π Part 6: Writing Cards That Actually Stick
Anki is only as good as the cards you feed it. This is the part that separates people who quietly quit after two weeks from people who remember what they study for years. The good news: card-writing is a learnable craft, and a handful of principles do most of the work.
π What You'll Learn
By the end of this part, you'll understand:
- Why card quality β not app settings or willpower β is the single biggest lever on your Anki success
- The Minimum Information Principle and how to write atomic, one-fact-per-card questions
- How to make questions unambiguous, and why vague fronts and long lists quietly sabotage you
- Why you should understand material before you card it, and how to personalize cards so they stick
- How to spot and rewrite a bad card β and why a "leech" is usually a card problem, not a brain problem
In This Part
Why Card Quality Dominates
Here is the uncomfortable truth that no amount of tweaking intervals will fix: a bad card wastes every single future review. Anki will faithfully show you a confusing, overloaded, or ambiguous card hundreds of times over the coming years β and each of those repetitions costs you real time and real frustration. A good card, by contrast, almost reviews itself. You glance at the front, the answer springs to mind, you press a button, and you move on in three seconds.
When people say "Anki didn't work for me," the cause is almost never the algorithm. It's the cards. Poorly written cards are the number-one source of Anki frustration and the number-one cause of review overload β a mountain of failed cards that reappear again and again, souring the whole habit. Fix the cards, and the mountain melts.
π§ Mindset
If your first cards are clumsy, that's completely normal β everyone's are. Card-writing is a skill, and like any skill it improves with reps. You don't need to get it perfect the first time; you need to notice when a card fights back and be willing to rewrite it. This part is a reference you can return to any time you sit down to make cards.
π‘ The leverage point: A one-minute investment in writing a card well saves you hundreds of seconds of struggle across its lifetime β and, more importantly, it's the difference between actually remembering the material and just shuffling flashcards around.
The Core Principles
Most great cards come from a small set of principles, drawn largely from Piotr WoΕΊniak's classic essay "20 Rules of Formulating Knowledge" (the intellectual foundation of SuperMemo and, by extension, Anki). Learn these, and the rest is practice.
1. Understand before you memorize
This comes first for a reason. Anki maintains memories; it cannot manufacture understanding. If you don't understand something, memorizing the words will leave you with a card you can parrot but not use β and it'll be miserable to review because nothing about it feels meaningful. Learn the concept first from a book, video, or teacher. Then make cards to keep it fresh.
β οΈ Watch Out
Don't turn Anki into a to-do list of things you haven't learned yet. A card you don't understand is a leech in waiting. Close the gap in understanding first β the card will practically write itself afterward.
2. The Minimum Information Principle
π Definition
Minimum Information Principle: Formulate the smallest, simplest questions you possibly can. Simple items are easier to remember, faster to review, and β crucially β schedule independently, so you can keep the parts you know and drill only the parts you don't.
A big, complex card behaves like a chain: get any link wrong and the whole card counts as failed, dragging even the parts you knew perfectly back to the start. Break the material into many small cards and each fact lives its own life on its own schedule.
3. Make cards atomic β one fact per card
An atomic card tests exactly one thing. "What year did the Berlin Wall fall?" is atomic. "Describe the causes, course, and consequences of the fall of the Berlin Wall" is an essay prompt wearing a flashcard costume. Whenever a card asks for two things joined by "and," that's usually two cards.
4. Make questions unambiguous
A good card has one clear, predictable answer. If the front is vague β "Tell me about the heart" β the answer changes every time you see it, so there's nothing stable to grade yourself against. Was it the four chambers? The blood flow? The rate? An unanswerable question is worse than no card at all, because it trains hesitation.
β οΈ Important: Before you save any card, ask: "If I showed this front to myself in six months, would I know exactly what answer is expected?" If the honest answer is "it depends," rewrite it until it doesn't.
