Blog

Lyra — Experimentation You Can Trust

Causal Inference

Experimentation

I built an experimentation platform where every estimator is certified against a known ground truth — because on real data, you can never check whether your A/B test is…

Two Pricing RCTs, Zero Real Customers

Causal Inference

Experimentation

Two questions every SaaS team argues about: can we raise the price, and should we push annual billing? Instead of guessing — or risking real revenue — I ran both as…

The Balcony Paradox: Causal Renovation Effects in Berlin Rents

Causal Inference

Machine Learning

Spatial Analysis

In Berlin’s rental data, apartments with a balcony rent for slightly less per square meter than apartments without one. Taken literally, that says rip the balcony off to…

When Synthetic Consumers Are Too Smart: A Perfect-Separation Post-Mortem

Discrete Choice

Conjoint Analysis

What if you could run a pricing survey without recruiting a single human — by having an LLM role-play hundreds of buyers? That’s the idea behind BeeSignal. But the first big…

Five Ways to Improve Your Event Study Plots

Causal Inference

Difference-in-Differences

The standard event study plot — point estimates with 95% pointwise confidence intervals — is ubiquitous in applied economics. But it leaves a surprising amount of…

Interpreting Event Studies Across Modern DiD Methods

Causal Inference

Difference-in-Differences

Modern event study plots are the go-to visualization for Difference-in-Differences analysis. But when you overlay results from multiple estimators — Dynamic TWFE…

Sensitivity Analysis for Parallel Trends

Causal Inference

Difference-in-Differences

Passing a pre-trends test does not mean parallel trends holds — it may just mean your test has low power (Roth, 2022). And even a well-powered test cannot rule out…

DiD vs Synthetic Control vs Synthetic DiD

Causal Inference

Difference-in-Differences

Synthetic Control

A unified comparison of Difference-in-Differences, Synthetic Control, and Synthetic Difference-in-Differences — when each method works, when it fails, and why SynthDiD…

Causal Forests: Heterogeneous Treatment Effects

Causal Inference

Machine Learning

Heterogeneous Treatment Effects

Estimating heterogeneous treatment effects using Causal Forests and Meta-Learners in Python.

Causal ARIMA Approach to Estimate Price Policy Changes on Sales

Causal Inference

C-ARIMA to estimate the causal effects in time series settings where no control unit is available.

Bayesian Approach to A/B Testing

Bayesian Statistics

Causal Inference

Bayesian A/B testing Analysis using Python.

Merger Simulation of Broadband Internet Services

Competition & Antitrust

Binary Choice Models

Assessing merger competitive effects through merger simulation.

Retail Gasoline Merger in Chile: An Ex-Post Merger Evaluation

Causal Inference

Competition & Antitrust

Spatial Analysis

Gasoline Industry

Ex-post assessments of merger effects in Chilean’s retail gasoline market.

A/B Testing Analysis

Causal Inference

A/B testing Analysis using Python.

Nonparametric & Semiparametric Binary-Choice Regressions

Nonparametric Regression

Semiparametric Regression

Binary Choice Models

Kernel Density Estimation

Nonparametric Kernel Density, Nonparametric Regression and Binary-Choice Models.