Skip to content

Amazon online assessment

Amazon online assessment (OA) is the first hard filter for most fresher, campus, and off-campus SDE-style drives. This page covers the Amazon OA format, what happens after you clear it, sample Amazon OA questions with approaches, an 8-week prep plan, and mistakes that sink otherwise-strong candidates.

Patterns below are candidate-reported for 2025-2026. Platforms and section names can change by drive. Always confirm the invite mailer for your window.

Stage What candidates report What you must show
1. Online assessment Often 70-120 min: 2 coding problems + work simulation / debugging + LP-style behavioral MCQs Correctness + reasonable complexity under time
2. Technical interviews Multiple DSA rounds; every round mixes coding with Leadership Principles Explain approach, edge cases, complexity
3. Later rounds Bar Raiser is mandatory on many loops; system design depth scales with level Depth beyond the OA pattern
4. HR / offer Role, location, CTC discussion Consistency with resume
Component Candidate-reported details
Platform HackerRank or Amazon OA platform
Duration About 70-120 minutes depending on invite
Question mix 2 coding (DSA) + work simulation / scenario tasks + behavioral MCQs
Languages Java, C++, Python, C# (pick one you can debug fast)
Scoring Automated tests on coding; simulation and MCQs scored separately; partial credit varies
Advance rate Roughly 15-25% of applicants clear OA in many campus-style reports
  • Coding alone is not enough: work simulation and Leadership Principles signals appear early
  • Problems often need clean edge-case handling more than exotic algorithms
  • Amazon-tagged LeetCode patterns (graphs, sliding window, heaps) show up frequently in candidate reports
  • Time pressure is real: leaving one coding problem unsubmitted usually ends the OA
  • Arrays, strings, hashing, sliding window
  • Graphs and trees (BFS/DFS, topological ideas)
  • Heaps / priority queues, intervals
  • Dynamic programming and greedy when constraints allow
  • Occasional bit manipulation or design-lite coding
  • Work-simulation style judgment (emails, debugging a small system, prioritizing tasks)
  • Behavioral MCQs mapped to Leadership Principles
  • Debugging snippets where the bug is logical, not just syntax
  • Full system design interviews (those come later)
  • Long aptitude batteries like service-company NQT papers
  • HR salary negotiation (offer stage)

Question 1: Maximum profit with one transaction

Section titled “Question 1: Maximum profit with one transaction”
Q: Prices of a product over n days are given. Buy once and sell once later. Maximum profit?

Track the minimum price so far and the best price - minSoFar. O(n) time, O(1) space. Edge cases: empty array, strictly decreasing prices (profit 0).

This is the classic Best Time to Buy and Sell Stock pattern frequently reported in Amazon-style screens.

Q: Given a stream of product IDs, return the k most frequent IDs.

Count with a hash map, then a heap of size k (or bucket sort by frequency). Average O(n log k) with a heap.

Watch ties: clarify whether any valid top-k is accepted.

Question 3: Number of islands / warehouse zones

Section titled “Question 3: Number of islands / warehouse zones”
Q: Grid of 1/0 cells. Count connected components of 1s (4-directional).

DFS/BFS flood fill. Mark visited in-place if allowed. O(rows * cols).

Amazon reports often include grid/graph connectivity variants.

Question 4: Merge overlapping delivery windows

Section titled “Question 4: Merge overlapping delivery windows”
Q: Merge overlapping intervals [start, end].

Sort by start. If the next interval starts at or before the current end, extend the end; else push current and move on.

O(n log n) from sorting.

Q: A hit-counter should return hits in the last 300 seconds but returns stale counts.

Typical bug: forgetting to evict timestamps outside the window before counting. Fix with a queue of timestamps and pop while the front is older than the window.

Question 6: Leadership Principles judgment (MCQ style)

Section titled “Question 6: Leadership Principles judgment (MCQ style)”
Q: A teammate wants to ship a feature with known data-quality risk to hit a date. What do you do?

Amazon-style answers favor Customer Obsession and Dive Deep: surface the risk, propose a measured mitigation, and avoid silent ship. Practice STAR stories for Bias for Action vs Insistence on Highest Standards trade-offs.

How to Prepare for Amazon OA (8-Week Plan)

Section titled “How to Prepare for Amazon OA (8-Week Plan)”
Weeks Focus Exit criteria
1-2 Arrays, hashing, two pointers, sliding window 40+ easy/medium solved cleanly
3-4 Trees, graphs (BFS/DFS), heaps Can code BFS/DFS without notes
5-6 DP + greedy patterns (plus Amazon LP story outlines in parallel) 25+ medium DP/greedy with recursion to bottom-up
7 Timed full mocks matching About 70-120 minutes depending on invite Finish both coding problems with solid tests with passing tests
8 Weak-topic repair + complexity review Re-solve every miss within 48 hours
  1. One timed medium problem (45-60 min hard cap)
  2. Write complexity and 3 edge cases before coding
  3. If stuck past 20 minutes, peek pattern name only, then re-solve tomorrow cold
  4. Log misses by topic on a simple sheet
  1. Spending 50 minutes on problem 1 and submitting nothing for problem 2
  2. Ignoring constraints and shipping O(n^2) when n is clearly large
  3. Skipping work simulation / MCQs because “coding matters more” (it all counts)
  4. No Leadership Principles stories ready even though OA/interviews expect them
  5. Copy-paste templates you cannot explain in the next interview
  1. Stable internet, charged laptop, quiet room
  2. Confirm language template compiles on the platform before starting hard problems
  3. Skim all problems for 3-4 minutes, order by familiarity
  4. Submit a correct brute force early if constraints allow, then optimize
  5. Leave 8-10 minutes for edge-case tests you invent yourself

Google · Microsoft · Meta · Apple · Netflix · Flipkart