Amazon Data Analyst Interview Guide

Amazon's Data Analyst interview loop is designed so that a single weak behavioral round can sink an otherwise flawless technical performance. The Bar Raiser, an interviewer from a completely different org with veto power, will spend 45 minutes probing your Leadership Principles stories for specifics. Candidates who treat that round as a soft toss tend not to make it through.

Amazon Data Analyst Role

This role is less "dashboard builder" and more "the person who explains why Subscribe & Save churn spiked in Q3 and what the team should do about it." You'll own the metric narratives that feed Amazon's Weekly Business Reviews, write root-cause analyses in six-pager format against Redshift data, and build self-serve QuickSight dashboards that replace manual Excel reporting. After year one, success looks like stakeholders pulling your dashboards into their own planning docs without asking you to double-check the numbers.

A Typical Week

Writing eats almost as much of your week as analysis does. That's the part most candidates don't anticipate. Six-pager drafts, JIRA documentation, narrative appendices for the WBR: these aren't side tasks, they're the job. The other surprise is how much time goes to data infrastructure firefighting, chasing duplicate rows from upstream ETL deploys or fixing QuickSight calculated fields that broke after a schema change.

Projects & Impact Areas

Your project mix depends on which org you join. In Stores, you might build a 90-day churn cohort analysis for Subscribe & Save by joining subscription events against order history in Redshift, then write the recommendation doc that kicks off a re-engagement experiment. Ads work looks completely different: stitching together advertiser spend and purchase conversion data to measure campaign attribution across retail media. Across all orgs, you'll maintain data collection processes and automate reporting workflows, not just query tables that someone else set up.

Skills & What's Expected

Business acumen is rated higher than any technical skill on the internal scorecard, and most candidates underweight it. Amazon wants you to frame a problem in terms of customer impact or revenue before you open a query editor. SQL proficiency is the technical foundation, but strong metric formulation and written communication will separate you from candidates who over-index on Python or R. ML knowledge isn't part of the role's requirements.

Levels & Career Growth

Most experienced external hires land at L5, where you're expected to own analyses end-to-end with real autonomy. The L5-to-L6 jump is where careers stall: it requires demonstrating influence beyond your immediate team, like shaping a product roadmap or leading a cross-functional project that touches multiple stakeholder groups. One genuine perk is that lateral moves across orgs (Stores to AWS, Ads to Devices) don't reset your level, so you can diversify your experience without sacrificing progress.

Work Culture

Your analysis narratives will get red-penned by your manager before stakeholders ever see them, and leaders will drill into your data tables during reviews without hesitation. Leadership Principles show up in every performance review and hiring debrief, not just on posters. The source data describes a three-days-in-office norm (Tuesday through Thursday), though Amazon's return-to-office policies have been tightening, so confirm the current expectation for your specific team before accepting an offer.

Amazon Data Analyst Compensation

The gap between your offer letter and your actual year-one paycheck can be 20-30% wider than you expect. Sign-on bonuses mask the backloaded vesting in years one and two, but they taper right as your stock starts catching up. Refresher grants kick in after your second performance review, though they vary significantly by rating, so don't count on them to fill the gap. Leaving before year three means walking away from the fattest portion of your equity.

Base salary has limited room to move, so don't burn negotiation capital there. Your real levers are the RSU grant size and sign-on bonus, especially the year-two sign-on, which covers the period where vesting is still thin and the year-one bonus has already dropped off. A competing offer from another large tech company is the single strongest card you can play; from what candidates report, recruiters have more flexibility on RSUs and sign-on when you bring a credible competing number to the table.

Amazon Data Analyst Interview Process

The widget above maps every round, but here's what it can't show you: the loop rounds (SQL through Bar Raiser) are scheduled back-to-back in a single virtual day, so you're doing five consecutive 60-minute interviews with short breaks. That's a marathon. Shallow behavioral stories are one of the most common reasons candidates wash out, right alongside insufficient technical depth in SQL or statistics. The difference is that most people prep for the technical rounds and underestimate how relentlessly interviewers probe STAR examples, especially for Dive Deep, Bias for Action, and Customer Obsession.

