Skip to content
Tech Interview Prep
Technical interview guide

Behavioral Interviews & STAR

Structured behavioral interviewing: STAR letters with numbers, conflict and failure without theater, and job-related scoring.

Read
34 min
Practice MCQs
25
Interview Q&A
25
Edition
v4
Editorial status
Reviewed

Scope: STAR as used in DDI-style behavioral interviews; structured interviewing per U.S. OPM; U.S. federal pre-employment guidance per EEOC and DOJ; principle-mapped loops as published by Amazon.

Curated: · Written: · Reviewed:

25 questions

MCQ-2You sit on an interview loop that scores behavioral answers on a four-anchor form - Situation, Task, Action, Result - and each entry is taken from the candidate's own words. A backend candidate answers a question about a hard week: 'Last Black Friday, checkout p99 went from 400 ms to 1.8 s under a 4x traffic spike. The team decided to add cache nodes. We moved session reads to Redis with a 5-minute TTL and a stampede lock, and things went well - the team learned a lot about caching.' Which set of entries can you actually record on the form?
MCQ-4A candidate's answer to 'tell me about a time you handled a production problem' opens with: 'Last quarter checkout had a bad outage and the team needed to make things more reliable.' The interviewer's rubric awards Task points only for a decision the candidate personally owned, with its constraint and deadline. Which rewrite of that opening line earns the Task points?
MCQ-3You record a 4-minute mock answer to "tell me about a time you cut latency." The breakdown is Situation 95 s, Task 20 s, Action 70 s, Result 15 s. The interviewer's only note: "I still don't know what you decided or what changed." Which revision fixes the answer the note is describing?
MCQ-5A candidate answers "tell me about a time you improved reliability" in five sentences, and the interviewer's Action row comes back 1 of 3. The rubric scores Action as what the candidate personally did, in verbs a skeptic can check: the trade-off you chose and why, the artifact you wrote, the number you measured, the thing you reverted. 1. Our checkout service was burning its error budget at 4x, p99 at 900 ms. 2. I asked the SRE team to look at the latency dashboards. 3. I set up a weekly sync with the pricing team and I pinged the on-call about the cache. 4. We cut p99 to 120 ms and halved the error rate after the cache shipped. 5. We learned a lot about cache invalidation. Which single edit moves Action from 1 to 3?
MCQ-8The interviewer opens a project question with: 'I've read your resume and I know the team shipped it. I want the decisions you personally made.' You're describing a migration five engineers delivered. How should you frame the answer?
MCQ-6An interviewer asks for 'the result' of the checkout latency project you led. Your notes: p99 checkout latency was 1.8 s before the rollout and 320 ms in the 6 weeks after it, against a 4-week pre-rollout baseline; error-budget burn was 12% a month before and 2% a month after; the change shipped 2 days ahead of a 3-week estimate; four other services now use the cache client; finance projects about $18k a month less compute at current traffic; the VP called the rollout 'transformative'. Which closing sentence should you give as the Result?
MCQ-7Senior backend behavioural round. Your story is a retry-path fix for duplicate card charges that you shipped on 2 March, and no other change to checkout went out in either window below. The support queue tags duplicate-charge tickets, and you pulled two windows: 5 January to 1 March (eight weeks) shows 96 tickets against 240,000 checkouts; 2 March to 29 March (four weeks) shows 20 tickets against 180,000 checkouts. The interviewer asks how you knew it worked. Which Result statement do you give?
MCQ-20A behavioral interview. The candidate's answer so far: 'Duplicate signups in the payments service. Three backend engineers and one SRE on the team. My task was to stop duplicate inserts, so I added a unique index on (account_id, idempotency_key) CONCURRENTLY, cut the old write path, and the duplicates stopped.' Their dashboard recorded 12,041 duplicate inserts in the 47 minutes before the index went live and 0 in the 47 minutes after. The interviewer follows up: 'How did you know it worked?' Which single reply answers the follow-up?
MCQ-24You are in a senior-level loop at a 400-engineer company. The req reads: 'Owns a service end to end; leads a cross-team migration with one partner team.' The interviewer asks: 'Tell me about a time you drove a technical change through another team.' Your story bank holds three entries: (A) cut p95 on your team's checkout service from 400 ms to 180 ms with a read-through cache, no other team involved; (B) led a dual-write migration of the orders table across 3 services with one partner team, 7 days of dual-write, then cut over, aborting on replication lag > 2 s; (C) wrote an RFC that 11 teams adopted as the standard for schema changes, with an org-wide freeze during rollout. Which story do you tell?
MCQ-21The loop email lists the two principles each interviewer scores, one evidence paragraph per principle, and slot 3's line reads Dive Deep and Earn Trust. That interviewer gives you 15 minutes of behavioral time before the rest of the hour turns to a design prompt. Three stories are banked, with the running time each needs so its outcome survives: the checkout trace runs 6 minutes (the team blamed GC; the traces put p99 at 180 ms on a connection pool capped at 20 connections against 400 rps, and two hypotheses died on the data before the fix took p99 to 40 ms), the 01:12 rollback runs 9 (you argued the platform lead out of holding the release, then wrote the runbook both teams still use), and the 31% cost cut on the batch tier runs 4. The interviewer opens with 'Walk me through the hardest problem you have ever had to debug.' How do you spend the 15 minutes?
MCQ-25Five behavioral rounds next week: conflict, failure, saying no, an incident, and influencing without authority. Your only story with hard numbers is a checkout migration that cut p99 from 900 ms to 220 ms, and you plan to tell it in every round, re-angled to fit each prompt. In round two the interviewer asks for a time you disagreed with a peer; you start the migration again and they cut in: "we've heard that one." What was the actual gap?
