I started with monthly active users and worked down to concurrent listeners, which felt right.
Start by clarifying the scope (e.g., audio streaming only, global, peak vs. average) and then break the problem into a simple equation: total bandwidth = number of concurrent streams × average bitrate per stream. Use round numbers and state assumptions clearly, then sanity-check the result against known industry figures.
Pro tip: Anchor your estimate with a known data point, such as Spotify's ~600 million users or typical audio bitrate of 160 kbps, and explicitly call out that you're ignoring video/podcast traffic unless asked. This shows you can prioritize and communicate assumptions like a PM.
Ask whether to include all content types (music, podcasts, video) and whether to estimate peak or average bandwidth. State that you'll focus on audio streaming and use global figures.
Start with total users (e.g., 600M), estimate daily active users (e.g., 30% = 180M), and then estimate peak concurrency (e.g., 10% of DAU = 18M) or average concurrency (e.g., 5% = 9M).
Use a typical streaming bitrate for Spotify: 160 kbps for normal quality, or 96 kbps for low. Convert to bytes: 160 kbps = 20 KB/s.
Multiply concurrent streams by bitrate: 18M × 20 KB/s = 360 GB/s (or 2.88 Tbps). For average, use 9M × 20 KB/s = 180 GB/s (1.44 Tbps).
Compare with known figures (e.g., Netflix uses ~100 Tbps peak) and adjust for overhead, caching, and regional distribution. Acknowledge that actual bandwidth may be lower due to CDNs and caching.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.