Newly iconNewly
Articles · App ExamplesUpdated September 2026

The flashcard app that knows when to ask again.

A flashcard app is the smallest idea in educational software: a front, a back, a tap to flip. The cards are not the product. The product is the decision the app makes after you answer: when does this card come back? Get that decision right and a deck of a few thousand Spanish words fits into a short daily session; get it wrong and you own a shuffled pile with a progress bar. That decision is also the quiet test behind most educational app development: whether the software knows something about the learner that a paper deck does not.

This page is the plumbing: what a card actually stores, the three ways apps schedule reviews (a shuffle, Leitner boxes, and the SM-2 algorithm most spaced repetition apps descend from), what changes for a math flashcard app, a Spanish flashcard app or a letter flashcard app for a four-year-old, and when the honest answer to "best flashcard app" is one you build yourself.

See what a card has to store

The short version

Three parts, and the third one is the app.

A flashcard app is a card model (a note with a front, a back and extras, plus a note type that decides how many cards the note produces), a review loop (show the front, let the learner recall, reveal, grade) and a scheduler that turns the grade into a due date. Most apps that say spaced repetition run a descendant of SM-2, a published 1987 algorithm with a formula you can read in five minutes; some run Leitner boxes; a surprising number just shuffle. One question tells them apart: does every card carry its own due date?

The card model

What a flashcard app has to store.

The screen shows two sides. The database holds three kinds of rows, and every feature people ask for later (reverse cards, cloze deletions, per-card statistics, an honest streak) depends on that split being made at the start:

RowFieldWhy it exists
NoteFront, back, extras: audio, image, example sentence, hintContent is written once. Cards are generated from it, so fixing a typo fixes every direction at the same time.
NoteNote type: basic, basic and reversed, clozeDecides how many cards one note produces: one, two, or one per blank.
NoteDeck and tagsThe deck picks what is studied together; tags cut across decks (chapter 4, verbs, exam).
CardDirection or cloze indexSpanish to English and English to Spanish are two cards with two separate histories.
CardRepetition countSuccessful reviews in a row. SM-2 uses it to pick the first two intervals.
CardEase factorHow fast this card's interval grows. Starts at 2.5 under SM-2 and never drops below 1.3.
CardInterval and due dateThe scheduler's output. Today's queue is every card whose due date has arrived.
CardLapses and a suspended flagCounts restarts. A card that keeps lapsing is a leech and should be rewritten, not drilled harder.
Review logCard, timestamp, grade, interval before and after, time takenThe only place the truth lives. Streaks, statistics and even the schedule can be recomputed from it.

The review loop

One session, step by step.

  1. 1Build today's queue: every card whose due date is today or earlier, plus new cards up to the daily cap.
  2. 2Show the front only. Nothing on the screen should hint at the back.
  3. 3The learner recalls, then reveals. Recall before reveal is the whole point, so a swipe must not skip it.
  4. 4Take a grade. SM-2 uses 0 to 5, Anki uses four buttons, and a math drill can grade by response time instead.
  5. 5The scheduler updates the repetition count, ease factor, interval and due date for that one card.
  6. 6Append one review log row. Never overwrite the previous one.
  7. 7Cards graded under 4 come back later in the same session until they score at least 4.
  8. 8Stop when the queue is empty, and show how many cards are due tomorrow before anyone adds more.

The scheduler

Shuffle, Leitner or SM-2: how a flashcard app decides when to ask again.

Four families cover almost every flashcard app and spaced repetition app on the market, and they differ in one measurable way: what the app remembers about each card between sessions.

ApproachWhat it stores per cardWhat a grade doesWhere it breaks
Random shuffleNothing, or a known / unknown flagMoves the card between two pilesEvery card is treated the same, so a deck of a few hundred cards becomes a daily slog of things you already know
Leitner boxesA box numberRight moves the card up a box, wrong sends it back to the first boxIntervals are fixed per box, so an easy card and a hard card in the same box come back on the same day
SM-2Repetition count, ease factor, interval, due dateCorrect grows the interval by the card's ease; a fail restarts it; the ease drifts with every gradeCards that keep failing sink to the 1.3 floor and return often; the fixed 1 and 6 day openers do not suit every learner
FSRSDifficulty, stability and retrievability, fitted to the learner's own review historyUpdates the memory state; the learner sets a target retention instead of tuning intervalsNeeds a review history to fit, and is harder to explain to a learner than wrong goes back to box one

SM-2, stated plainly

The whole algorithm fits in six rules.

