Not getting the speed you pay for? How to prove it

The four numbers people confuse

Speed complaints go nowhere because the customer and the provider are arguing about different numbers, and only one of them is written down anywhere.

The tier name — “500 Mbps,” “Gig” — is a product name and an up-to figure in the advertising. Nothing obliges a provider to deliver it to a specific device at a specific hour.

The label figure is different, and this is the one nobody tells you about. Since the FCC’s 2022 broadband label order, providers must display, for each plan, their typical upload and download speeds and typical latency. That disclosure survived the 2026 rewrite intact: the order that streamlined the label rules left the speed and latency fields in place, and it added a requirement that telephone sales staff orally summarize “typical download and upload speeds; latency” among the fields a caller must hear. A March 3, 2026 filing in the docket — cited in the Commission’s own order — examined bills from Charter, Comcast, Cox, AT&T and Verizon and found none contained an actual measured download speed, upload speed or latency. Your bill has a price and a plan name. The label has a number.

Here is the part the competing pages get wrong. The FCC has not defined a statistic behind “typical” in the label rules. In the 2022 order the Commission wrote, “We decline to adopt a requirement that providers tie their actual speed reporting to ‘peak usage periods.’” It also set out where the number may come from: a fixed provider that participates in Measuring Broadband America may publish its MBA results, and one that does not may use the MBA methodology, its own internal testing, its own consumer speed test data, or reliable, relevant data from third-party sources. Five permitted routes to the same field, and no defined statistic behind any of them. One carve-out matters: a provider that uses MBA results or the MBA methodology is required to disclose mean upload and download speeds by tier during the “busy hour,” so for those providers the label figure is a peak-hour number.

The Commission was asked again to define the term, and again declined. In the Report and Order it adopted on July 22, 2026 and released the next day, it closed the inquiry it had opened in 2022 into changing the label’s performance fields. One commenter had urged it “not to close its inquiry into defining ‘typical’ performance metrics, arguing that the current approach results in disclosures that are not comparable between providers”; the Commission answered that “we agree with the commenters who argue that the current rule is sufficient.”

So the label figure is a self-defined number. It is still the only thing the provider wrote down, which is what you want when you open a ticket.

The 80/80 figure is the one the internet has attached to the wrong document. It belongs to Measuring Broadband America, the FCC’s panel-based measurement program, where it “measures the minimum percentage of the advertised speed experienced by at least 80% of subscribers for at least 80% of the time over peak periods.” Read that again: at least 80% of subscribers — a research statistic about a tier across a panel of metered routers, not a floor under your account. Any page telling you that “typical speed” on your label means what you personally get 80% of the time between 7 and 11 p.m. has fused a research metric with a disclosure rule, and the 2022 order declined to require providers to tie their speed reporting to peak usage periods at all.

Your speed test result measures less than people think: throughput between one device and one test server, over one path, for a few seconds, through every piece of your own equipment on the way.

What caps the number before your provider does

The FCC is unusually blunt about this in its own report. Connections in the home, it warns, may not have enough capacity to support peak loads: “some Wi-Fi connections may be limited to tens of Mbps which would be the maximum achievable throughput to a specific device using that type of connection, independent of whether a consumer may subscribe to a service delivering; e.g., hundreds of Mbps to the home.” That is the regulator saying that a Wi-Fi speed test on a gigabit plan can be meaningless.

This is why the FCC does not measure over Wi-Fi. Its own fixed measurements run on a wired “whitebox” router, and the download test “measures the download speed of each whitebox over a 10-second period, once per hour during peak hours (7 p.m. to 11 p.m.) and once during each of the following periods: midnight to 6 a.m., 6 a.m. to noon, and noon to 6 p.m.” If you want a number that behaves like the FCC’s numbers, you have to test the way the FCC tests.

The last row is the one that produces a fixable answer. A wired result that sits at the same ceiling at 2 p.m. and at 9 p.m., across two servers and two days, is not a congestion story and not a Wi-Fi story. It points at how the line is configured or at the line itself.

The protocol that produces a claim

Each step below exists to close off one thing a provider can say back to you. Run them in order.

