QuestionCard

Spaced Repetition Schedule: A Plan You Can Run

By Mark Fulton · 2026-08-17 · 14 min read

Spaced Repetition Schedule: A Plan You Can Run

A spaced repetition schedule is a fixed list of dates on which you re-test a specific batch of cards, with the gaps between those dates getting longer each time. The useful version is built backwards from the date you actually need the material, not forwards from a canned interval list. For a test three weeks out, five sessions per batch at day 0, day 2, day 6, day 12 and day 19 works. For a test five days out, the same five sessions compress into hours and days rather than weeks. You do not need an app to run it, and if you miss a day you shift the remaining sessions rather than starting over.

That is the whole answer. The rest of this is how to build it for your date, what the research actually supports, and the operational details that decide whether the schedule survives contact with a real week.

What does a spaced repetition schedule actually look like?

A schedule has three parts, and most people only ever get given one of them.

The part everyone gets is the interval ladder: 1, 3, 7, 14, or 2, 3, 5, 7, or whatever your source prefers. The two parts nobody gives you are the batch, which is the specific set of cards those dates apply to, and the anchor, which is the date the ladder is measured against.

Without a batch, a ladder is meaningless. You cannot review "biology" on day 3, because by day 3 you have added forty new cards to biology and the deck is no longer the thing you learned. So the unit of scheduling is a batch: a chunk of cards, usually 20 to 40, that you first learn in one sitting. That batch gets its own ladder. The next batch gets its own ladder, starting from its own day 0.

Here is the five-session ladder for a three-week runway, which is the most common real situation: you have a test, quiz, or interview three weeks out, and material to cover.

The three-week schedule, per batch

Session Day Gap since last What you actually do
1 Day 0 first pass Learn the batch. Full pass, grade honestly, two piles only.
2 Day 2 2 days Full pass. Expect it to feel worse than day 0 felt.
3 Day 6 4 days Misses from session 2 first, then a full pass.
4 Day 12 6 days Misses only, unless your first-try accuracy was under 80 percent.
5 Day 19 7 days Full pass, two days before the test. Not the night before.

Two things about that table are deliberate.

The final session is pinned to two days before the test, not the night before. The night before belongs to the batches that are still shaky and to sleep, and a review that happens the evening before mostly measures your short-term memory rather than building anything.

And session 2 lands on day 2, not day 1. Reviewing the next morning feels responsible, but it is the cheapest review on the ladder because almost nothing has faded yet. Which brings us to the part most schedule guides get backwards.

What the research supports, and what it does not

The spacing effect is old and well replicated. Hermann Ebbinghaus ran the first experiments on himself in the 1880s, and it is worth reading what he actually measured, because the popular retelling has drifted a long way from it. In his chapter on retention as a function of time, the numbers are savings: how much of the original learning effort was saved when he relearned the same list later. After 24 hours the saving was about a third of the original work. That is not the same claim as "you forget 70 percent of what you learn in a day", which is the version that gets copied around, and I am not going to repeat a figure whose original does not say it.

The more useful modern finding for scheduling comes from Nicholas Cepeda and colleagues, who taught facts to more than 1,350 people and varied the gap before a review and the delay before the final test, with gaps running up to three and a half months and test delays up to a year. Their result: the best gap gets longer as the test gets further away, and as a proportion of the delay it was around 20 percent for delays of a few weeks, falling to roughly 5 percent when the delay was a year.

That is the rule worth carrying: the right gap scales with your horizon. A test three weeks out wants gaps measured in days. A test a year out wants gaps measured in weeks and months. Copying a "1, 3, 7, 14" ladder onto a professional exam eighteen months away is applying week-scale gaps to a year-scale problem.

Two honest caveats, because this is where most schedule advice quietly overclaims. First, that study tested one gap and one review, so a five-rung ladder is an extrapolation from it, not a finding of it. Second, expanding intervals are not magic. Karpicke and Roediger tested expanding against evenly spaced retrieval and found that expanding practice helped ten minutes later while evenly spaced retrieval was better two days later, concluding that making the first retrieval difficult mattered more than the shape of the ladder after it. And a review of the underlying mechanisms in Nature Reviews Neuroscience is explicit that no single universally optimal interval exists, because different gap lengths engage different processes.

