How Back-of-the-Envelope Math Can Make or Break a System Design Interview

System design interviews often trip up candidates not because the math is difficult, but because they fail to state assumptions, derive meaningful conclusions, or let the numbers guide their architecture decisions. A structured approach moves from user counts to requests per second, storage needs, and bandwidth, using simple approximations like treating one day as 100,000 seconds. For example, a social feed with 100 million daily active users yields roughly 200 writes per second and 30,000 reads per second at peak, immediately revealing a 50:1 read-to-write ratio that shapes the entire design. Latency benchmarks — such as memory being roughly 1,000 times faster than SSD — complement throughput estimates by exposing what is physically feasible within a given response-time budget. The goal is not precision but identifying the binding constraint, whether read throughput, write throughput, storage, or bandwidth, so design choices follow logically from the numbers.