1. Pull your label and screenshot it. Sign in to the account portal and open the broadband consumer label for your current plan. Providers that give customers an online portal must make the current plan’s label easily accessible there; the 2026 order keeps that requirement and adds that a prominent hyperlink or icon in the portal may stand in for the label itself, though that amendment is not yet in force. Capture typical download, typical upload and typical latency, with today’s date visible.
2. Read your ports. Check the Ethernet port markings on the router and the computer, and confirm your tier can physically fit through them.
3. Wire one device, quiet the rest. One computer on a cable, other devices paused or off, backups and streaming closed on the test machine.
4. Run an off-peak block. Around 2 p.m. local, run three tests against your usual server and three against a second server on a different network.
5. Run the same block during peak. Repeat on a weeknight inside 7 p.m. to 11 p.m. local, the FCC’s defined peak usage period.
6. Log everything. Date, clock time with time zone, wired or wireless, device, server, download, upload, latency, screenshot. A number without a timestamp is not evidence.
7. Repeat on a second day, so one bad afternoon cannot account for the result.
8. Compare to the label. Put your median wired off-peak download beside the typical download speed on the label, not beside the tier name.

For a reference point against the FCC’s own data: in the Thirteenth Measuring Broadband America Fixed Broadband Report, “eight measured download speeds were 100% or better than advertised speeds during the peak hours (7 p.m. to 11 p.m. local time)” out of 12 provider and technology configurations, and “the four other ISP/technologies provided between 86% to 90% of their advertised speed.” On the stricter 80/80 measure, DSL was weakest, “with providers achieving between 63% to 72% of the advertised speed.” Those results come from measurements taken in September and October 2022.

If your wired result is far below your label figure at every hour, on two servers, on two days, you have something specific to raise. If it only sags between 7 and 11 p.m., that is a congestion pattern rather than a provisioning one — usually a shared segment, not your line.

What to ask for, and what the process actually is

Ask the provider to compare your logged measurements against the typical download and upload speeds on your plan’s own label, and to open a ticket referencing that figure. Say it plainly:

I am comparing wired, single-device results to the typical download speed on my plan’s broadband label. Here are ten timestamped tests across two days and two servers. Please open a ticket against the label figure and tell me the provisioned rate configured for my line.

Two things worth knowing about the formal route. The 2022 order created no special enforcement machinery for labels: the Commission declined to adopt new rules, practices or procedures specifically for enforcement of the label, saying its existing enforcement mechanisms should enable it to enforce the new requirements. And under the Commission’s long-standing practice, reflected in 47 CFR 1.717, 30 days is the standard period for a carrier to respond to an informal consumer complaint. That is the documented process, not a prediction about your case.

There is also a reason to screenshot today rather than eventually. The 2022 order required providers to archive labels for at least two years after a plan stopped being offered to new customers, and the 2026 order restates why: the earlier order “explained that archived labels would help the Commission and state authorities investigate potential inaccuracies in labels, including in cases arising from consumer complaints.” The order adopted on July 22, 2026 reverses that — “We eliminate the requirement that providers archive all labels for at least two years after a service plan is no longer available to new customers” — although, as set out below, that particular amendment had not taken effect as of September 12, 2026. Your current plan’s label stays in your portal either way. The historical trail is the part on its way out, and a dated screenshot is a record you hold.

Caveats, confidence and limits

The throughput number can be the wrong metric entirely. A connection that tests at full speed and still feels broken during calls is usually a delay or loss problem, not a bandwidth problem. Cloudflare’s published scoring thresholds are a useful reference: packet loss under 1% scores best and over 5% worst; jitter under 10 ms best and over 20 ms worst; and the difference between loaded and unloaded latency is best under 10 ms and worst over 50 ms. If that describes your complaint, the throughput test will keep telling you everything is fine — start with latency, jitter and packet loss instead.

The FCC’s comparison figures are older than they look: the Thirteenth Report is built on measurements taken in September and October 2022, and SamKnows served as the FCC’s contractor for the program only through July 2023. Treat those percentages as a reference band, not as this year’s market.

And a boundary: this page is about measuring and documenting a shortfall against a stated figure. It is not legal advice and does not predict the outcome of any complaint or ticket. If you are still unsure whether the problem is inside your home at all, isolate that first with is it my Wi-Fi or my internet.