Three rules I keep coming back to when applying causal inference in product and business settings: start with the problem, skip causal inference when you can, and 80/20 the project itself.

A real-world RDD analysis on marketplace search rankings: practical setup, modelling choices, placebo testing, and lessons learned.

The core ideas behind RDD: how the continuity assumption works, why the running variable acts as an instrument, and what LATE really means.

Take a dive into the math and exemplifying use cases of the Poisson distribution.

How Marktplaats built Fisher — a Python A/B testing package — and how it scaled into Adevinta's group-wide stats engine.