SM-2 is the algorithm SuperMemo shipped in versions 1.0 to 3.0 between December 1987 and March 1989, written up in Piotr Wozniak's 1990 thesis and published in full on the SuperMemo site. It is short enough to state without simplifying:

  • Every new card gets an ease factor of 2.5.
  • After the first correct review the interval is 1 day; after the second, 6 days; after that, the previous interval multiplied by the ease factor, rounded up to a whole day.
  • Each review is graded 0 to 5: 5 perfect, 4 correct after a hesitation, 3 correct with serious difficulty, 2 wrong but the answer felt easy once shown, 1 wrong and only remembered when shown, 0 complete blackout.
  • After every review the ease factor becomes EF + (0.1 - (5 - q) x (0.08 + (5 - q) x 0.02)). A 5 adds 0.1, a 4 changes nothing, anything lower subtracts. It never drops below 1.3.
  • A grade below 3 restarts the card at the 1 day interval.
  • Cards graded below 4 are repeated again the same day until they score at least 4.

Rules, grade descriptions and formula from Wozniak's published description of SM-2. Later SuperMemo versions moved on; SM-2 stayed in circulation because it was published and small.

Try it

Grade one card and watch SM-2 schedule it.

Each button is one review of the same card, graded on the 0 to 5 scale from the published algorithm. The intervals below are exactly what SM-2 would produce, starting from an ease factor of 2.5.

Next review

Grade the card to see its first interval. Under SM-2 the first correct answer schedules it for tomorrow, the second for six days later.

0 of 10 reviews. Intervals round up to whole days, as the source specifies.

What the popular apps actually run.

Anki documents its scheduler in its own FAQ: as of Anki 23.10 it offers two, one based on SM-2 and one called FSRS. Its SM-2 variant departs from the original in ways worth copying: configurable learning steps instead of the fixed 1 and 6 day openers, four answer buttons instead of six grades, and no ease penalty for failing a card while it is still in learning, which avoids the low interval hell the original can produce for a badly written card. FSRS describes each card with three numbers, difficulty, stability and retrievability, fits its parameters to the learner's own review history, and lets the learner set a target retention instead of tuning intervals. Read Anki's algorithm FAQ

Name it honestly. If the app you are evaluating, or the one you are building, has no per-card due date, it is a shuffle with labels and should say so. The label changes what the learner does: with a real scheduler you finish the queue and stop; with a shuffle you keep going and burn out on cards you already know.

The evidence

Do flashcards actually work?

Yes, for two reasons that live in the app rather than in the cards. In a 2013 review for the Association for Psychological Science, Dunlosky, Rawson, Marsh, Nathan and Willingham assessed common study techniques and found two of high utility across ages, ability levels and subjects: practice testing, and distributed practice, meaning study spread out over time. Techniques students reach for by habit, rereading and underlining among them, came out with surprisingly low utility. See the APS summary of the review

A flashcard app is practice testing (recall before reveal) plus distributed practice (the scheduler) in one loop. Which is also why a deck that is flipped instead of recalled, and crammed in a single sitting, loses most of the benefit while looking exactly the same on screen.

review when recall drops to herehighlowday 0day 1day 7day 22day 60
SM-2 with a grade of 4 at every review: the card returns on day 1, day 7, day 22 and day 60. Each interval is the previous one multiplied by an ease factor that stays at 2.5, rounded up, so the same amount of forgetting takes longer each time.

By subject

Math, Spanish or letters: the same flashcard app, three different jobs.

Most searches around flashcard apps are for a subject, and each subject bends the card model somewhere. The scheduler stays; the cards, the grading and the screen do not.

Facts, generated by rule

Math flashcard app

Cards
Generated from a rule (every fact from 2 x 2 to 12 x 12), not typed one by one. The note is the rule; the cards are its products.
Grading
Response time, not a self-rating. Fluency means fast and correct, so the app times each answer and treats a slow correct answer as a 3, not a 5.
Scheduler
SM-2 works for facts, but a fact family (the sevens) is studied as a block, so the new-card cap should apply per block.
Screen
A number pad and a typed answer, not a flip. Self-grading a math fact is how children grade themselves generously.

Authored notes, two directions

Spanish flashcard app

Cards
Notes with the article on the front (la mesa, not mesa), audio and an example sentence. Each note produces a recognition card and a production card with separate histories.
Grading
A 0 to 5 self-rating for recall, or a typed answer with accent-tolerant matching so a phone keyboard does not fail a word the learner knows.
Scheduler
SM-2 or FSRS. The production direction earns lower grades and drifts to shorter intervals on its own, which is correct.
Screen
Audio before text on the reveal, and a leech list, because the cards that keep failing are usually badly written cards.

