First, check the units

Before anything else: divide your plan by 8. A 500 Mbps line that downloads at “55” is delivering 55 MB/s = 440 Mbps — that’s a healthy connection, not a broken one. Half of all “slow internet” reports end right here, at the bits-vs-bytes confusion. If your numbers survive that check — 500 Mbps plan, 55 megabits of actual throughput — read on.

The taxes, in the order they hit

1. Protocol overhead — a fixed ~3–8%. Every transfer is chopped into packets, and each packet carries headers: Ethernet, IP, TCP, plus TLS record framing for HTTPS. Retransmissions and TCP’s congestion control add a little more. On a clean wired link this costs 5–10%, which is why the calculator on this site defaults to 90% efficiency — that’s what “normal” looks like.

2. Wi-Fi physics — the biggest variable. Wired Ethernet is a dedicated lane; Wi-Fi is a shared, half-duplex conversation over contested airtime. Speed collapses with distance, walls, neighboring networks on the same channel, and the band: 2.4 GHz tops out around 40–100 Mbps in real homes while 5/6 GHz and Wi-Fi 6/6E/7 hold far more. A gigabit plan measured over congested 2.4 GHz Wi-Fi will “test” at a fraction of the plan and there’s nothing wrong with the line.

3. The server on the other end. Your ISP only controls the path to its edge. The far server’s rate limits, its own congestion, and the peering between networks decide whether that download saturates your link. This is why Steam or a big CDN fills a gigabit pipe while a small software site drips at 8 MB/s — the bottleneck moved upstream.

4. Peak-hour congestion. Shared access technology — cable DOCSIS especially, mobile networks always — slows when the neighborhood streams at 9 PM. “Up to” speeds are quoted for the best case; DOCSIS nodes and cell towers divide capacity among active users.

5. Your own stack. A VPN hop, an antivirus web filter, a powerline adapter under a fridge’s interference, a Cat5 cable negotiated to 100 Mbps, a 10/100 router port — each is a quiet 90% cut hiding inside the house.

How to measure honestly

1. Wired, direct to the router, gigabit port, Cat5e or better — eliminates taxes 2 and half of 5.
2. Close background traffic — cloud sync, streams, other devices.
3. Test twice: a public speed test (Mbps) for the plan check, then a real large-file download from a fast CDN (MB/s) for the practical check. The calculator reverse panel converts that timed download into your achieved Mbps.
4. Repeat at a different hour before concluding — one bad window isn’t a bad plan.

What to do with the number you get

• Within ~90% of the plan on a wired test? You’re getting what you bought; any slowness lives elsewhere.
• Wired is fine but Wi-Fi collapses? The fix is placement, band steering, a wired AP or mesh node — not a bigger plan.
• Speed tests fine but one service crawls? Their limit, not yours — try the same file from a faster mirror.
• Wired tests consistently under ~70% of plan, multiple days? Document them and escalate to the ISP — that’s a genuine under-delivery.

The companion piece Mbps vs MB/s covers the unit side, and download-time math turns the speed you actually get into real transfer times.

Frequently asked questions

I pay for 500 Mbps — why do downloads show ~55 MB/s?

That's almost exactly right: 500 Mbps = 62.5 MB/s theoretical, and ~56 MB/s after normal protocol overhead. If you're seeing 5–6 MB/s instead, check the units — that's a tenfold gap pointing to a 50 Mbps bottleneck somewhere: old Ethernet cable, 2.4 GHz Wi-Fi, or a limited server.

How much speed does Wi-Fi actually cost?

Depends on the band and distance. A close-range Wi-Fi 6 client can sustain 80–95% of a gigabit; 2.4 GHz through two walls may hold only 20–40 Mbps total. Wi-Fi is also half-duplex — one device talks at a time — so a busy household divides the airtime before your download even starts.

Why is my speed test fast but my actual download slow?

Speed tests hit a nearby, uncapped server chosen for the measurement; real downloads depend on the far server's own bandwidth and peering. Game platforms and OS mirrors usually saturate your link — if Steam is slow but the test is fast, the limit is upstream of your house.

Does a VPN slow downloads?

Yes — encryption overhead plus the VPN server's own capacity, typically 10–40% on a decent provider and far worse on crowded free ones. Test with the VPN off before blaming the plan.

What does 'up to 500 Mbps' legally mean?

In most markets, the advertised figure is the maximum the technology can deliver at the access link — not a guaranteed minimum. Regulators increasingly require a 'typical' or minimum speed in contracts; check yours, and document wired tests at off-peak hours if you're disputing it.