Updated September 2026 · 10 min read · US multinational technology corporation
The first is that Microsoft assesses candidates against "Leadership Principles." It does not. Leadership Principles are Amazon's, and the confusion is so widespread that entire prep guides are built on it. Microsoft publishes core values and a growth mindset framing instead: "We look for respect, integrity, accountability, and growth mindset."
The second is that "Model, Coach, Care" is Microsoft's interview rubric. It is not. It is a framework for how Microsoft's existing managers are expected to lead their teams, rolled out around 2019. No source shows candidates being scored against it.
The third thing — the AA round — is real, but the story behind it is more interesting than the version prep sites sell, and Microsoft's own engineering blog quietly contradicts most of what is written about it. That is where this guide starts.
Microsoft publishes one generic funnel for every role: Explore open jobs → Submit application → Application review → Interview → Offer → Pre-onboard → Hired.
It then declines, explicitly, to standardise beyond that. Its own wording: "hiring teams may approach interviews differently. In some cases, after reviewing resumes, the team might schedule a brief screening conversation." On the loop itself: "Most interviews include 2-4 conversations with potential teammates and cross-functional colleagues, each lasting up to an hour."
For university and new-grad hiring, Microsoft names two rounds and describes them in a way that matters for how you prepare.
At the Initial Interview, you are "assessed for a functional role at Microsoft, like mechanical engineer or software engineer" — the function, not the team. At the Final Interview, you are "being assessed for a specific role within an organization," with team matching drawn from interview feedback plus your stated preferences. All virtual. Microsoft says the decision window can run up to 90 days.
That two-stage structure explains something candidates often find disorienting: your first Microsoft interview may not be for any particular job. You are being sorted into a discipline first, and into a team second.
For experienced hires, Microsoft publishes no separate process. Recruiter screen, then a loop of 2 to 4 conversations, then an offer decision.
One notable absence: the online assessment is not documented anywhere on Microsoft's careers pages. It appears only in candidate reports and prep material. Microsoft's own statement is only that candidates "may be asked to write code, share a creative portfolio or project, or provide examples of your work."
Microsoft names four technical evaluation areas, and unlike most employers it says what each one means.
Problem-solving. "Break down information and navigate through problems," and "ask clarifying questions for any ambiguities." That second half is graded behaviour, not politeness — a candidate who starts coding before establishing constraints is being marked down while they type.
Design. Working with an existing platform and understanding "what it takes to build a new system."
Testing. "You test your own and other team members' code," including identifying "corner cases."
Technical excellence. Algorithms, data structures, system design, AI/ML, data science and networking, weighted by role.
Microsoft also scores code quality directly: "demonstrate best practices and ensure your code is clean, concise, and bug free."
For students, Microsoft frames the purpose more broadly — to "discover your base knowledge in the area of focus, evaluate critical thinking skills, assess how you solve problems and your ability to learn," plus identifying "your strengths to pair you with the right team."
Online assessment (university and new-grad software roles). Microsoft names no vendor and does not acknowledge the stage publicly. The best-supported attribution is Codility, from first-hand candidate reports — a Blind post in August 2024 described a new-grad software engineering assessment as "2 questions, 110 min Codility test," and Glassdoor carries an interview entry titled "1 Question Codility Assessment online."
Prep sites variously name Codility, HackerRank and CodeSignal for the same assessment. Only Codility has first-hand candidate support, and no Microsoft source confirms any of them. Treat the vendor as unconfirmed and possibly variable by region and team.
In-loop coding. Microsoft confirms this happens in a "third-party coding tool where you can run and compile your code" — real, runnable code, not whiteboard pseudocode. Its instructions are unusually direct: "you'll be asked to code only in the language you're strong in," and "don't use pseudocode." Interviews run 45 minutes, and Microsoft states its process is "currently entirely virtual."
The loop. 2 to 4 conversations mixing behavioural and technical content.
The AA round. The final conversation, typically with a senior interviewer.
