E-commerce & retail Data Scientist Interview Loop
E-commerce data science is closer to machine learning than in most industries. Forecasting, search ranking, recommendations, pricing and fraud are all ML problems, and data scientist interviews test for that. There's a recruiter call and a technical screen on SQL and Python. Then come four to six rounds: ML breadth, ML depth on your own work, model evaluation, a product or business case, and behavioral. The first step is where they differ. Walmart's is a timed assessment run by an outside company, and it stops you on a single failed test case. Shopify's first real round is the hour-long "Life Story" conversation, which decides more than candidates expect. eBay adds rounds on search, ranking and recommendations.
Companies behind this pattern
Practice by round
SQL and Python basics under time on orders, products, customers and sessions. That's joins, aggregations, filters and a top-N, then list and dictionary processing.
| Question | Difficulty |
|---|---|
| Customer Average Orders | Easy |
| Products Report Summary | Easy |
| Shipped by Speedy Express | Easy |
| Account Registrations | Easy |
| Lowest Priced Orders | Easy |
| Order Details | Easy |
| Calculate Samantha's and Lisa's total sales revenue | Easy |
| Total Cost Of Orders | Easy |
The hour with the highest order volume, the best-selling item, share of total spend, product market share, negative reviews in new locations, and revenue per employee. Then comes the Python data handling that happens before a model.
| Question | Difficulty |
|---|---|
| Salary Less Than Twice The Average | Medium |
| Employee with Most Orders | Medium |
| Most Sold in Germany | Medium |
| Process a Refund | Medium |
| Negative Reviews in New Locations | Hard |
| Cookbook Recipes | Hard |
| Sales Growth per Territory | Hard |
| Product Market Share | Hard |
The business or product case: a north-star metric for e-commerce, customer engagement and disengagement, how to grow the number of customers, and identifying eBay objects. Scope it, measure it and put a number on it. In the real loop, the depth round takes your own project apart. The projects here, like which customers to target, Black Friday purchases, laptop price prediction and a treadmill buyer profile, are practice for both.
| Case study | Difficulty |
|---|---|
| Graph of Sales and Revenue | Easy |
| McD's Big Macs | Medium |
| Identify Ebay Objects | Medium |
| Customer Engagement and Disengagement | Medium |
| North Star Metric for E-Commerce | Medium |
| Measuring Social Media Campaign Success | Medium |
| Data project | Difficulty |
|---|---|
| Sales Data Analysis | Easy |
| Beer Data Analysis | Easy |
| Which Customers to Target | Medium |
| Laptop Price Prediction | Hard |
ML breadth and evaluation: ridge versus lasso, bagged versus boosted models, random forests, the assumptions of linear regression, sorting new observations into categories, and purchase intent as a model. There's also conversion rate and the chance of a coupon being used.
| Question | Difficulty |
|---|---|
| Purchase Intent | Easy |
| Categorize New Observations | Easy |
| Rolling Two Dice | Easy |
| Conversion Rate | Easy |
| Random Forest | Medium |
| Bayesian Logistic Regression | Medium |
| Cost of Discount Coupons | Medium |
| Discount Coupons Usage | Medium |
| Two Coupons | Medium |
| Long Term Probability | Medium |
| P-Value, Power, and Sample Size in A/B Testing | Medium |
| Geo Experiment Across Regions | Medium |