Why your speed test results vary (and what to do about it)

Speed isn't a single number — it's a distribution

Your “200 Mbps” broadband plan isn't a guarantee that you get 200 Mbps at every moment. It's a contract for an upper-bound under normal conditions. The actual throughput at any instant depends on a dozen separate factors, and almost all of them change minute to minute. So a speed test result is one sample from a noisy distribution — not the speed.

The seven things that move your number around

1. Time of day

Cable broadband shares one local segment between hundreds of homes. Between roughly 7–11 pm, that segment is saturated with streaming traffic, and your slice of it shrinks. Full fibre is much less prone to this because the contention ratio is far lower, but it isn't immune. A test at 3 pm on a Tuesday and a test at 9 pm on a Sunday will read very differently on the same line.

2. Wi-Fi conditions

Anything that changes between the router and your device changes the result: someone walking through the room, a microwave running, a neighbour's router on the same channel. Wi-Fi tests can swing 20–40% from one minute to the next without anything obvious changing.

For a stable reading, test over Ethernet. We have a full guide on Wi-Fi vs Ethernet with the specifics.

3. Background traffic

Your phone backing up photos, Steam updating a game, a smart-TV pre-loading a stream — all of these consume bandwidth in the background. The speed test only sees what's left. A laptop with a Dropbox sync running can hide 100 Mbps of your line without you noticing.

Before testing: close other tabs, pause sync clients, end video calls, and ideally pause any other devices on the network.

4. The server you hit

Different speed test tools route to different measurement servers, sometimes on different networks. A test that hits a server two miles away over a fast peering will read higher than a test that has to traverse a congested cross-network link to reach a server in another city.

Our test always measures against the closest Cloudflare edge for your location, which is generally well-peered with most ISPs. That doesn't mean every tool will agree — and that's normal.

5. The TCP slow-start window

The protocol behind almost all web traffic (TCP) starts cautiously and ramps up. Short tests don't give the protocol enough time to find the ceiling, so they read low. Long tests give the line enough air to settle in. We size each phase to be long enough to clear the ramp but short enough to finish in 8–12 seconds.

6. Your device

A speed test puts real load on your CPU and Wi-Fi chipset. An old phone, a hot laptop thermally throttled, or a browser tab with too many extensions can all bottleneck the test before the network does. On gigabit and multi-gigabit plans this is increasingly the limiting factor, not the line.

7. ISP-side dynamics

A handful of legitimately-ISP-controlled things can shift your speed without notice:

• Routing changes — an ISP rebalancing traffic between peering partners.
• Maintenance — a node being taken offline temporarily reduces local capacity.
• Traffic shaping — some ISPs prioritise certain traffic types in busy hours.
• Faults — an underground cable degrades, a card in an exchange fails, weather affects copper lines.

How to get an honest reading

The goal isn't a single magic number — it's a confident range. Here's the five-minute routine:

1. Plug into your router with Ethernet if you can. If not, sit in the same room and use 5 GHz.
2. Close other tabs and apps. Pause cloud sync and video calls.
3. Run the test three times in a row.
4. Note the median (the middle of the three) as your number for this moment.
5. Run it again at a different time of day — ideally one peak (evening) and one off-peak (afternoon).

When to escalate to your ISP

Variance is normal. A complaint is justified when you see a pattern, not a single bad reading. Specifically:

• Median wired speed is well below the plan's minimum guaranteed speed across multiple tests, multiple days. (In the UK, the ISP industry's Voluntary Code of Practice covers this; in the US, the FCC's broadband labels specify the figures.)
• Peak-time speed is below 50% of off-peak speed consistently.
• Latency or jitter regularly spikes above the guidance in our latency guide.
• You can reproduce the issue at the modem with Wi-Fi turned off entirely.

Screenshots of multiple tests, with dates and times, are how you turn a vague complaint into something an ISP support team will actually act on.

The mindset shift

Stop chasing a single peak number. The useful question isn't “what's my fastest reading?” — it's “what does my line do reliably, at the times I actually use it, on the devices I actually use?” A 300 Mbps plan that consistently delivers 280 at peak is a better plan than a 1 Gbps plan that drops to 90 every evening.