Here is where Microsoft's own record contradicts the folklore. Raymond Chen, writing on Microsoft's official developer blog in October 2023, explains the origin: "The final interviewer was called the 'as appropriate' interviewer because the final interview took place only if they felt it was appropriate." The AA was a senior gatekeeper who read the earlier feedback and decided whether the final round was worth holding at all. Chen then records the change: "the 'As appropriate' stopped being optional. Everybody got an 'As appropriate' interview," and he characterises the term as dated internal jargon that no longer reflects current procedure.
So: the round is real, it is usually last, it is usually behavioural and fit-focused, and the name is a fossil. Claims that it can single-handedly override an otherwise strong loop are not documented by Microsoft anywhere.
Non-engineering roles. Microsoft says candidates "may be asked to… share a creative portfolio or project, or provide examples of your work in other ways." There is no current official description of the sales, marketing, finance or program-management loop. The only Microsoft-domain account is an archived employee blog post from 2008 — historically interesting, 18 years old, and not a basis for preparation.
Software engineering. The only track with confirmed live coding. University candidates report an online assessment; experienced-hire reports rarely mention one. The loop is coding-heavy, in a runnable environment, in your strongest language.
Program management. No coding. Portfolio, project examples, and product-sense discussion.
Sales, marketing and business roles. No current official description. Scenario-and-reasoning formats are the historical pattern, where the assessed quality was the defensibility of your logic rather than a correct answer.
Graduate versus experienced. Graduates get the explicit two-round, function-then-team structure with decision windows up to 90 days. Experienced hires go straight into a team-specific loop, and the sorting question — which team — is usually already answered.
Microsoft publishes no sample assessment questions. What follows are original practice items written to match the evaluation areas Microsoft names. They are not Microsoft questions and are not drawn from any real assessment.
Coding practice, in the style of a timed online assessment.
You are given a list of integers representing daily readings from a sensor. A reading is "anomalous" if it differs from the median of the 5 readings centred on it by more than a given threshold. Return the indices of all anomalous readings. Readings near the start and end of the list, where a full window of 5 is unavailable, should be ignored.
Before writing code: state your approach, name the time complexity you expect, and identify 3 edge cases. Microsoft grades clarifying questions, so in a live setting the correct first move is to ask what should happen with an even-length window, with duplicate values, and with an empty input.
Testing practice — an area Microsoft names explicitly and most candidates neglect.
You have written a function that merges two sorted lists into one sorted list.
Name 6 test cases you would write, and say what each one is protecting against.
Strong answers cover both lists empty, one empty, equal lengths, wildly unequal lengths, duplicate values across both lists, and values at the type's boundaries. Weak answers list three variations of the happy path.
Design practice.
You are asked to add a "recently viewed" list to an existing product page, showing the last 10 items a user looked at, persisting across devices.
Talk through what you would need to know before designing anything: how many users, how often reads happen versus writes, what happens when the user is signed out, whether ordering must be exact, and what is acceptable to lose. The design area Microsoft describes is about understanding "what it takes to build a new system," and the fastest way to fail it is to start proposing a schema before establishing the constraints.
Behavioural practice, for the final round.
Prepare a genuine answer to: Why this role, specifically? This sounds trivial and is the single most commonly reported stumble at the AA stage — see the candidate accounts below.
The most useful candidate material about Microsoft concerns the final round, and it comes with an unusual bonus: interviewers turn up in the threads.
On Blind in September 2026, a candidate interviewing at L62 described their AA round with evident anxiety: "I did very well and answered every questions with good datapoints. But, I didn't know how to answer 'why this role?'" A reply in the same thread came from a self-identified AA interviewer: "I'm an AA. You handled it fine. I wouldn't worry."
Other Blind commentary consistently frames the round as fit rather than gate: "this is mostly manager round. You have already won 90% battle," and "AA rounds generally is about how good you fit in the team… Yeah standard behavioral, it might even end in 30 mins if they think you are fit."
