Goldman Sachs Data Analyst Interview Guide

Most candidates prep for this role like it's a tech job that happens to sit inside a bank. That's the wrong frame. Goldman's data analyst position lives at the intersection of AWM Operations, Tax, and other divisions where your SQL output feeds directly into regulatory filings and MD-level decisions, so the people who struggle are the ones who can't translate a financial question into a data requirement without help.

Goldman Sachs Data Analyst Role

You'll work across divisions like Asset & Wealth Management Operations and Tax, writing production SQL against Snowflake and Oracle source systems, maintaining Tableau dashboards that track metrics like AUM flows and fee revenue, and fielding same-day ad-hoc requests from VPs who need answers before their next meeting. Success in year one means owning a recurring reporting workflow end-to-end (the weekly AWM executive dashboard, for instance) while earning enough trust that senior stakeholders route their data questions to you directly.

A Typical Week

The ratio that should reshape your prep strategy: analysis and writing together dominate your hours, while pure coding occupies a surprisingly thin slice. You'll spend real time in Confluence documenting metric definitions for Goldman's data governance process, where every new metric (like net-new-assets logic) needs sign-off from an AWM data steward before it can go live. That governance layer is invisible from the outside but eats meaningful calendar space inside.

Projects & Impact Areas

AWM Operations is where the hiring volume concentrates, and a typical quarter might have you building a fee revenue attribution pipeline that joins across three Snowflake schemas and a manual Excel crosswalk to segment advisory fees by client AUM tier. From there you could shift to standing up data quality monitors for a less-structured asset class like private credit, or partnering with engineering to resolve schema changes in a legacy Sybase IQ source table that broke an Alteryx workflow over the weekend. The connective thread is that your outputs don't just inform dashboards; they get consumed by compliance teams and regulators, which is why Goldman's review-before-publish norm exists for anything MD-facing or client-facing.

Skills & What's Expected

The source data scores software engineering (Python), data architecture, and business acumen all as high-priority dimensions, while machine learning and GenAI land at low. The implication for your prep is sharp: stop practicing XGBoost and start practicing how to explain what "net-new-assets" means in the context of AWM fund flows, because Goldman weights your ability to hold that conversation on par with your ability to write the query behind it. Candidates from pure tech backgrounds consistently miss this.

Levels & Career Growth

Most hires enter at the Analyst level with 0-3 years of experience, and the source data lists the hierarchy as Analyst, Associate, VP, then MD. Lateral moves into engineering, quant strategy, or product roles within Goldman are a real path because the firm invests in retaining institutional knowledge. What the levels don't show you is that advancement depends heavily on whether senior stakeholders trust you to operate independently on their requests, not just on your technical output.

Work Culture

Goldman requires five days in-office at 200 West Street or your assigned location (Dallas and Chicago are common), with no formal hybrid option for analyst-level roles. Expect 8:30 AM to 6:30 PM as your baseline, with late pushes around quarter-end reporting or regulatory deadlines. The upside is genuine structure: your analysis gets peer-reviewed before release, metric definitions go through a formal data steward approval, and the pace, while intense, is predictable enough that you can plan around it.

Goldman Sachs Data Analyst Compensation

Equity and RSUs become more common as you move up Goldman's hierarchy, so at the Analyst level your total comp skews heavily toward cash: base salary plus an annual discretionary bonus driven by firm performance and individual results. That bonus isn't formulaic, so treat any number HR mentions as a target rather than a guarantee.

Your most negotiable levers are level placement (Analyst vs. Associate), sign-on bonus, and anchoring to competing offers with scarce skills like advanced SQL, data modeling, or finance domain expertise. One detail worth confirming early: whether you'll be eligible for a full bonus in your first year, since a mid-year start can meaningfully reduce that payout. Locking in a signing bonus or full-year eligibility before you sign protects you from that gap.

Goldman Sachs Data Analyst Interview Process

The process spans roughly 8 weeks across 6 rounds. Weak SQL fundamentals are the most frequently cited rejection reason, with incorrect joins, wrong grain, and sloppy null handling on financial transaction data all surfacing in candidate reports. If you're coming from a tech background, the standalone Statistics & Probability round (round 3) is the other surprise: it's not folded into a case study, and questions skew toward finance-flavored probability (conditional default events, expected value of trading strategies) rather than standard A/B testing setups.

Goldman's Superday format packs 3 to 5 back-to-back interviews, and the overall tips in their process explicitly warn not to assume interviewers share context with each other. That means you'll re-introduce your background and re-anchor your stories multiple times in a single day. Inconsistency across rounds, like nailing the case study but rambling through the behavioral, signals low stakeholder readiness, which the data flags as a distinct deal-breaker in Goldman's high-visibility, controls-first environment.

Goldman Sachs Data Analyst Interview Questions

WITH ranked_trades AS (
  SELECT
    a.region,
    t.trade_id,
    t.account_id,
    t.symbol,
    t.trade_ts,
    t.quantity,
    t.price,
    ROW_NUMBER() OVER (
      PARTITION BY a.region
      ORDER BY t.trade_ts DESC, t.trade_id DESC
    ) AS rn
  FROM trades t
  INNER JOIN accounts a
    ON a.account_id = t.account_id
)
SELECT
  region,
  trade_id,
  account_id,
  symbol,
  trade_ts,
  quantity,
  price
FROM ranked_trades
WHERE rn <= 5
ORDER BY region, trade_ts DESC, trade_id DESC;

