"Up to 1 Gbps." You've seen the billboard. Maybe it convinced you to sign up. The question nobody on that billboard answers is: up to 1 Gbps under what conditions? And what's the realistic number you'll actually see at 8pm on a Tuesday when your neighbours are all streaming?
We have data. Not focus groups, not industry reports, not press releases — actual speed test measurements. Here is what they show.
The "up to" problem
ISPs in most markets are legally allowed to advertise "up to" speeds — meaning the theoretical maximum under ideal single-user conditions. This is a bit like advertising a car as capable of "up to 180 mph" when the speed limit is 70 and you live in a city. The number is technically accurate and practically useless.
The FCC in the United States — and Ofcom in the UK — have both published broadband performance studies showing persistent and significant gaps between advertised and delivered speeds. The FCC's Measuring Broadband America programme has documented this gap for over a decade. It hasn't closed.
What the gap looks like in practice
Here is a simplified summary of what we observe from our test data, grouped by connection technology. The "delivery ratio" is measured speed divided by advertised speed at plan signup time.
| Connection Type | Typical Plan Speed | Median Measured (Off-Peak) | Median Measured (Peak Hours) | Avg. Delivery Ratio |
|---|---|---|---|---|
| FTTH Fibre (symmetric) | 1 Gbps | 940 Mbps | 910 Mbps | 94% |
| Cable (DOCSIS 3.1) | 600 Mbps | 520 Mbps | 310 Mbps | 72% |
| Cable (DOCSIS 3.0) | 200 Mbps | 175 Mbps | 90 Mbps | 66% |
| DSL / VDSL2 | 100 Mbps | 75 Mbps | 65 Mbps | 70% |
| Fixed Wireless 5G | 300 Mbps | 220 Mbps | 130 Mbps | 58% |
The pattern is consistent: fibre connections deliver close to their advertised speed regardless of time of day. Cable and wireless connections — which share infrastructure between many users — show substantial peak-hour degradation. The difference between off-peak and peak on cable can be 40–50% of the advertised speed.
Latency tells a different story than throughput
Speed (throughput) is only half of the connection quality picture. Latency — the reaction time — matters at least as much for interactive applications.
This is significant because a 30ms latency baseline is already at the edge of what competitive gamers tolerate. Add bufferbloat on top — which affects the majority of cable connections — and you're looking at 100ms+ effective latency during evening hours on what was marketed as a fast, low-latency plan.
Fibre connections — specifically FTTH where the fibre runs all the way to your router, not just to the street cabinet — maintain low latency under load far more consistently. The shared coaxial plant on cable and the radio spectrum on fixed wireless are both contention-based, which means your neighbours' usage directly affects your latency.
Why peak hours hit cable so hard
Cable internet operates on a shared medium — your DOCSIS node serves anywhere from 50 to 500 households depending on the ISP's infrastructure investment. During peak hours (typically 7–10pm), all those households are simultaneously streaming, downloading, and gaming.
The CMTS (Cable Modem Termination System) at your ISP's headend has to schedule upstream and downstream bandwidth across every device on that node. When the node is congested, it's doing its best to share fairly — but "fairly" in this context means everyone gets less, including you.
ISPs respond to node congestion by splitting nodes — adding more equipment to serve fewer homes per node. This is expensive. Whether they do it proactively or only after customers start complaining depends on the ISP. Our ISP Performance Index tracks this over time using aggregate test data.
What to do with this information
First, measure your own connection at different times of day. A single speed test at 2pm on a weekday tells you almost nothing useful. Run tests at 8pm, at midnight, and on a weekend afternoon — then compare. If your evening speeds are consistently half your off-peak speeds, you have a congestion problem and a reasonable basis to contact your ISP about node upgrades or a plan change.
Second, check your bufferbloat score. The peak-hour latency numbers above get significantly worse when bufferbloat is factored in. A cable connection with poor SQM settings on a congested node can reach 300–400ms effective latency during prime time, which makes online gaming essentially unplayable.
Third, if you're shopping for a new ISP, weight fibre options heavily. The delivery ratio data is not subtle: FTTH consistently outperforms cable on both throughput delivery and latency consistency. If fibre is available in your area and you're still on cable, the performance difference justifies seriously evaluating a switch.
A note on our ISP index
Our Global ISP Performance Index tracks aggregate test results by ISP and region over rolling 90-day windows. It is not a ranking of which ISP is "best" globally — local network conditions vary too much for that to be meaningful. It is a transparency tool: you can see how your ISP's measured performance compares to what they advertise, and how it trends over time.