Amazon Online Assessment (OA): Format and Examples

The Amazon online assessment (OA) is the first technical gate for SDE candidates: a timed, proctored session, usually run on HackerRank, with two coding problems plus non-coding sections built around Amazon's Leadership Principles. The Amazon OA is administered on HackerRank and is timed, proctored, and completed individually, and for technical roles like SDE I, SDE II, and internships the structure typically includes coding challenges (70-90 mins) with one to two algorithmic problems solved in a live coding environment. The catch most people miss: passing every test case does not guarantee you advance. The work simulation and work style survey carry real weight, and weak answers there sink strong coders.

This guide covers the exact 2026 format, the coding patterns that repeat, and where candidates actually lose points. If you want the full pipeline after the OA, we cover it in our Amazon interview breakdown.

What the OA actually contains

Amazon splits the SDE assessment into distinct parts, and the exact mix depends on role, level, and country. Based on 2026 candidate reports and prep guides, here is the typical shape:

Section	Time	Scored?
Coding (2 problems)	70-90 min	Yes, heavily
Work Simulation (day in the life)	~50-60 min	Yes
Work Style Survey	15-20 min	Yes (fit)
Debugging (some batches)	20 min	Yes
Logical / cognitive reasoning (some batches)	~35 min	Varies
Feedback survey	~5 min	No

Depending on the role and location, you'll face a multi-part assessment: a coding section (2 problems, 70-90 minutes), a Work Simulation (~50 minutes), and a Work Style Assessment (15-20 minutes). The levels differ in emphasis. For the current SDE II Online Assessment, Amazon gives 90 minutes for two technical questions, followed by 20 minutes of System Design scenarios and an 8-minute Work Style survey. University and new-grad tracks lean the other way. The current full-time university guide lists coding at about 70 minutes, Workstyles at about 15 minutes, a Work Simulation at about 60 minutes, and a short feedback survey.

One thing to note: the old rigid "OA1 debugging, OA2 coding, OA3 simulation" naming is fading. Older Amazon preparation materials described a separate OA1 debugging stage, but Amazon's current university SDE guidance no longer publishes this 7-question debugging step as a universal standalone assessment, so use the current coding instructions in your invitation instead. Read your invite email carefully. It tells you which sections you actually get.

The coding section: what to expect

Two problems, roughly LeetCode easy-to-medium, occasionally one hard. There were 2 coding questions, one was leetcode easy and the other was leetcode hard. They are auto-graded, and brute force will not cut it. Amazon's recruiters will be looking for optimization and efficiency from your solutions, and not simply brute force. Historically the OA supported a wide language set: C++, C#, C, Java, Python, Swift, Ruby, and JavaScript.

The scoring is based on hidden test cases, which is where most points are won or lost. Hidden test cases specifically target edge cases, empty input, single element, all negatives, large values, and overflow, so explicitly handling them often lifts your score more than micro-optimising the main logic. In our mock interviews we see the same failure repeatedly: a candidate nails the core algorithm, then loses a third of the score to an unhandled empty array or a 32-bit overflow. Handle the boundaries before you optimize the average case.

Coding patterns that repeat

Amazon rotates problems constantly, but the underlying patterns are narrow. From DevInterview's Amazon question bank, ordered by how often candidates report them, these are the ones worth drilling:

Longest Substring Without Repeating Characters (sliding window plus a hash set). The canonical Amazon warm-up. If you can code this cleanly with a moving left pointer and a seen-map, you have the sliding window template for a dozen variants.
Add Two Numbers (linked list traversal with a carry). Tests pointer discipline and a clean while loop that handles unequal lengths and a trailing carry.
Longest Palindromic Substring (expand around center, or DP). Watch the even/odd center cases.
Reverse Integer and String to Integer (atoi). These two are secretly overflow tests. The real problem is not reversing digits, it is returning 0 (or clamping) when the result leaves the signed 32-bit range [-2^31, 2^31 - 1].
Hand of Straights (greedy plus counting). Group cards into consecutive runs of a fixed size using a frequency map or ordered map. A classic "sort, then consume greedily" problem.
Zigzag Conversion (simulation). Pure index bookkeeping; easy to get wrong under time pressure.
Median of Two Sorted Arrays (binary search on partitions). This is the hard-tier outlier. Most SDE I loops will not require the log(m+n) solution, but SDE II and staff candidates should know it cold.