So: get the first real review far enough out that recalling is work, then expand. The exact rungs matter less than people arguing about rungs would like.

How do you set intervals when the exam is in nine days, not nine weeks?

You compress, and you accept that you are buying a smaller thing.

Apply the scaling rule. If the horizon is five days, gaps of a few days are proportionally enormous, so the ladder moves down into hours. Here is the five-day version, same five sessions, same batch:

The five-day schedule, per batch

Session When Gap since last What you actually do
1 Day 0, morning first pass Learn the batch. Keep it small, 20 cards, not 60.
2 Day 0, evening 6 to 8 hours Misses only. Fast pass, five minutes.
3 Day 1 1 day Full pass. This is the session that decides the batch.
4 Day 3 2 days Misses first, then full pass.
5 Day 4, evening 1 day Full pass, misses last so they are the final thing you saw.

Three rules make the compressed version work rather than turn into cramming with extra steps.

Smaller batches. A five-day runway with 20-card batches gives you real coverage. The same runway with 60-card batches gives you one exhausting pass and a lie about having studied.

Front-load the new material. Every batch needs four days of runway to complete its ladder, so you introduce new batches on day 0 and day 1 only. Anything you meet on day 3 gets two sessions, and you should know going in that it is the weakest material you own.

Do not add rungs to compensate. Running the batch eight times in five days does not beat running it five times. The extra passes land while the memory is still fresh, which is the definition of a wasted review.

And the honest part: a five-day schedule buys you the test. A three-week schedule buys you the test and a decent chance the material is still there in a month. If the material matters past the date, the runway is the variable to fix, not the ladder.

What happens when you miss a review day?

You shift. You do not restart, and you do not skip.

This is where most self-built schedules die, because people treat a missed day as a broken streak and abandon the whole plan. The ladder is not a streak. It is a set of gaps, and a gap that ran two days long is still a gap.

The rules I use:

Missed by one or two days: run the session late and keep the remaining dates exactly as planned. A day-6 session run on day 8 is fine. Your day-12 session stays on day 12.

Missed by longer than the gap you were on: drop back one rung, not to the start. If you were due a 6-day session and eleven days have gone by, run the session as if it were the previous rung, misses first and then a full pass, and rebuild the ladder from there.

Missed several sessions across several batches: triage by test date, not by guilt. Run the batch with the worst last first-try accuracy, then the oldest batch, then the rest. Skip whole batches rather than doing half-passes on all of them. A batch reviewed properly and three batches skipped beats four batches skimmed.

Any card you have now missed twice goes into tomorrow's pass regardless of where its batch sits on the ladder. That single rule does most of the work that a per-card algorithm does, and you can run it in your head.

The thing to internalise is that a late review is a slightly harder review, and a slightly harder review is not a problem. The retrieval being effortful is the mechanism, not the failure.

Do you need an algorithm, or is a calendar enough?

For most people, for most of what they are studying, a calendar is enough. Here is where the line actually falls.

A per-card algorithm tracks an interval for every individual card and stretches it after each success, so card 41 might be due in nine days while card 42 is due tomorrow. That is genuinely better bookkeeping. It is also bookkeeping you cannot do by hand, which is the entire reason the software exists.

A calendar schedules batches, not cards. It is coarser: some cards in the batch get reviewed sooner than they strictly need. The cost of that coarseness is a few minutes of redundant review per session. The cost of per-card scheduling is a setup afternoon, a settings screen you will misconfigure, and a review queue that punishes you for a week off.

The crossover point is roughly this. Under a few hundred cards over a semester, batch scheduling on a calendar is the right tool and the redundancy is not worth eliminating. Over a couple of thousand cards across years, professional exams and long language projects, the per-card bookkeeping stops being optional and software wins.

The middle path is what I would suggest for most study projects: schedule batches on a calendar, and let the card player handle the within-session part. Run a batch, grade honestly, then re-drill only the cards you missed until they clear. That is a per-card queue operating inside a batch schedule, and it costs you nothing to set up. The same honest-grading discipline that makes flash cards work at all is what makes the schedule readable, because your first-try accuracy is the only number telling you whether any of this is working.

How many separate batches can you keep in flight?

More than people expect, because the sessions are short and they spread out.

