The simple answer

The honest target is zero. A healthy connection should not be losing packets at all under normal conditions, and any sustained loss is technically a problem worth investigating. That said, in the real world tiny amounts of loss happen constantly without anyone noticing, and the practical question is when loss becomes bad enough to actually affect what you’re doing.

The rough rule that most network engineers work to is that anything under 1% is generally fine for everyday use, 1% to 2.5% starts to be noticeable on live services like calls and gaming, and anything above 5% makes real-time applications close to unusable. Browsing and downloads can tolerate much higher loss than calls because the protocol just resends the missing data, you only notice that things feel slow rather than broken.

How much loss different activities can handle

Web browsing, email, and file downloads are the most forgiving. These use TCP, which automatically resends any packet that goes missing, so the data always gets there in the end. You can often have 5% or even 10% loss and still load a web page successfully. It will just feel sluggish because everything is being requested twice.

Video streaming sits in the middle. Services like Netflix and YouTube buffer ahead of what you’re watching specifically to absorb short bursts of loss, so up to around 1% is usually invisible. Above 2% you start seeing quality drops as the player switches to a lower bitrate to cope, and above 5% you get buffering and stalls.

Voice and video calls are far less forgiving. VoIP and video conferencing use UDP, which doesn’t resend lost packets because by the time they’d arrive the conversation has moved on. Loss above 1% starts to cause noticeable audio glitches. Above 2% you get robotic voices, frozen video, and people asking “are you still there?”. Above 5% the call is effectively broken.

Online gaming is the strictest of all. Competitive players consider anything above 0.5% loss to be a real problem because even occasional dropped packets cause rubber-banding, missed shots, and disconnections. Esports environments are typically engineered to maintain loss well below 0.1%.

The context that matters

Brief spikes are different from sustained loss. A connection that runs at 0% loss but briefly hits 4% during a single ten-second window is normal. A connection sitting at a steady 2% all day is not, even though the average might look acceptable on paper. What hurts user experience is consistency of loss, not just the headline percentage.

Distance matters too. A test to a server in your own country should show essentially zero loss. A test to a server on the other side of the world passes through far more equipment and may legitimately show a small amount of loss without anything being wrong. That’s why network engineers compare like with like, testing the same route at different times rather than comparing two unrelated routes.