A tiny deck, a hard interface

Letter flashcard app

Cards
26 letters, upper and lower case, plus the letter sound. That is the whole deck, so scheduling is small and interaction is everything.
Grading
The adult grades, or the child taps the matching letter out of three. A self-rating means nothing at four years old.
Scheduler
Leitner boxes are enough for a deck this size and far easier to show to a parent than an ease factor.
Screen
Full-width tap targets, audio on every card, no text menus, and a parent gate in front of settings.

Two neighbours are different jobs wearing a similar name. If a tutor assigns the deck and walks through the failures live, the review data belongs inside a tutoring app, where the deck is one tab and the session notes are another. And due counts tell a learner what to review, not when the lesson is; if the real need is periods, rooms and reminders, that is an app for class schedule, and bolting a deck onto it makes both worse.

Best flashcard app

Which flashcard app is best? It depends where the cards come from.

"Best flashcard app" has a different answer for a medical student with thousands of cards, a class sharing one set, and a parent with a four-year-old. The honest comparison is by scheduler, by who owns the cards, and by what it costs. Prices and tiers below were checked on 20 September 2026.

AppSchedulerWhere the cards come fromCostFits when
AnkiSM-2 based by default, FSRS as an optionYour own notes, synced free across devicesDesktop free; AnkiMobile for iOS is a one-time $24.99 on the US App StoreHigh volume, self-directed study where you want control over every interval
QuizletSession study modes: Learn rounds and practice testsSets you or a class createFree with ads and monthly limits; paid Plus tiers remove ads and raise or remove the limitsA class sharing one set, mixed study modes over a strict schedule
BrainscapeSpaced repetition, in the vendor's own termIts decks and yoursFree Basic tier; paid Pro plan billed yearlyA ready-made deck for a standard subject
Built for your processWhichever you specify: SM-2 is the safe default, Leitner for tiny decksWhatever shape your cards need: generated, tutor-assigned, picture-onlyFrom $25 a month for the build tooling, plus your timeThe card model or the card owner does not fit the three rows above

Anki's price and free sync from the AnkiMobile listing on the App Store; its scheduler from the Anki FAQ linked above. Quizlet and Brainscape tiers from their own pricing pages on the same date.

Which row fits depends less on features than on who owns the cards. A class that needs one shared set does not need SM-2; a learner with thousands of cards needs little else. Even a build-it-yourself platform like Newly starts at $25 a month, and that buys the tooling, not the deck: the card model, the grading rule and the scheduler still have to be decided by you, which is what the rest of this page has been about.

When to build your own

Building a flashcard app that fits your cards.

The reason to build a flashcard app rather than install one is rarely the scheduler, which is published and small. It is the card model: facts generated by rule, decks a tutor assigns and audits, cards a child can use without reading, or a deck that has to live inside your own course app under your own name. When the cards or their owner do not fit an off-the-shelf tool, teams end up building their own, and that is what Newly is for: describe the note fields, the card directions, the grades and the scheduler, and it builds a real native app around them. It builds the iOS app and uploads it to App Store Connect, you submit for review yourself, and on v2 there are no Android release builds, so plan a Play Store launch separately.

Before you describe the first screen

  • Write the scheduler out in rules, not in a label. Ease factor 2.5, intervals 1 and 6 then previous times ease, grades 0 to 5, floor 1.3, a fail restarts.
  • Decide the note types and how many cards each one produces. Reversed cards and cloze deletions are cheap now and expensive later.
  • Pick the grade: a self-rating, a typed answer, a timed answer, or an adult tapping for a child.
  • Set the daily new-card cap and show tomorrow's due count on the home screen.
  • Read what came back and name it honestly. If every card has a due date that moves, it is spaced repetition. If not, it is a shuffle, and the home screen should not say otherwise.

One more decision belongs before the first screen. Reviews happen on trains, in queues and in waiting rooms, so the queue has to be computed on the device and the review log synced when a connection returns. That changes how the app stores data, and it is a separate topic: offline app functionality is worth reading before you commit to how cards and reviews are saved.

Sources

FAQ

Flashcard apps, answered.

A flashcard app stores notes (a front, a back, and extras such as audio or an example sentence), turns each note into one or more cards, and runs a review loop: show the front, let you recall the answer, reveal the back, take a grade. The part that makes it more than a shuffled deck is the scheduler, which turns your grade into a due date for that card. When every card carries its own interval and due date, the app can show you only what is due today and let the easy cards drift out to weeks or months.

Start with the scheduler, not the deck.

A flashcard app earns its place with one decision made well: when to ask again. Write the card model and the rules down, then build the app around them in Newly.

Start building your flashcard app