S07 · interview simulation

System design: catching wrong fares before they sell

People, data and outputs in this scenario are simulated. Each stage is recorded when you commit; the reveal comes after.

5stages
40 minabout
0stages committed

The brief

A system-design interview (simulated). A synthetic airline's pricing system publishes about 2,000 fare changes a day across its network. Occasionally a fare is wrong: a decimal slip, a currency mismatch, or a value driven by a stale input. A wrong low fare can sell many seats within minutes. The interviewer (simulated) says: "Design a system that catches wrong fares before they sell."

Facts given (all hypothetical): about 0.5% of changes are wrong (about 10 a day); three analysts can review about 30 alerts a day between them; the pricing system can place a fare on hold through an API; the team is two data scientists and one engineer. No labelled history exists beyond analysts' override notes.

Stage 1 of 5

You have a few minutes to ask before designing.

Which questions decide the design? Select all that apply.

How fast a wrong fare sells and therefore how quickly an alert must reach someone, or act automatically.What a missed wrong fare costs against what a false alert costs, and how that differs by route and fare size.What actions exist (hold, revert, notify) and who is allowed to trigger them automatically.What labelled history exists and how it was produced, since analysts only noted what they happened to see.Whether the anomaly model should be an autoencoder, so that new kinds of error are covered without writing any rules.