AWS certification study has a specific failure mode, and it is not laziness.
It is that the material is a large set of services with overlapping names, similar purposes, and quirks that only show up at the edges. The exam does not ask "what is S3." It describes a situation and asks which service fits, which means the thing being tested is a decision, not a definition.
Most AWS decks are built to test definitions. That is why people who can recite every service still miss the questions.
Start from the exam guide, not from a service list
Every AWS certification publishes an exam guide with the domains covered and how they are weighted. Get the current one from the certification page before you build anything. The domains and weightings are updated over time, so a study plan written against an old guide studies the wrong proportions.
Use the guide to decide how much deck each domain deserves. A domain with a heavier weighting should get proportionally more cards than a light one. This is the single most common sizing mistake: people study what is interesting rather than what is weighted, and then wonder why practice scores do not move.
Write the domain names down as your top-level organization. Every card belongs to exactly one domain. If a card seems to belong to two, it is usually two cards.
One card, one decision
The rule that fixes most bad AWS cards: a card should ask for one thing you can say in a breath.
Compare these.
- Bad: "Explain the difference between SQS, SNS, and EventBridge."
- Better, split into three decisions: "You need one producer and many independent consumers, each getting every message. Which service?" / "You need to buffer work so a slow consumer does not lose tasks. Which service?" / "You need to route events by content rules to different targets. Which service?"
The second version is longer to write and far better to review, because you can fail it for one specific reason. The first version fails for reasons that have nothing to do with what you were testing — you might know all three services and still produce a rambling answer.
The shape to aim for is a situation → service, not a service → description. The exam gives situations. Train on situations.
Card types worth having
Four formats cover most of what an AWS exam tests.
Situation to service. "A workload needs a relational database with automatic multi-region replication. Which service?" This is the closest thing to exam format and should be the largest share of your deck.
Service to distinguishing feature. "What does X do that a plain load balancer does not?" Tests the edge that separates similar options.
Constraint checks. "What is the maximum size of an object in X?" "Is X regional or global?" These are the pure recall cards, and they are worth having precisely because they are not interesting.
Anti-cards. "A team uses X for Y and it goes wrong because of Z." The negative case is where most exam distractors live. If you only learn what a service is for, you will pick it for jobs it cannot do.
Resist the urge to write a card for every service you have heard of. A deck with 300 situation cards beats a deck with 2,000 definition cards, because you will actually finish the first one.
The wrong-answer loop
This is the highest-value part of the whole plan, and it is the part people skip.
Every practice question you miss contains a specific gap. Not "I should study compute more" — something narrower, like "I did not know that X cannot span regions."
The loop:
- Miss a practice question.
- Write down why you missed it, in one sentence, while you still remember.
- If the cause is a fact or a distinction, make one small card for it, immediately.
- If the cause is that you did not know the service existed at all, make a situation card for it, not a definition card.
- Group these cards so you can review them as a set.
Distractors are the most valuable material in the whole study process, because they are the exam telling you exactly where the edge of your knowledge is. A practice exam you scored 70% on is worth more than one you scored 90% on, provided you process the misses.
One warning: do not make a card for the exact question you missed. It will be a card about a question, not about a concept, and it will not generalize. Make a card about the underlying distinction.
The services you keep confusing
Every certification has a set of services that blur together. Rather than fighting this with more cards, build one comparison card per pair and make the distinguishing axis explicit.
Ask, for any pair you confuse:
- What is the trigger? Push-based versus pull-based is a common axis.
- What is the scope? Regional, global, single-account, cross-account.
- What does it guarantee? Ordering, at-least-once, exactly-once, durability.
- Who consumes it? Humans, applications, other services.
Most AWS confusion reduces to one of those four axes. A card that names the axis is far more useful than a card that lists features.
Where to put your practice exams
Cards build the raw material. Practice exams test whether the raw material is organized well enough to make decisions.
A workable split near the end:
- Cards daily, but with new cards tapering off. The last weeks should be mostly review.
- Practice sets regularly, with a full timed run at the end to check pacing.
- Immediately after each practice run, do the wrong-answer loop before you do anything else. The value of the miss decays fast.
If your practice scores plateau while your cards all pass, the issue is usually application rather than recall. More definition cards will not move it. Situation cards, or more practice questions, will.
Building the deck without losing a weekend
If your source material is already written down — a course's slides, a set of notes, a study guide as a PDF or a Word document — the mechanical part of writing the cards is a reasonable thing to hand off.
WakeCards takes a document and produces a first pass of question-and-answer cards in the same language as the source. For AWS that means you get a draft deck to edit rather than a blank page, and editing is much faster than writing.
Two honest cautions. AWS changes service capabilities regularly, so any generated card about limits or features should be checked against current documentation before you memorize it. And generated cards tend toward descriptions, which is the format that tests worst on this exam — so your job is to convert them into situation cards. That conversion is the studying.
A plan you can start this week
- Download the current exam guide for your certification. Note the domains and weightings.
- Pick the heaviest domain. Build situation cards for it only.
- Register for one practice question set, and take it before you feel ready.
- Run the wrong-answer loop on every miss.
- Next week, add the second domain and keep reviewing the first.
The plan is small on purpose. The thing that fails people on AWS is not the size of the deck. It is building a deck about services instead of a deck about decisions.
Frequently asked questions
How many cards do I need for an AWS certification?
Far fewer than most people build. A focused deck of a few hundred situation cards that you actually review will teach you more than a few thousand definition cards you never finish. Size the deck to what you will review daily.
Should cards test definitions or scenarios?
Scenarios. The exam gives you a situation and asks which service fits, so situation cards match the test format. Keep a small set of definition and limit cards for the facts you must know cold, but do not let them dominate.
Are pre-made AWS decks worth using?
They can save time, but AWS changes often, and pre-made decks age badly on limits, features, and service names. If you use one, verify anything about capabilities or quotas against current documentation.
How do I handle services with similar names?
Build one comparison card per confusing pair and name the axis that separates them — trigger, scope, guarantee, or consumer. Naming the axis generalizes to questions you have not seen.
Is it worth making cards from practice exam questions I missed?
It is arguably the best source of cards you will find. A miss identifies a precise gap. Convert it into a card about the underlying distinction, not a card about the question itself.