The brief
M
Your Manager
“Experiment wrapped today. Raw event logs are in the shared drive. I've dropped the CSV in the folder. Conversion is up, I think, but I need you to tell me whether to ship before Monday's standup. The CEO wants a recommendation, not just p-values.”
◷ Monday standup
That last sentence is the whole job.
Your role: data analyst at Finch, a DTC footwear brand selling trainers, boots, and casual shoes across the UK and EU. Average order value around £112. The PM ran a 14-day A/B test on the checkout flow, a redesigned single-page checkout with Apple Pay, Google Pay, and trust badges above the CTA. 10,000 sessions split evenly between control and treatment.
The question isn't whether something changed. It's whether it changed enough, in the right places, to justify shipping to 100% of users.
The dataset
You'll be working with finch_ab_test.csv, a synthetic dataset designed to mirror the structure of real checkout event logs. 10,000 rows, one per user session, covering the 14-day experiment window (4–17 November 2024).