On Glassdoor in June 2025, a candidate in Seattle described a full software-engineering loop with no online assessment at all — "phone screening + full loop of 3 rounds" covering an LRU cache implementation, a medium-difficulty algorithm problem and a binary sorting problem. That matters: multiple 2025 loop reports contain no OA, which is consistent with Microsoft never documenting it as a required stage.
A Glassdoor report from October 2025 described "three rounds including behavioral, technical, and bar raiser round," covering system design, microservices and distributed systems. Note the candidate's use of "bar raiser" — Amazon's term. The vocabulary confusion between the two companies runs in both directions, and it is worth being precise in your own head about which company's process you are preparing for.
After the online assessment, where one is issued, Microsoft publishes no score, band or threshold.
After the initial interview for students, you have been assessed for a function rather than a team. Progression means moving to the final interview, where team matching happens.
After the loop, feedback from every interviewer is collected before the final conversation. This is the mechanism that gave the AA round its original purpose — the senior interviewer reads what everyone else said first, and in current practice will often use their time on whatever the earlier interviewers did not cover.
After the final round, Microsoft's student guidance says the decision window can extend up to 90 days. That is a long time to wait, and it is officially stated, so do not read silence as rejection.
Ignore anything built on "Microsoft Leadership Principles." It is a category error. Prepare against Microsoft's actual published values — respect, integrity, accountability, growth mindset — and against the four technical areas it names.
Practise asking clarifying questions out loud. Microsoft names this explicitly under problem-solving. In a 45-minute interview, spending the first 3 minutes establishing constraints is not lost time; it is scored time.
Write real code, in one language, well. Microsoft says to code only in the language you are strong in, and not to use pseudocode. Pick one and get fluent rather than spreading across three.
Prepare tests, not just solutions. Testing is one of four named evaluation areas and is where average candidates separate from strong ones. After solving any practice problem, write the test cases before you look at the answer.
Prepare the "why this role" answer properly. It is the most reported stumble at the final stage, and it is entirely preparable. Have a specific reason tied to the team's actual work, not a general enthusiasm for Microsoft.
Do not over-prepare for the AA round as a trap. Microsoft's own blog says the term is legacy jargon and the round is now universal. Candidate and interviewer accounts describe it as behavioural and fit-focused, sometimes ending early when the answer is already clear.
Follow Microsoft's published AI policy. It is explicit: "We encourage candidates to use AI tools responsibly and ethically during preparation, as long as it reflects their true capabilities. During assessments and interviews, candidates should demonstrate their own skills without outside assistance unless explicitly permitted."
No. Those are Amazon's. Microsoft publishes core values — respect, integrity, accountability — alongside growth mindset, and no source shows candidates being scored against Amazon's framework or against Microsoft's internal "Model, Coach, Care" manager framework.
"As appropriate." Microsoft's own developer blog explains that the final interview originally took place only if the senior interviewer felt it was appropriate, and that it later "stopped being optional. Everybody got an 'As appropriate' interview." The same post calls the term dated jargon.
Microsoft does not document any such veto. Candidate and self-identified interviewer accounts on Blind describe it as largely behavioural and fit-focused, with one interviewer calling a worried candidate's performance fine.
It is not documented on any Microsoft careers page. Candidate reports describe one for university and new-grad software roles, but multiple 2025 loop reports contain no OA at all.
Microsoft names no vendor. Codility is the only one with first-hand candidate support, including a 2024 Blind report of a test that ran 110 minutes with 2 questions. HackerRank and CodeSignal claims come from prep sites only.
Microsoft states that interviews run 45 minutes and that most loops include 2 to 4 conversations "each lasting up to an hour."
Microsoft's own guidance states that its interview process is "currently entirely virtual."
For student hiring, Microsoft says the decision window can run up to 90 days. It publishes no timeline for experienced hires.
Yes for preparation, no during assessments. Microsoft encourages responsible AI use in preparation "as long as it reflects their true capabilities," and states that during assessments and interviews candidates "should demonstrate their own skills without outside assistance unless explicitly permitted."
Use TestSolve to crack the Microsoft assessment
AI-powered answers and reasoning sent to your phone. Free first solve, no signup.
Try it free →See pricing