Almost every app in this category runs on the same idea: show a word just before you’d have forgotten it, and each successful recall buys a longer gap. The principle has held up under a century of memory research: a landmark 2013 review rated its two ingredients, distributed practice and practice testing, as the only high-utility techniques of the ten it examined. Nobody owns it.
Which makes the algorithm nearly worthless for choosing between apps. Every serious one runs FSRS or an SM-2 variant, both freely available, and the gap between a good scheduler and an adequate one is smaller than the gap between using one and not. An app leading its pitch with scheduling sophistication is telling you about the part it didn’t have to build. What actually decides whether you’re still reviewing in three months comes down to four questions.
How much work does one card cost?
This is the whole thing, and it never appears in feature comparisons. Between meeting a word and having a reviewable card sits: look it up, pick the right sense of four, find an example sentence, open the app, make the card, paste both sides, choose a deck, save, and then try to remember what the paragraph was about. Call it ninety seconds. The real cost isn’t the time; it’s that every unfamiliar word now poses a decision, and the answer is often “not worth it.” You skip one, then the next, and within a week you’ve stopped capturing, left with forty cards from a burst of enthusiasm and no habit of adding more. The apps that survive real life are the ones where the card builds itself and you keep reading. If saving a word involves typing the translation yourself, you’ll stop.
Does the session arrive, or wait for you?
The quieter failure: two diligent weeks, a busy Thursday, a busy weekend, and you reopen the app to two hundred due cards. The backlog isn’t a scheduling problem, it’s a motivation problem, and it usually ends the habit. Studying should arrive at a time you chose, not sit as a chore to remember. Watch how an app behaves after you skip three days: some show you the pile; some cap the session and quietly redistribute. The second kind is still installed at Christmas.
Does it hold the words you actually meet?
There’s a real fork here. One kind of app ships a fixed course: the most common couple thousand words, the same for everyone. Genuinely good early on; Drops and Memrise do it well. The other kind holds the words you ran into, in a novel, a film, a conversation where you nodded and hoped. Those are worth more per card: you’ve met them in a context you remember, and a word that turned up in your life will likely turn up again. Most people past the beginner stage need the second kind and install the first, because the first is easier to market. It’s fine for them to be different apps.
Relatedly: a card showing cumbre → summit and nothing else teaches you to
recognise cumbre on a flashcard, not in a sentence. Cards worth reviewing
carry the sense you met, a real sentence, and the grammar that will bite you,
and whether the app builds that automatically takes you straight back to
question one. Good cards that cost ninety seconds don’t get made.
Does it work when the internet doesn’t?
Your best words turn up away from your desk: a train, another country, one bar of signal. An app that can’t save offline loses exactly the words you most wanted, silently.
When you shouldn’t use one at all
At the very beginning of a language you need contact (a course, a teacher, hours of comprehensible input), not a review app that lets you feel productive while avoiding them. For a deadline three weeks out, cram instead: the spacing research shows the optimal gap scales with how far away you need the memory, and spaced repetition deliberately tunes for months and years, not Friday’s exam. And if you’re not regularly reading, watching, or talking in the language, no scheduler will save you. These apps keep the words you meet; they’re not a way of meeting them.
The short version
Ignore the algorithm. Ask: how many taps to save a word; does the session arrive on its own; is the card worth reviewing; does it hold the words I actually meet. Get those right and the scheduling looks after itself.
The mechanics are in how spaced repetition actually works, the algorithm comparison in FSRS vs SM-2, the card-making problem in spaced repetition without making decks, and the reading workflow in how to save words while reading.
References
- Dunlosky, J. et al. (2013). Improving Students’ Learning With Effective Learning Techniques. PSPI.
- Cepeda, N. et al. (2006). Distributed practice in verbal recall tasks. Psychological Bulletin.
- Cepeda, N. et al. (2008). Spacing Effects in Learning: A Temporal Ridgeline of Optimal Retention. Psychological Science.