Investigate Instacart Revenue Decline Using Weekly Data
Overview
This question evaluates a Data Scientist's skills in time-series analysis, anomaly detection, changepoint testing, decomposition and attribution, and segmentation within the Analytics & Experimentation domain, emphasizing work with constrained weekly data and with richer order-level signals.
Instacart
Jul 12, 2025, 6:59 PM
Data Scientist
Technical Screen
Analytics & Experimentation
18
0
Scenario
You are the on-call Data Scientist for Instacart. This week’s total revenue is down 4% versus the prior week. Initially, you only have access to the historical weekly revenue time series (no sub-weekly, no segmentation).
Task
- Part A: With only weekly revenue data, outline how you would investigate the 4% week-over-week decline. Be explicit about how you would assess whether this is expected (seasonality/holidays) vs. anomalous (trend break/changepoint), and how you’d quantify the expected range.
- Part B: If you later gain access to richer data (orders, AOV, geography, cohorts, etc.), describe the additional drill-downs and attribution analyses you would run to identify root causes.
Constraints and Hints
- Data constraint (Part A): weekly revenue time series only.
- Use appropriate time-series tools: seasonality and holiday effects, YoY/seasonal comparisons, changepoint tests, anomaly detection against forecasts, and uncertainty intervals.
- With richer data (Part B): perform decomposition (e.g., Orders × AOV × Take Rate), segmentation (geography, retailer, category, cohort), and drill-downs.
- State minimal assumptions when needed and describe decision criteria for whether −4% is noteworthy.
Solution
Login required
Show