Unpack business metrics to explain their evolution
summary
Provides a concrete method that decomposes KPI changes into inner and mix effects and quantifies contributions instead of handwavy statements: e.g. 2022 revenue up $52,500 split into two additive effects, or an average footprint change of +0.6 kg vs −0.69 kg by category. Implements an open‑source Python package called I can explain, ~400 lines, implemented via Ibis to SQL so it runs on SQL backends, produces additive tables and waterfall charts, and is already used at Carbonfact in production.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 1:25Analytics engineering challenges
- 2:28Metric trees basics
- 3:11Revenue decomposition with Airbnb
- 4:59Introducing I Can Explain
- 7:50Framework beyond revenue
- 8:54Average metric decomposition
- 12:27Decomposing funnel metrics
- 15:38Why method is simple
- 16:41Carbon Fact applications
- 17:22Community feedback
- 18:47Ibis and SQL implementation
- 20:55Open-sourcing package
- 21:58Future research directions
- 23:43Interactivity and anomaly detection
- 24:47Carbon Fact resources
- 25:30Q&A
- 26:12Linear models and simplicity
- 29:26Standardizing metrics
- 31:55Theoretical guarantees
- 34:26Closing remarks
related talks
01h 06mBig Ways Small Businesses are Leveraging Analytics to Gain Insights
25mStorytelling with Data
07mPortfolio Marketing Metrics and Insights
01h 10mStorytelling through data
51mThe ROI of Business Analytics: Calculating Financial Return
18mProduct Usage Metrics - Practical Lessons from the Trenches
38mRip the Data
28mMeasuring the impact of an IDP