About this blog

Stats that actually ship to production

Most statistics education stops at textbooks. This blog bridges the gap — translating causal inference, A/B testing methodology, and econometric theory into the kind of reasoning that holds up when a PM asks "is this experiment actually working?"

A data analyst who spent too long in academia

I'm a Senior Data Analyst at a e-commerce unicorn, working on experimentation, recommendation, and search analytics — close enough to product and engineering that the role often looks more like Product Data Scientist in practice. Before that, I finished a PhD in Economics, where I spent most of my time thinking about causal identification — instruments, difference-in-differences, and whether any of it would survive contact with real data.

The short answer: some of it does. A lot of it needs translation. That translation is what this blog is about.

Academia

PhD, Economics

Applied Microeconomics — causal inference, IV, DiD. Job market paper on social network formation.

Research

Research Institute

Applied econometrics on government policy.

Startup

Early-stage Startup

Applied A/B testing frameworks, surrogate index LTV models to estimate the impact early

Now

E-commerce Unicorn

Senior Data Analyst working like a Product Data Scientist — sequential ab testing, ad attribution vs. incrementality, and causal inference including synthetic difference in difference.

Business impact first, the right method always

These posts start from a real operational problem, not a method looking for an application. Teams want to end experiments faster — but peeking at results inflates false positives. They want protection from a bad launch — but waiting for a fixed window leaves guardrail metrics blind until the very end. Group Sequential Testing is a good example: it isn't here because it's mathematically interesting, it's here because it's the correct fix for exactly those two problems. Every post on this blog works the same way — start from the pain point, then bring in only as much methodology as it takes to resolve it safely.

Sequential A/B TestingGroup Sequential MethodsCausal InferenceDifference-in-DifferencesCUPED / Variance ReductionMultiple TestingUplift ModelingSurrogate IndexSynthetic Difference-in-DifferencesFeature Stores & FlagsLTV PredictionRetail Media & Ad Tech