MCQ-22Priya has a staff-level loop in five days and one migration story: 12,041 duplicate rows, a missing unique key on the orders table, 47 minutes of write downtime. She has memorized a 400-word script for it and can recite it in 2:10 without notes. In a mock, the interviewer let her finish, then asked three probes: how she would roll back, who got paged, and what she would do differently. Priya answered all three with the same third paragraph of her script. Which change to her preparation most directly fixes what the mock exposed?
MCQ-10When asked 'Tell me about a time you failed,' which answer structure best demonstrates genuine ownership and authentic learning without deflection or artificial redemption?
MCQ-11An interviewer asks how you drove a cross-functional initiative when no one on it reported to you. The case you have in mind: three service teams, each with its own latency dashboard green for two quarters, while end-to-end p99 on the shared call path stayed flat at about 900 ms and the three tech leads had begun blaming each other's hops. Which narrative earns credit under a structured behavioral evaluation?
MCQ-16You're in a staff-level behavioral loop and the interviewer asks for a "developing others" story. Your best material: over two quarters you ran weekly design reviews with Sam, a mid-level engineer, on the payments retry path. You wrote the first draft of the retry-budget spec; Sam rewrote it and shipped it, and p95 fell from 900 ms to 210 ms. Six weeks after you moved to another team, Sam shipped the next retry budget alone; the abort path fired once with no customer impact. Which story do you tell?
MCQ-12Your manager has committed the team to a launch date that depends on a service whose p99 latency under peak load has never been measured. You believe the date is unsafe. Which response best demonstrates professional upward disagreement?
MCQ-9You're 40 minutes into a staff-level behavioral interview. You've just described a rollout disagreement: your peer wanted to ship a payments service on the committed launch date with a 2 s client timeout and three retries; you wanted to delay for a retry-path rewrite. Your load test showed the retries amplified traffic 3x during a downstream slowdown and drove the service to 100% errors. You shipped on the launch date with an 800 ms timeout and a circuit breaker. The interviewer leans in: "So your peer's plan would have taken payments down?" Which response should you give?
MCQ-15You're a staff engineer on Payments. Search owns the index schema and freezes it in 24 hours. Your dual-write needs 72 hours to prove out, so you need the freeze pushed 48 hours. You have no authority over Search. A 12-hour canary of the dual-write path shows a 0.02% error rate against Search's current 0.4%. What do you do?
MCQ-13A PM asks for CSV export of the admin table to be added on top of the committed launch scope, two weeks before launch. The table holds 10M rows. The PM sizes the work at 2 engineer-days, from a 2023 export that read 40k rows. The sprint has 40 engineer-days and 36 are committed to launch blockers, including a three-day load test. The team's billing export is the only one built on this query path and it took 6 engineer-days. The full-table scan adds about 400 ms to the admin p95, which is 250 ms today against a 300 ms launch acceptance bar. Which reply should the engineer give?
MCQ-14You are the tech lead on a payments cutover booked for Wednesday 09:00. Monday 09:00: the dual-write backfill is 62% complete, and the 38% still queued is 1.8 billion rows at a measured 4k rows/s, which lands Saturday 14:00. Raising the write cap for six hours would clear it before Wednesday, but takes checkout p99 from 180 ms to 260 ms, and the load test for that change is not booked until Tuesday. The launch channel that owns the date meets at 10:00, and your last note to it said 'on track'. What do you do on Monday morning?
MCQ-18You're rehearsing a 90-second staff-level answer. The story: checkout writes go synchronously to the primary, moving p95 from 40 ms to 90 ms with no staleness. The design you rejected read from an async replica: p95 stays 40 ms, but the replica lags the primary by up to 800 ms. Fraud review must see the payment within 200 ms of authorization. Which version of the answer holds up?
MCQ-17A postmortem draft for last Tuesday's outage reads: 02:14 config merge removed a unique idempotency key; retries fired at 0 ms and produced duplicate charges; 5xx climbed from 0.3% to 7.1%; on-call reverted at 02:31. The root-cause line says 'Dana merged the change without checking the retry path.' What should the reviewer require before the postmortem ships?
MCQ-19A team runs a 30-minute weekly SLO review. The agenda is fixed: 5 minutes to open on the error-budget summary, 5 minutes at the end to assign actions, and the remaining 20 minutes split across the services on the dashboard. Policy: a service burning faster than 14.4×, the 1-hour fast-burn threshold for a 30-day SLO, gets a deep dive; every other service gets a 2-minute confirmation. This week 6 services are on the dashboard and 4 are above the threshold. The team wants to protect the closing block and use the whole 20 minutes. What should the facilitator do?
MCQ-1A team replaced its free-form loop with a structured one: five job-analysis questions asked in the same order, every interviewer scoring 1–5 against written anchors. Two habits survived the rewrite: interviewers fill in their score sheet during the debrief, and the debrief runs until the panel reaches consensus. A crossover study on 90 candidates, each seen by two panels, shows panel-to-panel agreement of 91% on candidates who end up clearly hired or clearly rejected, and 54% on the borderline band (mean score 3.0–3.8). Inside that band, the interviewer who speaks first in the debrief ends up on the winning side 78% of the time. What is the highest-value change?
MCQ-23A candidate is in a 30-minute recruiter screen for a backend role at a U.S. company. The recruiter asks three questions in a row: "Do you have any medical conditions that would affect your ability to work?", "Are you planning to have children soon?", and "Are you authorized to work in the U.S., and will you need sponsorship?" Which response fits U.S. federal rules?