5. Avoid enumerations and long lists
Recalling a long, ordered list is one of the most brittle things you can ask of memory β miss one item and the card fails, and the sheer size makes it fail often. Instead:
- Cloze each item β turn one sentence listing the items into several cards, each hiding one item.
- Chain it β "What comes after Y?" cards walk you through a sequence one step at a time.
- Use overlapping cloze β reveal most of the list and hide one piece, so context supports recall.
Big lists fail because there's no natural cue linking one item to the next, and because the all-or-nothing grading punishes you for 90% recall. Shrink the ask, and success rates climb.
6. Use context and cues wisely β but avoid pattern matching
A little context makes a card meaningful; too much (or the wrong kind) lets you cheat. Pattern matching (a form of interference) happens when the card is answerable from a superficial cue β an unusually long blank, a familiar phrase, the one card in the deck about that topic β rather than from real recall. You "know" the answer only because you recognize the card's shape. The fix is to keep fronts natural and consistent, and to phrase questions so the answer comes from understanding, not from a tell.
β οΈ Watch Out
The opposite failure is just as real: a card so stripped of context that it's meaningless in isolation. "What is the value?" β value of what? Aim for the middle: enough context to make the question unambiguous, not so much that a cue gives the game away.
7. Personalize and connect
Cards built from your examples, your images, your mnemonics, and vivid or even absurd associations stick far better than dry, generic phrasing. This isn't a soft suggestion β it leans on the picture superiority effect (we remember images far better than words) and on the simple fact that things connected to what you already know have more hooks to hang on.
β Tip
When a fact won't stick, invent a ridiculous image for it and add it (or describe it) on the card. The weirder and more personal, the better it sticks. Absurdity is a feature here, not a bug.
8. Use media when it carries meaning
Images and audio are powerful when they are the content β a pronunciation clip for a language, a diagram you must read, a face you must name, a chemical structure. Add them freely in those cases. But keep media relevant, not decorative: a stock photo that just fills space adds load without adding memory. (The mechanics of attaching media were covered in Parts 4 and 5 β here the point is knowing when and why.)
9. Decide: two-way or one-way?
For vocabulary especially, ask whether you need recall in both directions. Do you need to produce the word when you see the meaning (recall), recognize the meaning when you see the word (recognition), or both? A reversed card gives you both directions but doubles your reviews β spend that budget only where you truly need production, not everywhere by reflex.
Good vs. Bad Cards in Practice
Principles land better with examples. Here are the most common failure patterns, each paired with a rewrite.
Example A: The overloaded enumeration
| β Bad card | β Rewritten |
|---|---|
| Front: List all the functions of the liver. Back: Detoxification, protein synthesis, bile production, glycogen storage, hormone regulation, clotting-factor production. |
Six atomic cards (or six clozes), e.g.: β’ "Which organ stores glucose as glycogen?" β Liver β’ "The liver produces {{c1::bile}} to help digest fats." β’ "Clotting factors are synthesized by the {{c1::liver}}." |
The bad card fails whenever you forget any one of six items β and grading "did I get it right?" becomes a judgment call. The rewrite lets Anki drill the two functions you keep forgetting while leaving the four you know alone.
Example B: The vague front
| β Bad card | β Rewritten |
|---|---|
| Front: Photosynthesis? Back: The process by which plants convert light energy into chemical energy, using COβ and water to produce glucose and oxygen. |
Front: What two raw inputs do plants use in photosynthesis? Back: Carbon dioxide and water. Front: What does photosynthesis produce (besides oxygen)? Back: Glucose. |
"Photosynthesis?" could want a one-word definition, the full equation, the location, or the purpose β the answer is unpredictable, so self-grading is meaningless. Each rewrite has exactly one right answer.
Example C: The two-in-one
| β Bad card | β Rewritten |
|---|---|
| Front: What is the capital of Australia and what is its population? Back: Canberra; about 460,000. |
Front: What is the capital of Australia? Back: Canberra. Front: Roughly what is Canberra's population? Back: About 460,000. |
Any time you see "and" joining two facts, suspect two cards. You might know the capital cold but blank on the population β the joined card punishes you for the half you knew.