Five sessions per batch times five batches is 25 sessions. Spread across a 21-day runway, that is one session on most days and two or three on the busiest. At eight to ten minutes a batch, the worst day of a five-batch plan costs you under half an hour.

Here is what that looks like laid out, with batches A to E introduced every three days on a three-week runway. Later batches get fewer rungs, because there is not enough runway left to fit five:

Day Batches due Sessions
0 A learn 1
2 A 1
3 B learn 1
5 B 1
6 A, C learn 2
8 C 1
9 B, D learn 2
11 D 1
12 A, C, E learn 3
14 E 1
15 B, D 2
17 E 1
18 C 1
19 A final 1
20 B, D, E final 3

Look at day 12 and day 20. Those are the days that break plans, and they are entirely predictable from the start, which means you can move a batch introduction by a day to flatten them. That is the practical argument for writing the schedule down before you start rather than deciding each morning.

The real limit is not the number of batches, it is the number of new batches. Learning a fresh batch is the expensive session; every other session on the ladder is shorter than the one before it. Two new batches in one day is usually the ceiling, and one is more sustainable. If a subject needs eight batches and you can start one every three days, the runway you need is about four weeks, and finding that out on day 1 is much better than finding it out on day 15.

How do you tell whether the schedule is working?

By first-try accuracy on the same batch, session over session. Nothing else.

Not whether the session felt good, which is actively misleading, because a well-spaced review feels harder than a badly spaced one. Not how many cards you have covered. Not hours logged. First-try accuracy, meaning the percentage you got right the first time you saw each card in that session, before any re-drill.

Read it like this:

Rising across sessions on the same batch. Working. Session 2 at 55 percent, session 3 at 70, session 4 at 85 is exactly the shape you want. Keep going and do not shorten the gaps because it feels easy.

Flat across sessions. This is almost never a schedule problem. Flat accuracy on a batch means the cards are wrong: too big, ambiguous, or testing recognition rather than recall. Fix the cards, and there is a whole method to writing cards that stick rather than cards that merely exist.

High on session 2, lower on session 4. Your gaps are too long for the material. Halve the next gap for that batch only. Dense, interference-heavy material like vocabulary pairs or the multiplication facts that keep colliding genuinely needs shorter gaps than conceptual material does.

Near 100 percent on session 3 and after. The batch is done. Stop reviewing it, or move it to a single confirmation pass before the test. Reviewing material you already know is the most common way study time gets spent for nothing.

Write those percentages down. Four numbers per batch, in a notes app, takes ten seconds a session and turns "am I ready" from a feeling into something you can read off a page.

FAQ

What is the best spaced repetition interval?

There is no single best interval, and the mechanisms review linked above says so directly. The useful rule is proportional: the gap should scale with how far away you need the material. For a horizon of a few weeks, gaps of roughly 2, 4, 6 and 7 days across five sessions work well. For a horizon of a year, gaps of weeks and then months are the right scale. Start with a gap that makes recall genuinely effortful and expand from there.

Can you do spaced repetition without an app?

Yes, and for anything under a few hundred cards a calendar is a perfectly good tool. Schedule batches rather than individual cards, write the five dates for each batch into your calendar the day you learn it, and let re-drilling missed cards inside a session handle the per-card part. The bookkeeping only becomes unmanageable at the thousands-of-cards scale.

How long before an exam should you start spaced repetition?

Work backwards from the material rather than picking a number. Count your batches, allow four days of runway per batch, and allow one new batch every two to three days. Eight batches at one every three days plus a four-day tail is about four weeks. If the runway you have is shorter than the runway you need, cut the number of batches rather than compressing every ladder, because half-learned batches are the first thing to go on test day.

What happens if you skip a review session?

Run it late and keep the rest of the dates as planned. If the delay was longer than the gap you were on, drop back one rung and rebuild from there. Do not restart the ladder, and do not try to make up the missed session by doing two passes back to back, because the second one lands while the material is still fresh and buys almost nothing.

You do not need to plan the whole runway tonight. Pick one batch, run it once on any study deck and note the first-try accuracy on the session report, then put the next four dates in your calendar. The report tells you what you missed, so the next review has already written itself.

Sources

QuestionCard is a free library of flash cards, trivia quizzes, and question card games. No sign-up — open a deck and play.