Packet Loss
Packet loss happens when pieces of data travelling across a network fail to reach their destination.
Since internet data is broken into small packets for transmission, losing even a small percentage of them can cause noticeable problems: video calls freeze, web pages load partially, and online games stutter. It is one of the most common causes of poor network quality.
Packet Loss, explained at three levels
At home
When data travels across the internet, it is broken into small pieces called packets. Normally, all the packets arrive and are reassembled into the complete data. Packet loss occurs when some of those pieces do not make it: they get dropped along the way. This can make video calls glitchy, web pages slow to load, or downloads fail. A small amount of loss is normal, but too much causes real problems.
At work
Packet loss is a critical network quality metric that directly impacts employee productivity and customer experience. Even modest packet loss (1-2%) can make VoIP calls sound choppy, cause video conferencing to freeze, and make cloud applications feel unresponsive. When users report intermittent connectivity issues that are not explained by speed tests, packet loss is often the underlying cause. IT teams monitor packet loss continuously on business-critical network paths.
At home
Packet loss occurs at several points: congested network links where routers drop packets because their buffers are full, faulty cables or connectors causing data corruption (corrupted packets are discarded), Wi-Fi interference causing failed transmissions, and overloaded network equipment. TCP (used for web browsing, email, and file transfers) automatically detects and retransmits lost packets, adding delay but ensuring data integrity. UDP (used for voice, video, and gaming) does not retransmit, so lost packets result in visible or audible glitches.
At work
Diagnosing packet loss requires testing at multiple points in the network path. Tools like MTR (My Traceroute) show loss at each hop between source and destination, revealing whether the issue is in the local network, the ISP, or further along the path. Common enterprise remediation includes upgrading congested links, replacing faulty cabling or hardware, improving Wi-Fi coverage and channel management, and implementing Quality of Service (QoS) policies to prioritize latency-sensitive traffic over bulk transfers during congestion.
At home
At the transport protocol level, TCP's packet loss recovery mechanisms include fast retransmit (retransmitting after three duplicate acknowledgements rather than waiting for a timeout), selective acknowledgement (SACK, which identifies exactly which packets were lost rather than retransmitting everything from the loss point), and congestion control algorithms (Cubic, BBR) that adjust sending rate based on detected loss and delay. For UDP-based real-time applications, Forward Error Correction (FEC) adds redundant data that allows the receiver to reconstruct lost packets without retransmission, trading bandwidth for resilience.
At work
In large-scale networks, persistent packet loss typically indicates either a capacity or a hardware problem. Buffer bloat, excessive buffering in network equipment that causes high latency and eventual packet drops, can be mitigated through active queue management (AQM) algorithms like CoDel or PIE. Network telemetry platforms that correlate interface error counters, buffer utilization metrics, and traffic flow data enable proactive identification of loss-prone paths before users are affected. For mission-critical applications, redundant network paths with automatic failover and application-layer retransmission provide defence in depth against packet loss impacts.
Real-world examples
At home
When Lena is on a video call with her parents and the picture keeps freezing and their voices cut out, the most likely cause is packet loss on her network. Some of the data packets carrying the video and audio are getting dropped, so her device cannot display a smooth picture or play continuous sound. Moving closer to her Wi-Fi router often helps by reducing wireless packet loss.
At work
When the sales team at Jerome's company reports that their VoIP phone calls have been plagued by choppy audio and dropped words, the network team runs diagnostics and finds 3% packet loss on the office WAN link during peak hours. They upgrade the link and implement QoS policies that prioritize voice traffic, and the call quality problems disappear.
Common confusions
- Some people are convinced that packet loss means your internet is broken. A small amount of loss is normal on any network; it only becomes problematic when it exceeds about 1%.
- It's a popular misconception that faster internet will fix packet loss. Loss is caused by congestion, faulty equipment, or interference, and adding bandwidth without fixing the root cause does not help.
- Many people assume packet loss only affects large downloads. Its most visible impact is on real-time communication like video calls and VoIP, where lost data cannot be recovered in time.
For analysts
If your project involves real-time communication (video conferencing, VoIP, live data feeds), specify acceptable packet loss thresholds in your requirements (typically less than 0.5% for voice). Do not assume the network can handle it without checking. Work with the network team to confirm that the infrastructure can meet these requirements and that QoS is in place to protect real-time traffic during periods of congestion.
History & origin
Packet loss has been an inherent challenge since the earliest packet-switched networks in the 1970s. The TCP protocol, designed in the early 1980s, included mechanisms to detect and retransmit lost packets specifically because loss was expected on the diverse and unreliable networks of the early internet. As networks improved, acceptable packet loss thresholds dropped, and today even 1% loss can significantly degrade the experience for real-time applications.