Goldman Sachs Engineering — Data Analyst Case Interview (Global Markets TCA and Client Profitability)

This Goldman Sachs case simulates a fast-paced Global Markets analytics problem where you act as a Data Analyst embedded in Engineering, partnering with traders and sales to improve execution quality and client profitability while upholding risk and control standards. You will be given a simplified, de-identified trade and market data excerpt (equities and FX) and asked to: 1) clean and reconcile the data (detect nulls, time-zone drift, out-of-order timestamps, misjoined client IDs), 2) produce core SQL/Python queries for execution quality (arrival slippage bps vs arrival price, VWAP shortfall, implementation shortfall, venue fill analysis, cancel/replace loop rates), 3) compute profitability with a commercial lens (desk and client PnL after fees and explicit costs, revenue concentration, capacity/size effects, and outlier days), 4) flag anomalies and control breaks (suspicious mark-outs, unusually positive/negative slippage by venue or algorithm, missing market data for best-ex checks) and propose guardrails, and 5) craft a concise narrative recommending client and venue actions the desk can take next week. The interview mirrors Goldman’s culture: detail-first, hypothesis-driven, and execution-oriented. Expect rapid follow-ups challenging your assumptions, precision in window functions and time-series joins, and discussion of how you would productionize your approach (data lineage, auditability, SLAs, access control, PII handling). You may be asked to whiteboard query logic, explain metric choices (e.g., why median vs mean slippage; using P90 tails), and translate findings for non-technical stakeholders. Typical focus areas: • SQL proficiency (complex joins across trades, orders, quotes; partitioned window calcs; late-arriving data handling). • Python for quick EDA and validation (pandas time alignment, outlier detection, simple regression to explain slippage by volatility/liquidity). • Finance domain judgment (best execution/TCA concepts, liquidity buckets, venue/algorithm selection, client coverage implications). • Risk and controls (data quality checks, reconciliations to reference data, reproducibility, monitoring/alerts). • Communication under time pressure (executive-ready summary with clear next steps). Deliverables in-session: (a) query snippets or pseudocode, (b) 2–3 key tables/charts (e.g., slippage by venue and notional, client profitability waterfall), and (c) a 5-minute recommendation: which two clients to prioritize, which venues/algos to adjust, and what control to implement immediately.