How concurrency is measured
Players report regularly while they are playing. Counting the sessions that reported in the same short interval gives the number of concurrent viewers. The figure is tracked over time and the peak is recorded for each day, event or month.
Concurrency for on-demand and live
On-demand viewing is spread across the day, so peak concurrency is a small share of the audience. As a rough guide, a few percent of monthly active users watch at the busiest moment of an ordinary evening.
Live is different. Everyone arrives for the same start time, so a large share of ticket holders or subscribers may be watching at once, and most of them join within a few minutes. Plan capacity for a live event from expected attendance, not from on-demand history.
Why it matters
For example, 5,000 concurrent viewers at an average of 4 Mbps is 20 Gbps of delivery at that moment.
- Capacity: sign-in, entitlement and DRM license systems must absorb the peak
- Bandwidth: peak concurrency multiplied by average bitrate is the peak delivery rate
- Cost and plan limits: platforms commonly set a concurrent-stream allowance per tier
- Quality: buffering and failed starts appear first at the peak
Last reviewed