π A note on downloaded decks
It's tempting to download a giant shared deck and expect it to stick. Be careful: cards you write (or heavily edit) are understood and memorable because you built the connections yourself. Someone else's phrasing often means nothing to you, so it reviews badly. That said, premade decks can be perfectly fine for standardized content β a well-made deck for a language's most common words, or for a certification exam's fixed facts, saves real time. The nuance (when to download vs. build) gets its own treatment in Part 10; for now, if you use a shared deck, plan to edit the cards that don't click.
Editing is normal β and leeches are a signal
Pressing a card wrong three times in a row is not a verdict on your intelligence. It's feedback about the card. Anki even formalizes this: a card you fail repeatedly becomes a leech, which Anki flags (and by default suspends) so it stops wasting your time.
β Tip
When you meet a leech, don't grind it harder β rewrite it. Ask: Is it overloaded? Ambiguous? A list? Do I actually understand it? Nine times out of ten a leech is a bad card, not a bad brain. Fixing the card is the fastest way to finally learn the fact.
π The card-writing principles, distilled (reference)
Keep this list handy whenever you make cards β it's a reference, not a to-do:
- Understand first β never card what you don't yet get.
- Minimum information β smallest, simplest question possible.
- One fact per card β if there's an "and," suspect two cards.
- One clear answer β the front must be unambiguous.
- Avoid long lists β cloze each item or chain them instead.
- Cue, don't pattern-match β enough context to be clear, not enough to cheat.
- Personalize β your examples, images, and vivid associations.
- Media that carries meaning β relevant, never decorative.
- Choose your direction β reverse only when you need both.
- Edit freely β a leech is a card to rewrite, not a brain to blame.
π οΈ How To: Rewrite a Bad Card into Good Cards
What you'll do: Take one overloaded card and turn it into a small set of atomic cards (or clozes) that each have exactly one clear answer.
We'll use this real offender:
Front: What are the three branches of the U.S. government, and what does each do?
Back: Legislative (makes laws), Executive (enforces laws), Judicial (interprets laws).
Step by step
- Identify each distinct fact. There are (at least) three here: legislative β makes laws, executive β enforces laws, judicial β interprets laws. Notice the "and" and the enumeration β two red flags at once.
- Split into atomic cards, or cloze each item. Either write three Q&A cards ("Which branch interprets laws?" β Judicial) or write one sentence and cloze the pieces: "The {{c1::legislative}} branch makes laws, the {{c2::executive}} branch enforces them, and the {{c3::judicial}} branch interprets them." Cloze deletion (Part 5) shines here β three cards from one sentence, each hiding one branch.
- Add a cue or image. Attach a small hook to anything that won't stick: picture a judge's gavel for "judicial β interprets," or a factory for "executive β enforces." Your own mental image beats a textbook phrase.
- Test each front for exactly one clear answer. Read each card's front cold and check that only one answer fits. "Which branch interprets laws?" has one answer. If any front still feels like it could go several ways, tighten the wording before you save.
π‘ Tip
Do the rewrite inside the review session where the leech bit you. When a card frustrates you, open it (press E to edit), fix it on the spot, and reset it if needed. Fixing cards in the moment β while you remember exactly why it was confusing β is far more effective than promising to "clean up my deck later."
Best Practices
β Do's
- Understand the material first β Anki keeps memories alive, it can't build them from scratch.
- Write the smallest question that captures the fact β small cards are fast, robust, and independently scheduled.
- Give every front one predictable answer β if you can't grade it consistently, it's not done.
- Personalize with your own images and associations β memory rewards vividness and connection.
- Rewrite cards that keep failing β treat a leech as a bug report, and fix the card.
β Don'ts
- Don't cram multiple facts onto one card β you'll fail the whole card over the one piece you forgot.
