Design Search & Discovery

This is the most product-authentic design in the loop, because every reviewer has used the product they are asking you to build. A guest sees a map. The listings on it have to match where they're looking, the dates they want, how many people are coming, and a stack of filters like price and amenities. They drag the map, and a fresh search fires. The whole thing has to feel instant, the results have to be ranked well, and a place that got booked five seconds ago must not be sitting there as if it were free.

Anchor the scale first. Airbnb has on the order of 7 to 8 million active listings worldwide, and search reads dwarf everything else. At peak, that's plausibly low tens of thousands of searches a second once you count every map drag and filter toggle. Against that, the searchable data per listing is tiny, a few kilobytes of attributes, so the entire index is small enough to keep hot in memory. Name that asymmetry early. This is a read-heavy, latency-sensitive system over a small but constantly changing corpus, and most of the design follows from that one sentence.

The Airbnb angle

Airbnb's search is a map, and that shapes the design. The map is the query. The primary input is a viewport, the rectangle of the world the guest is currently looking at, rather than a text box. They pan, they zoom, they flip on "Search as I move the map," and each gesture fires a new bounding-box query. So the geospatial read is the hot path that everything else hangs off, and a few hundred milliseconds of latency is the budget for the entire round trip. Frame your design around the guest staring at a map and tugging it around, because that is the real product and the interviewer knows it cold.