Whatnot — Product Analytics — Product Case Questions

Role context: Data Scientist, Product Analytics (live commerce marketplace) · Est. study time: 60 min · 5 questions

How to approach product cases here

Every case is one move in the same chain: understand the goal, break it into a data problem, choose the metric or method, name the bias and the trade-off, land on a decision and an action.

Five things make Whatnot cases distinctive, and they all come from the same root: live auctions with ephemeral inventory.

1. GMV can rise from price or from volume, and they mean opposite things. Routing more viewers to a show with fixed inventory raises the clearing price. That is a transfer from buyers to sellers, not created value. Normal e-commerce has no equivalent, because more traffic means more units sold.

2. The feed is 65% of GMV, so it is the demand allocation mechanism. A ranking change is not a UX decision. With 1 in 8 sellers depending on Whatnot for primary income, it is somebody's rent.

3. The objective is dual: watch or purchase. A buyer who watches 40 minutes and buys nothing is the audience that makes the auctions clear high. Optimizing purchases alone empties the rooms.

4. Interference is structural. Buyer-randomized tests overstate the launch, because treated buyers take finite attention and inventory from control buyers. This is the single most important stats fact in the role.

5. The ranker starves the seller pipeline by construction. It predicts purchase likelihood, established sellers have the best history, so demand concentrates on them. That is the model working correctly and killing next year's supply.

The traps: reading a GMV lift as created value, treating a long non-buying watch as a funnel leak, trusting a buyer-randomized marketplace test, and optimizing one launch at a time while the feed drifts.

The five below weight diagnosis, measurement design, two launch decisions (one on the feed, one on the pipeline), and incrementality. If you only have time for two, do bc1 and bc3.