- Don't memorize long ordered lists as a single card β they're brittle; cloze or chain them instead.
- Don't write vague fronts like "Tell me about X" β an unpredictable answer can't be graded.
- Don't rely on a cue that lets you pattern-match β recognizing a card isn't the same as recalling the fact.
- Don't download a huge deck and assume it'll stick β plan to edit the cards that don't click for you.
π‘ Pro Tips
- Keep a "20 Rules" bookmark open while you make your first hundred cards; the principles become automatic fast.
- If a card takes more than ~10 seconds to answer because of how it's written (not because it's genuinely hard), it's probably too big β split it.
π Learning Journal
Keep a journal as you work through this guide β digital or paper. After each part, jot down:
- Key ideas you learned
- Things that clicked for you
- Questions or confusion points to revisit
- Ideas you want to try
- Your progress and how you feel about it
βοΈ This part's prompt: Find one card (or one fact) you keep failing. Write it out as it stands now, then diagnose it against the principles: is it overloaded, ambiguous, a list, or something you don't fully understand? Rewrite it smaller and clearer β and note what specifically made the original hard.
π Part Summary
π Key Takeaways
- Card quality is the single biggest determinant of Anki success β a bad card wastes every future review; a good one nearly reviews itself.
- The Minimum Information Principle and atomic cards (one clear fact per card) do most of the heavy lifting.
- Understand before you memorize, keep fronts unambiguous, avoid brittle long lists, and personalize with your own hooks and images.
- A repeatedly failed card (a leech) is almost always a card problem β rewrite it rather than grinding it.
π What You Can Now Do
You can now look at any flashcard β yours or someone else's β and tell whether it's built to stick. You can take an overloaded, vague, or list-heavy card and rewrite it into a set of small, clear, atomic cards that Anki can drill efficiently. That's genuinely the highest-leverage skill in the whole guide: master it and everything downstream (scheduling, review load, long-term retention) gets easier.
β Common Questions
Isn't it faster to make one big card than five small ones?
Faster to create, slower to live with. One big card fails often, drags known material back to zero, and clogs your reviews for years. Five small cards take a few extra minutes up front and then quietly work. You're optimizing for the hundreds of reviews ahead, not the one minute of typing.
Some facts really are a list β how do I handle those without breaking the rule?
Cloze each item so you get one card per list member, or chain them ("what comes after Y?"). You keep the list intact conceptually while asking your memory only one small thing at a time. Overlapping cloze β showing most of the list and hiding one piece β is a great middle ground when order and context matter.
Should I reverse every vocabulary card so I learn both directions?
Only where you actually need production. Recognition (see the word, know the meaning) and recall (see the meaning, produce the word) are different skills, and reversing doubles your reviews. Reverse the words you need to actively use; leave one-way the ones you only need to understand.
How do I know if a card is "bad" versus just hard?
Hard cards fail because the material is genuinely difficult and improve with reps. Bad cards fail because of how they're written β they're overloaded, ambiguous, or a list β and no amount of repetition fixes that. If a card keeps failing and rewriting it smaller/clearer suddenly makes it easy, it was a bad card.
π Up Next
In the next part, The Daily Review β Answer Buttons & Scheduling, we'll shift from making cards to reviewing them: what Again / Hard / Good / Easy really mean, how honest grading feeds the algorithm, and how to keep your daily load sustainable.
π Additional Resources
- "20 Rules of Formulating Knowledge" by Piotr WoΕΊniak β the classic essay behind everything in this part; read it at least once.
- Anki Manual β Getting Started β official documentation on cards and notes.
- Part 5: Cloze Deletion & Other Card Types β the mechanics that make atomic cards easy to build.
π Keep Going
Nobody writes perfect cards from day one β you get better by noticing which cards fight back and fixing them. Every card you rewrite makes the next one easier to write well. Come back to this page any time you sit down to make cards; the principles are always here waiting.