- Data Aggregation & Complex Joins – Performance considerations when joining high-volume session logs with transaction tables, handling duplicate keys, and avoiding cartesian products.
- Data Cleaning in Python – Manipulating unstructured data structures, parsing JSON payloads, and converting string timestamps efficiently using
pandas or base Python.
- Advanced concepts (less common) – CTE optimization, partitioning strategy trade-offs, and building scalable tracking schemas for new tracking events.
Example questions or scenarios:
- "Write a SQL query using window functions to isolate each user's second purchase on the platform and calculate the time elapsed since their initial registration."
- "How would you optimize a slow-running join between a multi-billion-row search event table and a seller dimension table?"
Experimentation & Statistical Inference
Given eBay's constant feature testing across search, UI, and live streaming, statistical experimentation rigor is non-negotiable.
Be ready to go over:
- A/B Testing Frameworks – Sample size determination, power analysis, baseline conversion rates, and minimum detectable effect (MDE) setup.
- Statistical Significance – Calculating p-values, confidence intervals, t-tests, z-tests, and understanding distribution assumptions.
- Experimentation Pitfalls – Identifying and mitigating sample ratio mismatch (SRM), spillover/network effects (buyers seeing modified seller listings), and metric novelty effects.
- Advanced concepts (less common) – Variance reduction techniques (e.g., CUPED), quasi-experimentation designs (diff-in-diff) when randomized control groups are impossible, and multi-armed bandit testing.
Example questions or scenarios:
- "You launch an A/B test for seller tools. Control shows higher listing volume, but Treatment shows higher final conversion. How do you evaluate net platform impact?"
- "Walk through the mathematical steps to detect an SRM, and explain what operational issues usually cause it."
Product Sense & Business Analytics
This evaluation tests your structured thinking around e-commerce product health, operational anomalies, and multi-sided marketplace dynamics.
Be ready to go over:
- Metric Drop Diagnosis – Methodical root-cause analysis (segmenting by device, geography, seller tier, traffic source, acquisition channel) to isolate unexpected metric shifts.
- Product Metric Design – Establishing primary KPIs, secondary metrics, and guardrail metrics for new product initiatives like eBay Live.
- User Journey Analysis – Tracking multi-device user funnels, buyer retention loops, and seller onboarding conversion paths.
- Advanced concepts (less common) – Marketplace elasticity, cross-platform attribution modeling, and long-term seller churn prediction framing.
Example questions or scenarios:
- "Search refinement clicks went up 12% following a UI deployment, but overall checkout conversion dropped 1.5%. Is this launch successful? How would you investigate?"
- "Design a complete metric framework for evaluating the trust and safety health of a high-value vertical like luxury watches."
Machine Learning Modeling & System Design
For data scientists on ML-heavy teams, this module focuses on modeling fundamentals, algorithm design, and scalable deployment architectures.
Be ready to go over:
- Supervised Learning & System Design – End-to-end framework for search ranking, user classification, and risk/fraud detection systems.
- Model Evaluation Metrics – Selecting appropriate evaluation metrics (Precision, Recall, ROC-AUC, PR-AUC, F1-Score, NDCG) based on business cost functions.
- Recommender Systems – Collaborative filtering, content-based approaches, matrix factorization, and real-time candidate retrieval.
- Advanced concepts (less common) – Implementing clustering algorithms (e.g., k-means) from scratch, graph neural networks for fraud network detection, and LLM fine-tuning for listing knowledge graphs.
Example questions or scenarios:
- "Walk through how you would build a machine learning pipeline to improve a recommendation system for the eBay homepage, from raw log processing to online scoring."
- "How would you select models and metrics to detect counterfeit goods while ensuring good sellers face minimal false positive suspensions?"
Behavioral & Leadership Strategy
This area measures soft skills, alignment with eBay values, stakeholder management, and project execution capabilities.
Be ready to go over:
- Cross-Functional Influence – Navigating disagreements with Product, Engineering, or Executive leads using data-driven arguments.
- Ambiguity & Ownership – Taking vague business problems and turning them into technical projects with measurable outcomes.
- Project Post-Mortems – Articulating key learnings from project failures, analytical errors, or unexpected modeling bottlenecks.
Example questions or scenarios:
- "Describe a time you had to pivot a data science project mid-way due to changing business priorities or flawed underlying data."
- "Give an example of how you communicated a complex statistical model result to non-technical operational leads."