The Bar Raiser round deserves special attention because this interviewer carries outsized influence in the post-loop debrief. They're from a different org entirely, trained to evaluate whether you'd raise the average quality of Amazonians at your target level. From what candidates report, a strong "no" from the Bar Raiser is very difficult for the rest of the panel to override. That makes this round functionally higher-stakes than any single technical interview, even though it blends behavioral and technical questions together.

Amazon Data Analyst Interview Questions

/*
Daily on-time shipment rate (order-level) for the last 14 days.
Assumes a Redshift-like dialect where DATEADD supports hour granularity.
*/
WITH orders AS (
  -- Collapse line items to one row per order (order grain)
  SELECT
    oli.order_id,
    MIN(oli.order_date) AS order_ts,
    CAST(MIN(oli.order_date) AS DATE) AS order_dt
  FROM order_line_items AS oli
  WHERE CAST(oli.order_date AS DATE) >= DATEADD(day, -14, CURRENT_DATE)
  GROUP BY
    oli.order_id
),
first_shipped AS (
  -- Get the first SHIPPED timestamp per order
  SELECT
    se.order_id,
    MIN(se.event_ts) AS first_shipped_ts
  FROM shipment_events AS se
  WHERE se.event_type = 'SHIPPED'
  GROUP BY
    se.order_id
),
order_flags AS (
  SELECT
    o.order_id,
    o.order_dt,
    CASE
      WHEN fs.first_shipped_ts IS NOT NULL
           AND fs.first_shipped_ts <= DATEADD(hour, 48, o.order_ts)
        THEN 1
      ELSE 0
    END AS is_on_time
  FROM orders AS o
  LEFT JOIN first_shipped AS fs
    ON o.order_id = fs.order_id
)
SELECT
  ofl.order_dt,
  COUNT(*) AS total_orders,
  SUM(ofl.is_on_time) AS on_time_orders,
  (SUM(ofl.is_on_time)::DECIMAL(18,6) / NULLIF(COUNT(*), 0)) AS on_time_rate
FROM order_flags AS ofl
GROUP BY
  ofl.order_dt
ORDER BY
  ofl.order_dt;

What jumps out isn't any single dominant area. It's that Amazon splits evaluation weight almost evenly across technical, product, and statistical reasoning, then layers a behavioral round scored against Leadership Principles on top. The compounding difficulty lives where A/B testing meets Amazon's marketplace reality: questions about experimentation assume you understand buyer-seller interference, delivery promise tradeoffs, and why a naive randomization on checkout widgets can contaminate seller-side metrics. Candidates from single-sided product companies tend to prep clean textbook experiments and then stall when asked how they'd isolate treatment effects in a two-sided marketplace like Amazon Retail.

Prep for these question types with Amazon-specific scenarios at datainterview.com/questions.

How to Prepare for Amazon Data Analyst Interviews

Amazon reported roughly $717 billion in revenue for FY 2025, up 13.6% year over year. The three bets that most directly shape DA work right now: AWS racing to make AI inference cheaper with custom silicon, Amazon Ads building AI-powered campaign optimization and creative tooling, and Stores expanding same-day pharmacy delivery to nearly 4,500 cities while pushing Prime speed records even further.

Which bet your target team sits under changes everything about the interview. An Ads DA needs to talk about connecting media exposure to shopping behavior across streaming and display. A Stores DA should speak to delivery promise accuracy or Subscribe & Save retention. AWS? Churn signals for enterprise accounts and inference cost metrics. Walk into your loop knowing the specific business problems your team owns, not just the segment name.

Candidates often fumble "why Amazon" by vaguely praising customer obsession as a philosophy. The Leadership Principles aren't decorative, though. They're the literal evaluation rubric in every interview round, including the Bar Raiser's. Instead of generic admiration, pick a concrete initiative (say, the Rufus AI assistant's focus on price tracking and automated purchasing) and explain which LP your past work maps to in solving a similar problem. That's how you show you understand Amazon's decision-making language, not just its press releases.