Notice the concentration: sliding window, hash maps, linked lists, greedy, and one binary search problem cover almost everything. That maps directly to the binary search patterns and hashing techniques Amazon leans on. Over-preparing on obscure DP is the most common misallocation we see; under-practicing clean, bug-free implementation of these bread-and-butter patterns is the real gap.

The work simulation and work style survey

This is the part LeetCode grinders ignore, and it is why they fail. Passing all test cases doesn't guarantee you advance; many rejections trace back to the simulation, not the code, so if you're only grinding LeetCode, you're preparing for half the test.

The work simulation drops you into an SDE's inbox. This section is about 50 minutes and simulates a day in an Amazon engineer's inbox: you'll see emails from teammates, messages from managers, bug reports, prioritization requests, and design tradeoff scenarios, formatted as multiple-choice and multiple-select questions with no coding, and every response is scored against Amazon's Leadership Principles.

The work style survey is shorter and personality-based. The Amazon Work Style assessment is a personality test with statements that assess your working preferences and behaviours by having you rate them on a scale of 1-5; it consists of 38 questions and lasts between 15 to 20 minutes. There are no purely right answers, but consistency matters. Answer honestly and consistently, leaning toward ownership and customer focus; inconsistent or extreme responses can hurt a strong coding performance.

Practically: know the 16 Leadership Principles well enough to recognize which one a scenario is testing. When a bug report and a feature request land at once, Customer Obsession and Ownership usually point to fixing the customer-facing bug and not punting it to another team. Bias for Action means deciding with incomplete information rather than stalling.

How it's scored and what happens next

The coding section dominates, but it is a gate, not the whole grade. The coding portion carries the most weight, Amazon's automated evaluation system scores your code solutions, and a low score on the coding section is almost always disqualifying regardless of your behavioral responses. Clear the coding bar and the non-coding sections become the tiebreaker.

The stakes are high because the OA is a silent filter. Pass it, and you move to phone screens and virtual onsite interviews; fail it, and you're out, with no feedback and no second chance for six months. If you advance, the loop that follows typically runs four to five rounds. Candidates who pass proceed to a phone or HireVue video screen, then to the interview loop, typically four to five rounds with Amazon employees, at least one of whom is a designated "Bar Raiser" whose role is to uphold hiring standards across the company.

How to prepare (a realistic plan)

Drill the patterns, not the answers. Don't just memorize the answers; strive to understand the reasoning and patterns behind the solutions, because Amazon changes its coding challenges frequently and you might encounter a question that hasn't been uploaded to LeetCode yet.
Test your edge cases first. Before you submit, feed your function empty input, a single element, and values near integer limits. That is where hidden tests live.
Rehearse the simulation. Read all 16 Leadership Principles until you can name them, then practice tagging each scenario to the LP it targets.
Practice under a clock. Two problems in 70 minutes rewards speed and clean implementation, not clever one-liners. Simulate the constraint. You can run timed company-specific coding practice to build that muscle.
Use a desktop. Amazon's assessment platform is optimised for desktop use, and the Work Sample Simulation in particular presents multi-column layouts with email threads and data tables that are difficult to navigate on a mobile screen.

FAQ

How long is the Amazon OA?

Plan for two to three hours total, though it varies by role. The coding section is 2 problems in 70-90 minutes, plus a Work Simulation (~50 minutes) and a Work Style Assessment (15-20 minutes); for SDE II, the format is 90 minutes of coding plus 20 minutes of Systems Design scenarios and an 8-minute Work Style Survey.

Do I need to pass all test cases to advance?

No, and passing them all still is not a guarantee. Coding uses partial scoring on hidden tests, and Amazon weighs the work simulation and work style sections too. Strong code with a poor simulation can still get you rejected, so prepare both.

Is there still a separate debugging section?

Sometimes. Some SDE online assessments include a debugging section with 7 to 10 code snippets, each containing one or more bugs you need to identify and fix, with 20 to 30 minutes for the section. It is no longer a universal standalone stage, so confirm from your invitation email.

What language should I use?

Whichever you are fastest and most bug-free in. The OA historically supports C, C++, C#, Java, Python, Swift, Ruby, and JavaScript. Python and Java are the most common choices for readability and speed.

What happens if I fail?

You are typically out for that role with no feedback, and Amazon usually enforces a cooldown of around six months before you can reapply. Treat the OA as a real interview, not a practice run.