When a flashcard app asks you to choose a "desired retention" target, it's asking a bigger question than it looks: How certain do you want to be that you'll remember this card when it comes back?
The default in most FSRS-based apps is 0.9 — a 90% probability. It's a sensible default, but it's not the right choice for every learner or every subject. Here's how to pick a better one.
What Desired Retention Actually Does
FSRS schedules each card for review at the moment your predicted recall probability drops to your target. Raise the target and intervals shrink. Lower it and intervals grow. Everything else — difficulty, stability, the order you see cards — is downstream of this number.
Concretely:
- 0.80 → long intervals, light workload, frequent forgetting.
- 0.85 → balanced toward workload reduction, some forgetting.
- 0.90 → the classic default. Strong recall, moderate workload.
- 0.95 → heavy workload, very strong recall.
- 0.97+ → exam-week territory. Not sustainable long-term.
The Workload Math
Here's the counterintuitive part: higher retention isn't always more efficient. Jarrett Ye's analysis on the FSRS4Anki wiki shows that the total number of reviews needed per memorized card over the long run is actually minimized somewhere around 0.85–0.90 retention, not at 0.95 or higher.
Why? Because at very high retention, you review each card so frequently that the total review count balloons faster than the retention gains. At very low retention, cards get forgotten and have to be relearned from scratch, which also costs reviews. The sweet spot for workload efficiency is in the middle.
Choosing by Goal
If you're building long-term knowledge (language learning, hobby topics, general knowledge): target 0.85–0.88. You'll forget occasionally, but your daily review load will feel sustainable for years.
If you're preparing for a dated high-stakes exam (USMLE, bar, professional cert): target 0.90 for most of the prep window, then raise to 0.93–0.95 in the final 4–6 weeks. This concentrates effort where it pays off most.
If you're maintaining already-learned material: target 0.85. You already know it; you just want to keep it from fading.
If you're a beginner who's never used SRS: start at 0.90 and don't touch it for three months. You need data before optimization is meaningful.
Signs You Should Change Your Target
- Your daily queue feels heavier than the material warrants → lower retention.
- You keep forgetting cards that feel important → raise retention, or reformulate the cards.
- Your exam is six weeks away and you're not confident → raise retention temporarily.
- You're burning out on reviews → lower retention before you quit SRS entirely.
One Mistake to Avoid
Don't change your retention every day. The scheduler needs accumulated reviews under a stable setting to give meaningful results. Pick a target, stick with it for at least two weeks, then evaluate.
Sources
- FSRS4Anki wiki on retention: github.com/open-spaced-repetition/fsrs4anki/wiki
- FSRS benchmark: fsrs-benchmark