Study Planning with Spaced Repetition: A Reverse-Engineering Approach

Most people start their SRS journey by making cards. They highlight a textbook, extract 200 flashcards, and hit "study." Three weeks later, they're drowning in reviews and haven't covered half the material.

I've watched this pattern destroy study plans for Step 1, the CFA Level II, and the JLPT N2. The problem isn't discipline—it's that they never did the math.

TL;DR
Reverse-engineer your study plan: start with your exam date, calculate your daily review capacity, subtract a 25% buffer for re-learns, then work backward to determine how many new cards you can actually acquire per day. This article walks through the spreadsheet math and three real case studies.

Why Forward Planning Fails

When you plan forward ("I'll do 50 new cards a day"), you're guessing. You don't know:

• How many reviews those 50 cards will generate in week 4
• Whether your retention rate will hold at 90% or collapse to 75%
• How much time re-learning lapsed cards will consume
• Whether you'll finish the deck before your exam

Forward planning optimizes for starting. Reverse planning optimizes for finishing.

The Reverse-Engineering Framework

Here's the structure I use for every high-stakes exam:

1. Set your exam date and total card count
2. Calculate your daily review capacity (time available ÷ seconds per card)
3. Subtract a re-learn buffer (typically 25% of capacity)
4. Determine your sustainable new-card rate
5. Validate against your acquisition deadline

Let's break down each step.

Step 1: Exam Date and Total Cards

This is straightforward. If you're studying for USMLE Step 1 with the AnKing deck, you're looking at roughly 30,000 cards. For JLPT N3, a solid deck runs 2,500–3,500 cards. CFA Level II candidates often build 4,000–6,000 cards from the curriculum.

Your exam date is fixed. Your card count is negotiable but should be realistic—don't plan to memorize Costanzo Physiology and First Aid and Pathoma if you have 12 weeks.

Step 2: Daily Review Capacity

This is where most plans fall apart. Your review capacity isn't "how many cards I want to do"—it's how many you can actually do given your schedule, energy, and the cognitive load of your material.

I calculate this in seconds, not cards:

Daily review capacity = (available minutes × 60) ÷ seconds per card

For most people:

• Easy recall cards (vocabulary, simple facts): 8–12 seconds
• Medium cards (mechanisms, multi-step concepts): 15–25 seconds
• Hard cards (clinical vignettes, case-based reasoning): 30–50 seconds

A second-year med student grinding Step 1 might have 3 hours daily for Anki. If their average card takes 20 seconds:

(180 minutes × 60) ÷ 20 = 540 cards/day

A working professional studying for CFA with 90 minutes on weekdays:

(90 × 60) ÷ 25 = 216 cards/day

Be honest here. If you're also attending lectures, doing practice questions, or working full-time, your actual available time is less than you think.

Step 3: The Re-Learn Buffer

Here's the thing nobody tells you: your review load isn't just mature cards coming due. It's also:

• Cards you fail and need to re-learn (lapses)
• Cards you mark "hard" that come back sooner
• Cards you suspend and later unsuspend

If you're running at 90% retention (which is aggressive), 10% of your reviews are lapses. Those lapsed cards re-enter the learning queue and generate extra reviews over the next few days.

I budget 25% of my daily capacity for re-learns and variability. So if my raw capacity is 540 cards/day, my working capacity is:

540 × 0.75 = 405 cards/day

This buffer also absorbs life: the day you're sick, the week of final exams, the weekend you travel.

Step 4: Sustainable New-Card Rate

Now we can calculate how many new cards to add daily.

The SM-2 algorithm (used by Anki) and FSRS (used by SmartRecall) both follow a predictable review schedule. A new card you learn today will come back:

• Tomorrow (if you pass)
• In ~3 days
• In ~7 days
• In ~14 days
• In ~30 days

The exact intervals depend on your retention target and the algorithm's parameters, but the shape is consistent: each new card generates roughly 6–8 reviews in its first 30 days.

So if you add 50 new cards today, you're committing to ~350 reviews over the next month from just those 50 cards.

Here's the formula:

Max new cards/day = (working capacity − steady-state reviews) ÷ review multiplier

The "review multiplier" is how many reviews each new card generates before it stabilizes. For most decks, this is 6–8 in the first month.

Let's say you have 405 working capacity and you're already doing 200 steady-state reviews from mature cards:

(405 − 200) ÷ 7 ≈ 29 new cards/day

Step 5: Validate Against Deadline

Finally, check whether your new-card rate gets you through the deck in time.

If you have 5,000 cards and 150 days until your exam:

5,000 ÷ 150 = 33 cards/day needed

If your sustainable rate is only 29/day, you have a problem. Your options:

• Increase study time (raise capacity)
• Reduce deck size (cut low-yield cards)
• Lower retention target (accept more lapses, but this is risky)
• Start earlier (if possible)

This is why reverse planning matters. You find the mismatch before you're 6 weeks in and panicking.

Case Study 1: USMLE Step 1 (AnKing Deck)

Profile: Second-year med student, 24 weeks until exam, using the 30,000-card AnKing deck.

Available time: 3 hours/day on weekdays, 5 hours/day on weekends (average 3.5 hrs/day).

Card speed: 20 seconds (mix of cloze deletions and image occlusions).

