Skip to content

A level deeper on your project

Being able to say how any part of your project worked, one layer below the headline, when the interviewer asks. In a project deep dive, that layer is what separates work you did from work you watched.

On this page

What it is

A project deep dive asks you to walk through something you built while the interviewer interrupts to ask about the parts. Going a level deeper means that for each tool, method, data model or metric you name, you can say how it worked in your project, in the terms your field uses, and why it was that and not something else. Ashby, which publishes a candidate-facing guide to its engineering past-projects interview, says the candidate leads and gets stopped with questions, sometimes very deep ones about a piece the interviewers find interesting, and is then brought back to where they left off.

Employers run the round this way because a project’s headline reads the same whether you made the decisions or sat near the person who did. A walkthrough of real work, with questions, shows how you gathered information, decided, planned and handled setbacks, which a timed exercise with little context doesn’t show well. Those things live a layer below the headline: in the field you added to the data model, the option you turned down, the week the rollout stalled.

You saidThe layer belowWhat it shows

“We added a queue between the API and the workers.”

What each message held, how it was keyed and ordered, what happened to a failed delivery, and why a queue rather than a direct call.

You understand the mechanism, not just the name.

“I built the data model.”

The main entities, the field that was hard to get right, and what changed in it after launch.

Recall of the kind of detail you have if you did the work.

“I ran an experiment.”

How users were split, how long it ran, the metric that decided it, and what you’d have done with a flat result.

Method, in the field’s own terms.

“It cut latency.”

Measured where, which percentile, before and after, and whether it held a quarter later.

A result you can defend.

“It shipped.”

When, what broke first, what the team changed after you left, and where it stands now.

What became of it.

What happened after. The after-story is part of the depth: lessons learned after release are on the list of things the interviewers want to understand, and a project you left before release is a weaker pick, because you didn’t see the outcomes of your own decisions. So carry the project past launch: whether the metric moved and whether it held, what broke first, what the team changed later, and whether it still runs. If you did leave before the end, say so and give what you know from former colleagues rather than the result you expected. Putting numbers on impact covers the figures themselves.

When to use it

Use it whenever you present your own past work and the other person can ask questions: a project deep dive or retrospective round, the project part of a hiring-manager screen, a walk through your resume. Ashby, for example, goes through a project twice: 10 to 20 minutes in the first call with an engineering leader, then 30 or 45 minutes in the final round, and allows the same project both times as long as there’s enough in it to fill the longer slot.

Pick the project with the depth in it. Ashby’s expectations ask for one you led and made the key decisions on, where you can still recall details such as properties on the data model, milestone dates and how customers reacted, and where you can explain why decisions went the way they did, including technology choices, trade-offs, testing and reliability. Projects it finds less interesting: mostly plain create-read-update-delete work, migrations from one technology to another, and projects that were mainly cross-team coordination or internal politics. If you’re unsure which to bring, asking the hiring manager is welcome there and isn’t held against you. Choosing the right story has more on picking.

Some employers publish values that tell you how much detail they prize. Amazon’s leadership principles include Dive Deep: leaders operate at all levels, stay close to the details, audit often, and are skeptical when the numbers and the anecdotes disagree. Our advice: when a company publishes a value like that, prepare the second layer of your project to that standard, including the figures and where they came from.

When not to

When the question wants the headline. “Tell me about a project you’re proud of” at the start of a round, or “walk me through your resume”, asks for the shape first: the problem, your part, the result, in a couple of minutes. Opening with the schema buries the point the interviewer needs in order to choose where to dig. Give the headline, pause, and let them pick. The two-minute project pitch covers that opening.

When the listener isn’t technical. In a technical deep dive you can assume the panel knows basic concepts; in a recruiter screen or a cross-functional panel you can’t. There, give the business layer, what it did for users and what changed, and make the deeper layer a short offer (“I can go into how the matching worked if that’s useful”) rather than the default.

When the piece wasn’t yours. Going deep on a part a teammate built, as if you had built it, invites the next question, which is the one you can’t answer. Say where your part ended, give what you understood of theirs, and move the depth to a piece that was yours. Saying “I”, not “we” covers the same boundary from the other direction.

When the question is about behavior, not mechanism. “Tell me about a disagreement on that project” asks about the decision and the people. The mechanism gets one sentence of context, and the depth goes to what you did and why. The STAR method is the shape for those.

Compared

Follow-upWeak answerStrong answerWhen the shorter answer is right

“How did that work?”

“It’s a standard caching layer in front of the database.”

What was cached, keyed by what, how entries were dropped when the source changed, and why a cache instead of a faster query.

A non-technical listener, or one who has signaled they want the shape: one sentence, then offer more.

“Why that choice?”

“It’s what the team used.” or “It’s best practice.”

The option you considered and rejected, the constraint that decided it, and what the choice cost you later.

You inherited the decision: say so, then say what you’d have chosen and why.

“What happened after?”

“It went well, the team was happy.”

When it shipped, the metric before and after and whether it held, what broke first, where it stands now.

You left before release: say so, give what you know from former colleagues, and name what you’d have watched.

“What were the exact numbers?”

“I don’t remember, it was a lot.”

The figure, or an honest range tied to something you do remember: “about a tenth of the traffic, because it was the EU region.”

You truly don’t recall: an honest range with its reasoning is more useful than a confident wrong number.

In a picture

Your headlineHow did itwork?Mechanism hereWhy that one?Option, reasonWhat cameafter?Outcome and now
A follow-up that asks for depth wants the next layer down. Answer it, then return to the walkthrough.

How to say it in an interview

The common mistake

Quiz

The interviewer asks how the message broker was configured. You wrote the producer side; a teammate set up the broker. What do you say?
Which detail does most to show you did the work yourself?
You left the company before the project shipped. The interviewer asks what happened after. Best answer?
Which preparation fits a project deep dive best?

Related

Sources

  1. Engineer Past Projects Deep Dive (Ashby)
  2. Amazon’s Leadership Principles (About Amazon UK)
Sign in to report a mistake