WITH daily_totals AS (
  SELECT
    as_of_date,
    account_id,
    SUM(qty) AS total_qty
  FROM positions_snapshot
  WHERE as_of_date BETWEEN :start_date AND :end_date
  GROUP BY as_of_date, account_id
)
SELECT
  as_of_date,
  account_id,
  total_qty,
  total_qty - LAG(total_qty) OVER (
    PARTITION BY account_id
    ORDER BY as_of_date
  ) AS delta_qty
FROM daily_totals
ORDER BY account_id, as_of_date;

WITH status_matches AS (
  SELECT
    t.txn_id,
    t.account_id,
    t.txn_ts,
    t.amount,
    s.status
  FROM transactions t
  LEFT JOIN account_status_history s
    ON s.account_id = t.account_id
   AND t.txn_ts >= s.effective_from
   AND t.txn_ts < COALESCE(s.effective_to, TIMESTAMP '9999-12-31 00:00:00')
), match_counts AS (
  SELECT
    txn_id,
    COUNT(status) AS matched_status_rows
  FROM status_matches
  GROUP BY txn_id
)
SELECT
  m.txn_id,
  m.account_id,
  m.txn_ts,
  m.amount,
  m.status,
  c.matched_status_rows
FROM status_matches m
INNER JOIN match_counts c
  ON c.txn_id = m.txn_id
WHERE c.matched_status_rows <> 1
ORDER BY m.txn_ts, m.txn_id, m.status;

SQL and data modeling dominate the distribution, but what makes Goldman's version uniquely painful is that both areas assume you already speak the language of Asset & Wealth Management. You won't design a positions schema in a vacuum; expect a hiring manager to press you on why your grain choice accounts for NAV calculation timing or how your SCD approach handles a desk reassignment mid-quarter. The biggest prep mistake is drilling SQL mechanics without learning the financial objects those queries touch (fund flows, trade lifecycle states, counterparty hierarchies), because Goldman's questions embed domain logic into the query itself, and clean syntax alone won't save you when you don't recognize what "as-of-date" means in a positions snapshot.

Practice Goldman-style questions across all six areas at datainterview.com/questions.

How to Prepare for Goldman Sachs Data Analyst Interviews

Goldman Sachs is pushing hard into the wealth channel. The firm oversees approximately $3.5 trillion in assets under supervision, posted $59.4 billion in revenue with 15.2% year-over-year growth, and just debuted co-branded model portfolios with T. Rowe Price aimed at financial advisors. For a data analyst sitting inside Asset Management, that expansion means more fund products, more advisor-facing analytics, and more cross-entity data to reconcile.

Your "why Goldman" answer needs to reference something you can't copy-paste for JPMorgan. Point to the T. Rowe Price partnership and what it implies about data complexity when two firms' portfolio models need to be stitched together. Or mention that Goldman forecasts alternative investment demand will outstrip origination supply and talk about why tracking performance across illiquid, infrequently priced assets is a harder analytics problem than public equities.

Try a Real Interview Question

| ledger_txns |
|-------------|
| ledger_id | account_id | txn_date    | amount | status    |
|----------|------------|-------------|--------|-----------|
| 1        | A1         | 2024-01-02  | 100.00 | POSTED    |
| 2        | A1         | 2024-01-02  | -20.00 | POSTED    |
| 3        | A1         | 2024-01-02  | 50.00  | CANCELED  |
| 4        | A2         | 2024-01-02  | 200.00 | POSTED    |
| 5        | A2         | 2024-01-03  | -10.00 | POSTED    |

| bank_txns |
|----------|
| bank_id | account_id | txn_date    | amount | status    |
|--------|------------|-------------|--------|-----------|
| 10     | A1         | 2024-01-02  | 80.00  | SETTLED   |
| 11     | A1         | 2024-01-02  | 5.00   | SETTLED   |
| 12     | A2         | 2024-01-02  | 200.00 | SETTLED   |
| 13     | A2         | 2024-01-03  | -10.00 | REVERSED  |
| 14     | A3         | 2024-01-02  | 15.00  | SETTLED   |

WITH ledger_daily AS (
  SELECT
    account_id,
    txn_date,
    SUM(CAST(amount AS DECIMAL(18,2))) AS ledger_net
  FROM ledger_txns
  WHERE status <> 'CANCELED'
  GROUP BY account_id, txn_date
),
bank_daily AS (
  SELECT
    account_id,
    txn_date,
    SUM(CAST(amount AS DECIMAL(18,2))) AS bank_net
  FROM bank_txns
  WHERE status <> 'REVERSED'
  GROUP BY account_id, txn_date
),
keys AS (
  SELECT account_id, txn_date FROM ledger_daily
  UNION
  SELECT account_id, txn_date FROM bank_daily
)
SELECT
  k.account_id,
  k.txn_date,
  COALESCE(l.ledger_net, 0.00) AS ledger_net,
  COALESCE(b.bank_net, 0.00) AS bank_net,
  COALESCE(l.ledger_net, 0.00) - COALESCE(b.bank_net, 0.00) AS net_diff,
  CASE
    WHEN COALESCE(l.ledger_net, 0.00) - COALESCE(b.bank_net, 0.00) <> 0.00 THEN 1
    ELSE 0
  END AS break_flag
FROM keys k
LEFT JOIN ledger_daily l
  ON l.account_id = k.account_id
 AND l.txn_date = k.txn_date
LEFT JOIN bank_daily b
  ON b.account_id = k.account_id
 AND b.txn_date = k.txn_date
ORDER BY k.account_id, k.txn_date;

Financial data problems tend to involve time-series positions, multi-entity joins, and window functions for running calculations, so practicing on schemas that feel like fund or trade data will serve you better than generic exercises. Drill finance-flavored SQL at datainterview.com/coding.

Test Your Readiness

Quiz yourself on stats, finance terminology (NAV, AUM, tracking error), and case structuring at datainterview.com/questions so you know where your gaps are before round one.