Design Reviews & Ratings

This is one of the most concretely reported DoorDash design prompts, and it is a clean one because it sidesteps the dispatch and ETA machinery that every Uber-style answer drowns in. The shape is simple to state. A delivery finishes. The consumer gets asked to rate it, and on DoorDash that means two separate ratings, one for the restaurant and one for the Dasher who brought the food. Those ratings roll up into the single number you see on every store card and every Dasher profile. It has to load instantly while a hungry consumer scrolls a feed of fifty restaurants.

Anchor the scale before drawing anything. DoorDash runs on the order of two billion orders a year, so call it roughly six million completed deliveries a day. Plenty of consumers skip it, yet each delivery can still produce up to two ratings, so the write volume lands in the low single-digit millions of ratings a day. Spread over a day that is only tens of writes a second, peaking maybe five times that during the dinner rush. Reads are a different planet. Every app open paints a feed where each store card carries a rating, so a single browsing session reads dozens of aggregates. Across millions of daily sessions, that is on the order of a thousand reads for every write. Name that asymmetry first. This is a write-light, read-heavy system, and almost every choice in the design follows from that one sentence.

The DoorDash angle

DoorDash is a three-sided marketplace, so a single "rate your order" tap is really two ratings going to two different places. The consumer rates the merchant, which feeds how that restaurant ranks and shows in the store feed, and separately rates the Dasher, which feeds something far more consequential than a display star. Frame the design around that split from the first minute. One submission flow, two rating entities, two aggregates with two different jobs.