Calculation:

• Raw capacity: (210 min × 60) ÷ 20 = 630 cards/day
• Working capacity: 630 × 0.75 = 472 cards/day
• Steady-state reviews by week 12: ~250/day
• Available for new cards: 472 − 250 = 222
• Sustainable new rate: 222 ÷ 7 ≈ 31 new cards/day

Validation:

• Total cards: 30,000
• Days available: 168 (24 weeks)
• Required rate: 30,000 ÷ 168 ≈ 178 cards/day

Problem: The required rate (178) far exceeds the sustainable rate (31). This student needs to either:

• Unsuspend cards earlier in M1 (spreading acquisition over 18 months instead of 6)
• Use a pre-filtered deck (e.g., Zanki Step 1 at ~20,000 cards)
• Increase study time to 5 hours/day

What they did: Started unsuspending cards by organ system during M1 coursework, front-loading 15,000 cards over the first year. Final 6 months focused on 15,000 remaining cards at 89 cards/day—still aggressive but achievable.

They used SmartRecall's workload heatmap to visualize future review load and adjust new-card intake during high-volume weeks.

Case Study 2: CFA Level II

Profile: Working finance professional, 6 months until exam, 5,000 self-made cards from curriculum.

Available time: 90 minutes on weekdays, 4 hours on weekends (average 2 hrs/day).

Card speed: 25 seconds (formula-heavy, requires calculation).

Calculation:

• Raw capacity: (120 min × 60) ÷ 25 = 288 cards/day
• Working capacity: 288 × 0.75 = 216 cards/day
• Steady-state reviews by month 3: ~120/day
• Available for new cards: 216 − 120 = 96
• Sustainable new rate: 96 ÷ 7 ≈ 13 new cards/day

Validation:

• Total cards: 5,000
• Days available: 180
• Required rate: 5,000 ÷ 180 ≈ 28 cards/day

Problem: Required rate (28) exceeds sustainable rate (13) by more than 2×.

What they did:

• Cut deck to 3,500 cards (removed redundant formula cards, focused on high-weight topics)
• Increased weekend study to 5 hours (raised average to 2.3 hrs/day)
• New sustainable rate: ~18 cards/day
• New required rate: 3,500 ÷ 180 ≈ 19 cards/day

Close enough. They finished the deck with 2 weeks to spare and used that buffer for practice exams.

Case Study 3: JLPT N3

Profile: Self-learner, 4 months until exam, 3,000-card deck (vocab + grammar).

Available time: 45 minutes/day (busy work schedule).

Card speed: 10 seconds (mostly single-word vocab).

Calculation:

• Raw capacity: (45 × 60) ÷ 10 = 270 cards/day
• Working capacity: 270 × 0.75 = 202 cards/day
• Steady-state reviews by month 2: ~80/day
• Available for new cards: 202 − 80 = 122
• Sustainable new rate: 122 ÷ 6 ≈ 20 new cards/day (vocab stabilizes faster)

Validation:

• Total cards: 3,000
• Days available: 120
• Required rate: 3,000 ÷ 120 = 25 cards/day

Problem: Required rate (25) slightly exceeds sustainable rate (20).

What they did: Started 3 weeks early (added 21 days), bringing required rate to 3,000 ÷ 141 ≈ 21 cards/day. Finished deck on schedule with 1 week buffer for review-only mode before the exam.

Used SmartRecall's FSRS optimizer to tune retention target to 88% (down from 90%), which reduced review load by ~8% and freed up capacity for the last 500 cards.

Building Your Own Spreadsheet

Here's the template I use. You can replicate this in Google Sheets or Excel:

If the gap is positive, you need to adjust inputs (more time, fewer cards, earlier start). If it's negative, you have buffer—use it.

Common Adjustments

If you're behind schedule: Don't just increase new cards. That makes the problem worse in 2 weeks. Instead:

• Suspend low-yield cards (use tag filters)
• Increase study time by 30 minutes/day
• Lower retention target from 90% to 87% (frees ~10% capacity)

If you're ahead of schedule: Bank that buffer. Don't add more cards. Use the extra capacity for:

• Practice questions (UWorld, Qbank, past exams)
• Weak-topic deep dives
• Rest (seriously—burnout is real)

If your retention is tanking: You're either overloaded or your cards are bad. Check your "again" rate in your SRS stats. If it's above 15%, you need to:

• Reduce new cards by 30–40% for 2 weeks
• Review card quality (are they too ambiguous? too complex?)
• Check for interference (similar cards confusing each other)

SmartRecall's retention analytics dashboard flags these patterns automatically, but you can also track them manually in Anki's stats screen.

The 80/20 of Study Planning

If you take one thing from this article, make it this: your constraint is review capacity, not motivation.

I've seen people burn out not because they lacked discipline, but because they committed to 50 new cards/day when their math supported 25. The algorithm doesn't care about your goals—it will surface every card you're due, and if you can't keep up, your retention collapses.

Reverse planning forces you to confront that reality before you start. It's not sexy. It won't fit in a motivational Instagram post. But it's the difference between finishing your deck with a 90% retention rate and abandoning it at 60% completion because the reviews became unmanageable.

Do the math. Build the spreadsheet. Adjust as you go. Your exam date isn't moving—your plan needs to be the thing that bends.