Stripe — Experimental Projects — Product Case Questions

Role context: Data Scientist, Experimental Projects (zero-to-one product incubation) · Est. study time: 60 min · 5 questions

How to approach product cases here

Every case is the same chain: understand the business goal, turn it into a data problem, pick the metric or method, name the bias and the trade-off, then land on a decision. On this team the decision is almost always one of three: invest more, iterate, or stop.

Three facts about early Stripe products sit under nearly every case:

• The units are businesses, and there are few of them. A pilot has tens of businesses, volume is dominated by a few, and buyer sessions are clustered inside those businesses. Intervals will be wide. Your job is to decide anyway, and say how sure you are.
• Adoption takes engineering work. A business has to integrate before any outcome exists. The funnel has a sandbox stage that consumer products don't, and that's where most pilots stall.
• Money moves, so risk lags. Disputes and fraud arrive weeks after the payment. Every "it's working" needs a "and it isn't losing money" next to it.

The metrics that matter here: retained live businesses (still running real volume at week 8), go-live rate and time to first live transaction, checkout conversion for buyer-facing bets, incremental versus moved volume, and the loss rate.

Traps specific to this domain:

• Treating stated interest (a waitlist, a friendly design partner) as demand.
• Reading a hand-picked, hand-held pilot as if it predicted self-serve adoption.
• Killing a bet on a non-significant result from 18 businesses, or scaling one on a single large customer's volume.
• Counting volume that moved from another Stripe product as growth.

This role weights Measure success and Launch or not, because defining success and making the invest-or-kill call is the core of the job. It adds one sizing question (Forecasting), since every bet starts with "is this worth building?", one Diagnose, and one Measure impact, because new features are usually opt-in and opt-in adoption is never random.