Try a Real Interview Question

| order_id | fc_id | shipped_at           | promised_delivery_at  | delivered_at          |
|----------|-------|----------------------|-----------------------|-----------------------|
| O1001    | FC_A  | 2023-12-10 08:00:00  | 2023-12-12 20:00:00   | 2023-12-12 19:00:00   |
| O1002    | FC_A  | 2023-12-20 09:00:00  | 2023-12-22 20:00:00   | 2023-12-23 10:00:00   |
| O2001    | FC_B  | 2023-12-15 07:30:00  | 2023-12-18 20:00:00   | 2023-12-18 18:00:00   |
| O3001    | FC_A  | 2024-01-05 10:00:00  | 2024-01-07 20:00:00   | 2024-01-08 08:00:00   |
| O3002    | FC_A  | 2024-01-18 11:00:00  | 2024-01-20 20:00:00   | 2024-01-20 17:00:00   |
| O4001    | FC_B  | 2024-01-09 06:00:00  | 2024-01-12 20:00:00   | 2024-01-11 12:00:00   |
| O4002    | FC_B  | 2024-01-22 06:00:00  | 2024-01-25 20:00:00   | 2024-01-26 09:00:00   |
| O5001    | FC_C  | 2024-01-03 05:00:00  | 2024-01-06 20:00:00   | 2024-01-06 19:30:00   |

WITH delivered AS (
  SELECT
    order_id,
    fc_id,
    delivered_at,
    promised_delivery_at,
    DATE_TRUNC('month', delivered_at) AS delivery_month,
    CASE WHEN delivered_at <= promised_delivery_at THEN 1 ELSE 0 END AS is_ontime
  FROM shipment_events
  WHERE delivered_at IS NOT NULL
    AND delivered_at >= TIMESTAMP '2023-12-01'
    AND delivered_at < TIMESTAMP '2024-02-01'
),
month_rates AS (
  SELECT
    fc_id,
    delivery_month,
    COUNT(*) AS total_delivered_orders,
    AVG(is_ontime::DECIMAL) AS on_time_rate
  FROM delivered
  WHERE delivery_month IN (TIMESTAMP '2023-12-01', TIMESTAMP '2024-01-01')
  GROUP BY 1, 2
),
jan AS (
  SELECT
    fc_id,
    total_delivered_orders AS jan_total_delivered_orders,
    on_time_rate AS jan_on_time_rate
  FROM month_rates
  WHERE delivery_month = TIMESTAMP '2024-01-01'
),
dec AS (
  SELECT
    fc_id,
    on_time_rate AS dec_on_time_rate
  FROM month_rates
  WHERE delivery_month = TIMESTAMP '2023-12-01'
),
final AS (
  SELECT
    j.fc_id,
    j.jan_on_time_rate AS on_time_rate,
    j.jan_total_delivered_orders AS total_delivered_orders,
    (j.jan_on_time_rate - COALESCE(d.dec_on_time_rate, 0)) AS rate_change_vs_dec
  FROM jan j
  LEFT JOIN dec d
    ON j.fc_id = d.fc_id
)
SELECT
  fc_id,
  on_time_rate,
  total_delivered_orders,
  rate_change_vs_dec
FROM final
ORDER BY rate_change_vs_dec ASC, total_delivered_orders DESC, fc_id ASC
LIMIT 1;

From what candidates report, Amazon's SQL questions tend to involve layered joins and require you to reason about how tables relate before you start writing. The Bar Raiser or technical interviewer may push you to explain why you'd structure a schema a certain way, not just whether your output is correct. Build that habit at datainterview.com/coding, where you can practice on e-commerce-style datasets that mirror the complexity you'll face.

Test Your Readiness

Spot your weak points across product sense, metrics, and statistics with Amazon-tailored questions at datainterview.com/questions.