraph about light, chloroplasts, water, carbon dioxide, glucose and oxygen.
with three separate cards:
Q: Where in a plant cell does photosynthesis take place? → A: In the chloroplasts.
Q: Besides light, what are the two main inputs of photosynthesis? → A: Carbon dioxide and water.
Q: What are the main products of photosynthesis? → A: Glucose and oxygen.
The big card fails every time you forget any one part, so you never learn it reliably. The small cards succeed or fail independently, and the scheduler can space each one properly. Three small cards are faster to review than one big card you keep missing.
Avoid lists on a single card. “Name the seven layers of the OSI model” is a card you will fail for months. Break it up, or use overlapping cloze deletions so that each card hides one or two items while showing the rest.
Use cloze deletions for sentences. A note like The {{c1::hippocampus}} is essential for forming new {{c2::declarative}} memories creates two cards: one hides the brain region, the other hides the type of memory. It is quicker to write than separate question-and-answer pairs, and it keeps the context. Just do not hide half a paragraph in a single cloze.
Use image occlusion for diagrams. Anatomy, maps, circuit diagrams and labeled charts work far better as images. Since version 23.10 this is built into Anki: choose the Image Occlusion note type, draw boxes over the labels, and Anki creates one card per label.
Add a context cue. “Which enzyme does it block?” is ambiguous six weeks later. “Pharmacology, lisinopril: which enzyme does it block?” is not. A short prefix and a tag take seconds and save a lot of confusion during reviews.
Understand first, then memorize. Anki is excellent at helping you keep what you already understand. It is poor at teaching you something for the first time. If a lecture confused you, turning it into thirty cards will just give you thirty confusing cards. Read, ask questions, make rough notes, and only then pull out the specific facts you want to keep.
Shared decks: when they help and when they do not
Making your own cards is part of the learning, because choosing what to keep and how to phrase it forces you to process the material. Still, “never use shared decks” is too strong. Large, well-maintained community decks built around a standard curriculum, such as medical licensing exams, are used successfully by many students. What tends to fail is downloading a random deck on a subject you are learning from a different book, with someone else’s wording and priorities. If you use a shared deck, suspend the cards you have not covered yet and unsuspend them as your course reaches them.
A daily routine that survives real life
The schedule assumes you show up most days, so the main goal is a routine that does not depend on willpower. Tie your reviews to something you already do every day, like your first coffee, the commute or right after lunch. The phone apps are good for spare minutes. Missing one day is fine. Missing a week is where the pile, and the temptation to quit, comes from.
People often ask whether music helps. There is no good evidence that background music improves flashcard recall. If you like having sound while you review, keep it instrumental and quiet. Speech and lyrics are a classic distraction in memory research, known as the irrelevant speech effect. That is one reason instrumental genres like lofi work as background for many people: there are no words competing with the ones on your cards. Plenty of people still do better in silence, and that is just as valid.
When the review pile gets out of hand
It happens to almost everyone eventually: an exam week, an illness, a holiday. Here is a way back that does not mean starting over.
- Stop adding new cards until the daily count is back to normal. Set new cards per day to 0 for a while.
- Spread the backlog out. In the Browser, search for
is:due, select the cards, and use Cards → Set Due Date with a range such as0-14. Anki spreads them over the next two weeks instead of showing them all tomorrow. - Keep grading honestly. Pressing Good on cards you have forgotten makes the pile disappear today and come back worse later.
- Suspend what no longer matters. If the exam is over and a deck is not useful anymore, suspend it rather than deleting it, in case you want it back later.
One setting makes backlogs worse without anyone noticing: a low maximum reviews per day. It does not reduce your workload. It only hides the overdue cards until they pile up. If you need to cut the daily load, lower the number of new cards instead.
Getting started this week
- Install it. Anki is free on Windows, Mac and Linux (apps.ankiweb.net), and AnkiDroid is free on Android. On iPhone and iPad the official app, AnkiMobile, is paid, and the purchase helps fund Anki’s development. A free AnkiWeb account keeps everything in sync.
- Set it up. In the deck options, turn on FSRS, leave desired retention at 0.90, and set new cards per day to about 10.
- Make your first fifty cards from material you already understand, following the principles above.
- Review every day for two weeks before you judge it. The early reviews feel almost too easy. The real payoff shows up around the second month, when cards from the first week come back and you still know them without having opened the notes.
Spaced repetition will not replace understanding, practice problems or sleep. What it does well is make sure that what you have already learned is still there when you need it, for about the cost of a coffee break a day.
Sources
- Ebbinghaus, H. (1885). Über das Gedächtnis. English translation: Memory: A Contribution to Experimental Psychology (1913).
- Murre, J. M. J., & Dros, J. (2015). Replication and analysis of Ebbinghaus’ forgetting curve. PLOS ONE, 10(7), e0120644.
- Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380.
- Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., & Pashler, H. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19(11), 1095–1102.
- Karpicke, J. D., & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319(5865), 966–968.
- Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58.
- Salamé, P., & Baddeley, A. (1982). Disruption of short-term memory by unattended speech. Journal of Verbal Learning and Verbal Behavior, 21(2), 150–164.
- Woźniak, P. (1999). Effective learning: Twenty rules of formulating knowledge. SuperMemo.
- The Anki Manual, including the sections on deck options, FSRS and the Browser.
