A student opens Anki in late March, six weeks before AP Bio. They've built a 700-card deck. Reviews look fine for the first two weeks. Then they notice something weird: cards they crushed yesterday are scheduled to come back in October. The exam is May 12.
This isn't a bug in Anki. It's the design. Classic spaced repetition assumes you want to remember the material indefinitely, so every successful review pushes the next one further out. If you have a deadline — and most students do — that assumption is the wrong one.
What classic SRS optimizes for
Piotr Wozniak built SuperMemo in the 1980s to solve a specific problem: how do you remember a foreign language for life? The answer turned out to be intervals that grow without bound. Easy cards eventually space out to months, then years. The math works because the goal is open-ended retention.
This is also what SM-2 (the algorithm Anki ships with) does by default. Hit Easy on a card a few times and the next interval can be six months out. Great for vocabulary you'll use for decades. Wrong for a test in 30 days.
What changes when you have an exam date
The optimization target flips. You no longer want long-term retention; you want maximum recall on a specific calendar day. Cepeda et al.'s 2008 study gave us the rule of thumb: the optimal gap between study sessions is roughly proportional to how long you want to remember the material. If your retention interval is 25 days, your spacing should compress as you approach that date, not stretch. Cards you've nailed should still come back inside the exam window — a 6-month interval is wasted on an AP test 30 days away.
How exam-date-aware scheduling works
The FSRS algorithm tracks each card's stability — roughly, the time it takes for retrievability to drop from 100% to 90%. Stability grows with each successful review. In an open-ended schedule, that growth is the only signal: bigger stability, longer next interval.
In an exam-date-aware schedule, the algorithm caps intervals so every card lands at high recall on the target date. Cards with high stability still get longer gaps, just clipped to fit inside the deadline. Cards with low stability cycle faster. As you get closer to the exam, intervals contract — every card you've learned gets one final pass close to the test.
StudyLess wires the exam date directly into the scheduler. The cards know your test is May 12, so they schedule themselves around that date. We covered the underlying mechanics in how spaced repetition works; this post is about how to set the inputs.
How far out should you start
For a single AP, IB, or SAT subject, four to eight weeks of daily review is the sweet spot. Less than four weeks: you can still benefit, but the deck has to be smaller and reviews have to be daily — don't try to learn 800 brand-new cards in 14 days. More than eight weeks: you'll be fine, but the first month or two will look closer to a normal long-term schedule, with the compression hitting in the last 3–4 weeks.
The wrong move is starting with no end date and trying to convert later. Set the exam date when you build the deck, even if it's two months out.
Set the date a few days early
Whatever the actual exam date is, set the date in your app slightly earlier. Three to seven days early is the sweet spot.
Two reasons. First, you want a buffer to take a mock exam without your flashcard schedule still adding new reviews on top. The day or two before a mock should be light. Second, the schedule peaks recall on the date you set. Setting it a few days early means you're still at peak recall on the actual test day, with a cushion for the inevitable last-minute cramming and nerves.
If your AP Bio exam is May 12, set the date as May 8 or May 9. Use the buffer for one final mock and a light review.
What to do per phase
The exam-date schedule moves through three rough phases. You don't manage them — the scheduler does — but knowing them helps you trust what's happening. Early on, mostly new cards plus the first review pass; queues are heavy on new material. In the middle (2–3 weeks out), mature cards cycle at intermediate intervals and the queue stabilizes around 20–30 minutes a day. In the final 7–10 days, every card passes back through, sometimes more than once — this is where the exam-date optimization earns its keep.
A normal SRS schedule never gets to that compression phase. It just keeps stretching intervals into next year, and you walk into the test with cards "scheduled" two months out that you haven't actually seen since March.
Don't change the date once you've started
The temptation to push the date out — "I'll take it next month, I'm not ready" — breaks the schedule. The algorithm has been compressing reviews around the date you set. If you push it out, intervals lengthen, and cards that should have come back tomorrow won't surface for weeks. If you genuinely need to delay, do it once and commit. And if you're using mock exam scores to decide, make that call by week 3 of prep, not the week of the test.
What about multi-exam stretches
A student taking AP Bio May 12, AP Chem May 14, and AP Calc BC May 18 has a four-week exam window, not a single date. One deck per subject, each with its own exam date — cards for Bio peak May 12, cards for Calc peak May 18. Don't combine subjects into a single super-deck with a single date; the schedule will be wrong for at least two of the three exams.
The takeaway
Spaced repetition without an exam date is a system optimized for someone who isn't you. The default SRS schedule assumes indefinite retention. Real students writing real exams have a specific day where their recall has to peak, and the schedule needs to know about it.
Set the exam date a few days before your actual test. Build the deck four to eight weeks out. Don't keep moving the date. Run one mock exam in the buffer week. The schedule does the rest. Pair this with the daily 5-minute habit and you've removed almost every reason students fall off a study plan in the final stretch.