Method
Root-cause analysis
Explaining metric changes by decomposition and slice discovery.
11 cases from 10 companies apply Root-cause analysis. 5 report results in production.
Typical tasks: Explain a metric change, Answer questions over data, Detect anomalies, Build a data platform and 3 more. Companies: Pinterest, LY Corporation, Grab, Meta, Celsus and 5 more. Often combined with: ReAct / single-agent loop, LLM-as-a-judge, Golden set / benchmark, Data drift monitoring, Tool / function calling and 21 more.
How much should we delegate to an LLM? Automating advertising analytics reports with an AI agent![]() | ![]() | adtech | 2026-10 |
| How AI is transforming analytics at Grab | food-delivery | 2026-08 | |
| Building an Explainable AI Research Agent With Databricks Genie | retail | 2026-06 | |
Your BI Copilot Is Blind - Crafting an Agent That Goes Beyond the Dashboard![]() | ![]() | general | 2026-06 |
| DrP: Meta’s Root Cause Analysis Platform at Scale | social | 2025-12 | |
| The Quest to Understand Metric Movements | social | 2025-02 | |
How ML Product Development for Manufacturing Works at EuroChem![]() | ![]() | manufacturing | 2024-05 |
Incidents in ML Services and How to Monitor Them![]() | ![]() | healthtech | 2024-05 |
| Handling Online-Offline Discrepancy in Pinterest Ads Ranking System | adtech | 2024-01 | |
| Ariadne: building a custom observability UI for personalized search | e-commerce | 2023-06 | |
| Multi-dispatch event logging | food-delivery | 2022-01 |



