Experimentation & A/B Testing for Analysts

Experimentation & A/B Testing for Analysts

Experimentation & A/B Testing for Analysts is a structured track of 12 lessons that build a complete, interview-ready understanding of the topic. Work through them in order, then use the quiz and flashcards in each lesson to revise.

What this course covers

  • Why Experiments Beat Observational Analysis - What randomisation buys you that no amount of controlling for variables can.
  • Designing an Experiment: Hypothesis, Unit & Success Metric - Writing a testable hypothesis and choosing the randomisation unit correctly.
  • Randomisation, Control Groups & Assignment Mistakes - How assignment goes wrong in practice, and the sanity checks that catch it.
  • Sample Size, Minimum Detectable Effect & Power - Sizing a test before launch, and what to do when the traffic is not there.
  • Guardrail Metrics and Protecting Against Harm - The metrics that must not move, and how to set an automatic stop.
  • Reading Test Results: Significance, Effect Size & Confidence - Interpreting a result set honestly, including a flat result that is still informative.
  • Peeking, Early Stopping & Sequential Testing - Why checking daily inflates false positives, and the methods that permit it.
  • Common Experiment Failures: Novelty, Contamination & Seasonality - The five ways real tests break, each with the diagnostic that reveals it.
  • When You Cannot Randomise: Quasi-Experiments & Difference-in-Differences - Getting a defensible causal read without a clean randomised test.
  • Multi-Armed Bandits and Continuous Optimisation - When adaptive allocation beats a fixed split, and what you give up for it.
  • Building an Experimentation Culture and a Test Backlog - Prioritising tests, documenting results and stopping the same test being rerun.
  • Case Study: A Full Experiment from Hypothesis to Ship Decision - One experiment run end to end, including the numbers and the final call.