Project deep-dive and retrospective rounds

Meta’s interview guide tells candidates to “be prepared to discuss projects in depth”, and suggests outlining two or three major projects ahead of time. Amazon’s page for software development managers says you may be asked to “dive deep into the most complex system architecture you’ve supported and explain the trade-offs you made”. One past project can then fill most of a 45-minute round, with the interviewer choosing where to dig.

Project deep-diveproject deep dive: A round, or part of one, that explores one of your past projects in detail: the problem, your decisions and their trade-offs, and the results. More in “Answer length, layering and follow-up probes” rounds and retrospective roundsretrospective round: An interview round, or part of one, that asks you to look back over a past project with hindsight: what went well, what went wrong and what you would change. both look back at work you did; neither asks you to design something new. Below are what each round asks, the ladder an interviewer walks down, and one project prepared for it.

A deep dive goes down one project; a retrospective looks back over it

In a project deep-dive, the interviewer picks one of your projects and goes down it one question at a time. The questions run from why it existed to the detail of what you built. Retrospective rounds ask you to look back over a whole project with hindsight: what went well, what went wrong and what you would change. A deep dive often ends with a retrospective question, and a retrospective often asks for depth on one decision, so prepare each project for both.

Neither is a system design round. A design round asks you to design something new, and the system design track teaches it; a deep dive asks about something you already built, so every answer can be checked by the next probe. Amazon’s manager page lists what it wants to understand in such a dive: the how and why, the obstacles, the lessons and “your level of involvement and specific contributions”.

Both rounds test what a single two-minute story cannot show at length: your role, your decisions, the alternatives you rejected and the results. A deep dive usually goes down five rungs, a MySmartCoPilot depth ladderdepth ladder: The order in which a project deep dive usually goes down: the business goal, the system's boundaries, a component you built, a hard bug or incident, and the numbers. The interviewer may move up or down it in any order.: the business goal, the system’s boundaries, a component you built, a hard bug or incident, and the numbers. Step down it with Ishaan, an invented backend engineer, and his reconciliation service:

The drawing shows the whole scene; its 7 steps are listed below.

Every step as text

1. The business goal. Ishaan, an invented backend engineer, starts where the interviewer starts: finance matched card settlements against bank files with hand-written scripts, and clearing the mismatches took most of a week each month.
2. The boundaries. Then the edges of the system: bank files come in once a night in three formats; matched records, a queue of exceptions for finance and daily reports go out. Nothing on the banks' side could change.
3. His component. The interviewer asks what he built himself: the matching engine, which pairs each settlement with its bank record and sends anything it cannot pair to the exceptions queue. This is where his own decisions live.
4. The hard bug. A rung lower: a bank re-sent a night's file after an outage. Keying the intake on a checksum of each file meant the repeat was rejected, so nothing was matched twice.
5. Back up a rung. The interviewer may climb back up at any moment, for example to ask why the files are not streamed instead. Each rung has to stand on its own, in any order.
6. The numbers. At the bottom, the numbers: the automatic match rate rose from about 90 % to above 99 %, and clearing month-end exceptions fell from 4 to 5 days to under 1. Ranges keep the exact figures private.
7. The whole ladder. Five rungs, from why the project existed to the numbers that show it worked. Prepare every rung for each of your two or three projects, and the interviewer can dig wherever they choose.

Every rung has an answer ready, and the interviewer can move up or down in any order. A one-page brief per project is the preparation for both kinds of round. It holds each rung of the ladder and what you would change, which is where retrospective rounds begin.

A one-page brief, with ranges where the numbers are confidential Pro: opens with a Pro, Premium or Ultimate pass

In this part:

• Thirty-five minutes down the ladder, probe by probe
• “Why not X?”: the alternatives you rejected, and hindsight
• Keeping confidential details out and the substance in
• Running a retrospective round aloud
• Practise: brief two projects
• Interview questions
  • Pick one decision in the project you just described. What was the strongest alternative, and why did you reject it?
  • Sketch the architecture of your last project, then tell me where it would break first at ten times the load.
  • Run a retrospective on your most recent project out loud. What went well, what went wrong, and what would you change?
• Key takeaways

1 diagram · 1 runnable example (JavaScript) · 3 interview answers · about 27 minutes

Opening this part with your pass…

This part opens with any MySmartCoPilot pass.

• Pro pass · 1 week · ₹79 · $2.99
• 30 days · ₹179 · $5.99
• 1 year · ₹1,699 · $45

One payment; a pass never renews. The price shown is the total you pay: no fees or charges are added at checkout.

Choose a passI have a licence key

It also opens every other in-depth part in Learn and every Pro tool. Sales are final, except where the law gives you a right to withdraw, for example in the EU and the UK (refund policy).

Terms in this lesson

depth ladder

The order in which a project deep dive usually goes down: the business goal, the system's boundaries, a component you built, a hard bug or incident, and the numbers. The interviewer may move up or down it in any order.

retrospective round

An interview round, or part of one, that asks you to look back over a past project with hindsight: what went well, what went wrong and what you would change.