Reading a growth launch as a supply problem, for Uber

The problem

A retail-goods marketplace launch, running on a courier network built and economically tuned for prepared food. The pitch was a growth story: get retail goods to a door using couriers who are already idle between food orders.

Walk into the analytics room and the real question was not how to drive demand. It was at what hourly incentive a Tuesday-afternoon courier accepts a heavier, lower-tipping retail trip instead of waiting for a better food order. That is a labour-supply question wearing a marketing costume.

What we did

Built and led an eight-analyst product analytics team across the launch: courier supply and incentive design, merchant onboarding and order frequency, named-partner reporting, and the experiment infrastructure underneath all of it.

Set the craft bar explicitly. Sample sizing done properly rather than by eye. Every incentive test carried a free-rider measure, which is the one most experimental designs skip: what share of these couriers would have shown up anyway. That is the question the finance team asks eventually, and the time to be ready for it is the week the programme launches.

The hardest problem was not analytical. A supply analyst is paid against courier acceptance and cost per incremental trip. A demand analyst is paid against merchant onboarding and order frequency. Those metrics fight each other. The team’s real job was keeping both analysts on the same row of the same spreadsheet every Monday.

The financial-products analytics group followed: debit, stored balance, gift cards, a card-network partnership, and the open-loop infrastructure that lets a courier’s earnings card work like a real bank card outside the app.

What happened

The structural finding is the deliverable, and it has held up. Marketplace launches are won and lost on the supply side. The demand side gets the press release. The supply side gets the late-night thread when Tuesday fill rate drops below what a merchant was promised.

The same mistake keeps repeating across the sector: importing a food-delivery elasticity curve into a retail launch and adjusting the intercept. Retail baskets are heavier, tips are lower per trip, and dwell time at pickup is longer. It is a different curve.

What it cost them to find out

Every week the two analysts sat in different reporting lines was a week the supply and demand tradeoff got argued in a leadership meeting where nobody had the per-region numbers in front of them. That is not a dashboard problem. It is an org-design problem wearing an analytics costume.

Courier elasticity, cohort decay, and the joint scorecard, on Jeff’s personal site.