The Amazon SDE II online assessment has three parts. You get a timed coding challenge, then a system design section, then a work style section. You must finish all three to be considered. Most prep goes into the coding. That makes sense. The system design section is the one candidates misread. It has no diagram and no whiteboard. Amazon’s own prep page says you rate how effective some possible actions are. Each action comes from a made-up work situation with simulated peers and managers. Amazon says the section looks for three things. They are designing for maintainability, weighing tradeoffs and handling ambiguity. So sketching load balancers is the wrong drill. The right drill is judging what a sensible engineer would do next. Below, one Android engineer’s invite shows the whole test. A sample scenario is worked through with a rating for each action and the reason behind it.
Amazon describes the system design section as a rating exercise
Amazon’s SDE II online assessment prep page lays out the format. The page shows no date. Coding is 90 minutes for two questions. It’s the only timed part. System design takes about 15 minutes. Work style takes about 10. Amazon suggests you set aside two hours for the whole thing.
The page’s FAQ explains the system design part. You see a few fictional situations with simulated peers, managers and stakeholders. You won’t draw a diagram or do a whiteboard exercise. You rate the effectiveness of possible actions in each situation. Amazon names three skills to show. They are designing for maintainability, determining tradeoffs and dealing with ambiguity.
Amazon’s numbers don’t fully agree across its own pages. Its SDE II interview prep page lists 20 minutes of systems design scenarios. It also lists an 8-minute work style survey. That page shows no date either. Your invite email is the version to trust.
Prep guides train scaling answers the test doesn’t ask for
DesignGurus’ Amazon assessment guide frames this section as a short design interview. The guide is dated September 14, 2026. It tells SDE II candidates to explain scaling, storage and caching in a few sentences each. DesignGurus sells interview prep courses. The guide links to them. That advice fits a live design interview. The assessment instead asks you to judge a teammate’s proposal under a deadline.
Take an Android engineer with five years on a shopping app. They apply to an SDE II role on a mobile team and get the invite. Their plan is two weeks of graph problems, then some caching notes for system design. The caching notes won’t help much, because nothing in the section asks for a design. It asks which response to a messy situation is sound.
A sample scenario with each action rated
The scenario below is our own invention. It isn’t a real Amazon question. Amazon doesn’t publish its scoring key, so the ratings are our reasoning from its three stated criteria. Your test’s rating scale may also look different.
Your team owns the cart screen in a shopping app. A product manager wants a “save for later” button live by Friday. The request doesn’t say whether saved items should sync across a customer’s devices. A teammate suggests copying the cart code into a new screen to hit the date. How effective is each action?
Here are four possible actions and how the engineer rates them.
Action A is to copy the cart code, ship by Friday and clean it up later. The engineer rates it ineffective. Two copies of the cart logic must now change together. A bug fixed in one stays in the other. It fails the maintainability test. Also, “later” has no owner or date.
Action B is to ask the product manager whether saved items must sync before picking a storage approach. The engineer rates it very effective. Sync decides whether the data lives on the device or on a server. One short question settles the biggest unknown before any code depends on it. This is the ambiguity criterion handled directly.
Action C is to build a full sync service with conflict handling before shipping anything. The engineer rates it ineffective. It solves a problem nobody has confirmed. It also misses Friday. Over-building under ambiguity is a tradeoff failure, just in the other direction from A.
Action D is to ship a local-only version behind a feature flag. The sync question gets written down as an open item. The engineer rates it effective. The date holds and the risk stays small. The model can still grow into sync later. It ranks below B because it still guesses before asking.
Why each rating follows from Amazon’s three criteria
Actions A and C fail in opposite ways. A trades maintainability for speed and hides the cost. C trades the deadline for a feature nobody asked for. Neither one weighs the tradeoff out loud, so both rate low.
Action B ranks first because the unknown is cheap to resolve. One message to the product manager costs minutes. A wrong storage choice costs a rewrite. When a question is cheap and the wrong guess is expensive, asking wins.
Action D is the strongest move when nobody can answer. It keeps the change reversible, because a flag turns it off. It also writes the open question down, so it doesn’t disappear. Android engineers who ship behind flags or Play Console staged rollouts already have this habit. It transfers straight into this section.
The engineer turns this into a rule for every scenario. Rate highest the action that settles the unknown cheaply. Rate next the action that keeps options open. Rate lowest anything that hides a cost or builds ahead of the facts.
The work style section scores how you approach work
The same FAQ says the work style section has two parts. One part is about software engineering. The other is about your work style in general. Both are multiple-choice questions about your approach to work. Amazon measures them against its Leadership Principles.
Read the principles once before the test. Don’t try to game the answers. The DesignGurus guide claims similar statements come back in different words. Amazon’s page doesn’t say that. Either way, answering as you actually work is the simplest way to stay consistent. Telling stories about those principles belongs to the later interview loop. This section is only a survey.
Amazon SDE II online assessment logistics that can cost you six months
Amazon’s prep page sets a clear deadline. You have seven days from the date the link was sent. Need more time? You can ask for the link at a later date in the Amazon.jobs Candidate Portal. Make that request before the seven days run out. If you don’t move forward, Amazon encourages you to reapply in six months.
The coding challenge accepts Kotlin, along with Java, Swift and several other languages. Amazon expects syntactically correct code, not pseudocode. It suggests getting comfortable coding in HackerRank without an IDE. Android engineers who live in Android Studio autocomplete should practice there first. Our post on the algorithm round in Android interviews covers writing Kotlin under time pressure.
One assessment covers multiple SDE II roles, according to the FAQ. If you passed it in an earlier process, a recruiter can say whether the result still applies. Amazon says you’ll hear within two days whether you’re moving forward. The next step is an interview loop of four 55-minute interviews, per the interview prep page. One of them may be run by a bar raiser.
For the Android engineer, the revised plan looks different. Two weeks of prep won’t fit in a seven-day window. So they use the Candidate Portal right away to ask for the link at a later date. They spend most of that time coding in Kotlin inside HackerRank. They spend two evenings rating made-up team scenarios like the one above. They set aside two quiet hours for one sitting, inside the seven-day window. Other companies run their own screening tests. The CodeSignal General Coding Assessment works differently, because one score can follow you to other companies.
1 thought on “Amazon SDE II Online Assessment: Why the System Design Section Has No Diagram”