Buffering bottlenecks: Understanding causes of latency

Let’s set the scene. You’re glued to the season’s most important match, be it a tennis fanatic fixating on the Wimbledon final or a football fan watching your country agonisingly take penalties at the culmination of the Euros, when you hear your neighbour’s triumphant cheers booming through the wall. Half a minute before the winning moment makes its way to your screen. In the meantime, your phone is blowing up with reactions and news notifications – all while you’re stuck watching the earlier play.

Data from the 2025 Super Bowl captures this frustration well. Tubi achieved the best performance when it came to the stream, lagging only 26 seconds behind real time, but most platforms were a lot worse – with many keeping viewers waiting over a minute.

The challenge is just as difficult for content creators themselves as it is for the end user. For example, trying to respond to message requests when there’s at times a minute’s delay between what you’re saying and your audience seeing it makes your interactive stream awkward – and building a connection with your viewers virtually impossible.

Latency is the formal word, and it’s a challenge that anyone involved in streaming, whether linear or over IP, has been wrestling with since the birth of live broadcast.

This delay experienced by viewers is known colloquially in the media-tech space as ‘glass to glass’, encompassing the entire journey of the video signal from the camera lens to the display.

Generally, the goal to achieve broadcast-grade latency in live video streaming sets the target of roughly five seconds or less. This benchmark originates from the performance characteristics of traditional television, where viewers have come to expect minimal delay between live action and its presentation.

We know that lowering latency is a key goal of media service providers, but what are the causes of latency? And how can we best combat them?

Capture and encoding

One of the root causes in delay build-up originates in the capture itself. Cameras, audio interfaces and capture cards often introduce delays due to internal buffering before outputting frames or audio. To overcome this, aim to use low-latency devices – in other words, choosing cameras and audio interfaces that guarantee ‘minimal latency’ or feature ‘real-time modes’. It’s also good practice to configure devices to output frames with the smallest buffer possible. Direct connections also reduce latency – so aim to avoid hubs or adapters that add extra delay.

Video compression is also computationally intensive. Software encoding adds an additional layer of latency, more so than hardware acceleration. Ways to overcome this include leveraging hardware encoders, using GPUs or dedicated streaming hardware. Optimising adaptive bitrate (ABR) can also alleviate issues. Only generate multiple renditions if necessary and encode in parallel rather than sequentially.

Packaging and segmenting

Protocols like HLS and DASH can also cause latency. These often divide video into small chunks (typically two to ten seconds each). Most players buffer multiple segments before playback to ensure smooth streaming, which can offer five to 30 seconds of latency. You can seek to address this through reducing segment lengths, as shorter segments decrease startup and live latency – although this can slightly increase overhead. On top of this, enabling low-latency protocols like LL-HLS or LL-DASH allows the player to start playback before full segments are available.

Additionally, muxing (or multiplexing) video into transport streams can also introduce further delay if the packaging process is not optimised. This can be avoided by configuring your packager to write segments as soon as possible, which minimises internal buffering. Also, it’s good practice to eliminate unnecessary conversions, by avoiding extra format changes or re-muxing steps in the pipeline to reduce cumulative delay.

Transporting and CDN distribution

One of the most common causes of latency is buffering at content delivery network (CDN) edges. CDNs store video segments closer to end users, but the process of caching and distributing these segments can introduce short propagation delays before playback begins.

Another key factor is geographic distance. The farther data has to travel, the longer it takes to complete a round trip between the user’s device and the server. These long-haul network hops increase round-trip time (RTT), which can slow the start of playback or cause interruptions during viewing.

Finally, congestion control plays a role. The Transmission Control Protocol (TCP) is designed to ensure reliable delivery of data packets, but if packets are lost or need to be retransmitted, the recovery process can create stalls in video playback.

Closing the gap

In the end, the causes of latency are not the result of one single bottleneck but rather the sum of many small delays that stack up along the streaming chain – from the camera lens all the way to the viewer’s screen. Capture devices, encoding, packaging, CDN distribution, player buffers and adaptive bitrate algorithms all contribute a few seconds here and there. Collectively, this can leave fans half a minute behind the action or creators struggling to engage with their audiences in real time.

The good news is that progress is being made. Hardware encoders, shorter segment lengths, low-latency protocols and smarter CDN strategies are already pushing performance closer to broadcast standards. Low-latency technologies like LL-HLS, CMAF DASH and WebRTC show what’s possible, offering delays of just a few seconds instead of nearly a minute. Still, these solutions come with trade-offs in complexity and scalability, meaning there is no one-size-fits-all approach.

For sports broadcasters, shaving seconds off the stream can be the difference between fans celebrating in sync with the stadium or hearing their neighbours spoil the winning moment. For creators, lower latency can mean transforming an awkwardly delayed live chat into a genuine, interactive conversation.

Ultimately, the industry’s challenge is to strike the right balance between quality, stability, scalability and speed. The future of streaming won’t be defined by eliminating latency entirely, but by minimising it to the point where it no longer breaks immersion. As viewer expectations continue to rise, the pressure will only grow for platforms to close the gap between ‘glass’ and ‘glass’ to deliver live content that truly feels live.

Watch our first FEED Forum: Lessons in latency here!