Layout
NORTHSTAR LAB · JOURNEY 02

Experimentation,
from change to evidence.

A faster store. But is it a better store?

Northstar is a fictional Amazon-like marketplace. Its engineering team wants to ship a new product-page cache. Will it make shopping faster, help buyers complete purchases, and keep checkout reliable?

Follow one A/B test from a causal question to a rollout decision. The running example uses sticky buyer assignment, seven-day purchase conversion, latency diagnostics, and reliability guardrails. All numbers are constructed for learning—not Amazon results.

24 stops · 0 marked as understood · no prerequisites

Start with why randomize

Start with a fair comparison. Measure buyer benefit and operational safety separately. End with an honest decision—not a search for a green p-value.