
AI Analyst Lab
AI Analytics by Sravya, Shane, and Hai
The Root-Cause and Experiment Playbook (37 pages). The flagship. All three parts end to end, from a red arrow in Slack to a causal claim someone can bet a roadmap on.
Why Did the Number Move? (14 pages). Root cause decomposition, the funnel debugging map, segment-first thinking, and the mix-shift trap that fools almost everyone.
Run the Experiment (14 pages). Testable hypotheses, power intuition without formulas, the result interpretation tree, mixed results, and the one-page experiment brief.
Prove Impact Without an A/B Test (13 pages). The correlation trap, input-to-outcome maps, and three practical methods: before-versus-after, compare-the-changes, and find-the-twins.
Split a moved number into its parts until one cause is left standing, instead of running a meeting on competing hunches.
Catch the mix-shift trap, where an average moves even though nothing about the product actually changed.
Write a hypothesis specific enough to be proven wrong, and tell whether a test is even big enough to detect what you are hoping for.
Read a result honestly, including the ugly case where the primary metric wins and a guardrail quietly gets worse.
Make a credible causal claim when you never got to randomize, and say out loud how much to trust it.
Free
Four free guides on why a number moved and whether your change actually worked. 78 pages. No SQL, no Python, no tools.