Broadband Labels Are Back — and the Internet Has Changed Since 2016

The Infrastructure Investment and Jobs Act directed the FCC to finalize rules for a consumer broadband label, reviving a proposal first drafted in 2016. The intent is straightforward: give people a simple, standardized disclosure of price, data caps, performance, and network management practices when they shop for home Internet service. The FCC has opened a comment period on the draft label, and the public docket is available for review.

Why the 2016 Template No Longer Fits

The original label design predates a very different Internet. In 2016, the global online population was 3.4 billion; by 2021 it had grown to 5.2 billion. Daily digital media use among U.S. consumers has climbed from roughly 5.5 hours in 2015 to nearly 8 hours today. Traffic volumes have exploded — 79 zettabytes were consumed in 2021, a figure expected to double within two years.

The pandemic accelerated the shift from recreational use to necessity. Ninety percent of American adults describe the Internet as essential or important to them personally during the outbreak, and 40 percent report using it in new ways. Video conferencing, remote work, telehealth, and online education are no longer optional applications; they are mission-critical.

That dependence has raised the bar for tolerance. In 2016, users were more forgiving of sluggish connections and latency-sensitive apps like VoIP were less prevalent. Today a poor-performing video call reads as “my Internet is cutting out,” and the experience is judged far more harshly.

Performance Means More Than Bandwidth

Raw throughput has improved dramatically. Median fixed download speed in the U.S. went from 39 Mbps in 2016 to 194 Mbps in 2021. As capacity has become less of a bottleneck, other metrics have grown in importance. Latency — the time for a data packet to travel between two points — and jitter, its variance, now matter immensely for real-time applications. The FCC’s 2016 definition of latency is technically correct but vague: it does not clarify which path is measured or whether the measured path reflects traffic users actually generate.

Internet service providers are not the only parties affected by these measurement gaps. Enterprises increasingly rely on low-latency access to cloud services for both customers and employees. The quality of broadband is now foundational to the economy, not just a matter of household convenience.

From the edge provider’s vantage point, latency patterns are visible from the server side. Rural states show measurably higher TCP connect times than denser regions, a disparity that a well-designed label could expose. Cloudflare’s own browser isolation product, for example, requires very low latency — a constraint that drives direct interconnection with 10,000 networks across over 270 data centers to keep 95 percent of global users within 50 milliseconds of Cloudflare-protected services.

To make the label’s performance section meaningful, several changes are worth considering:

  • Replace “speed” with “throughput,” “bandwidth,” or “capacity.” Colloquial usage aside, “speed” conflates two distinct concepts. Throughput measures bit-carrying capacity; true “speed” in the user-experience sense is better captured by latency and overall quality of service.
  • Add jitter as a disclosed metric. Pandemic-era reliance on video conferencing made jitter a primary cause of perceived connection problems — the “am I frozen?” effect that users blame on their ISP.
  • Standardize how latency, jitter, and packet loss are measured. Consumers can only compare offerings if ISPs measure the same way. Rather than an immediate mandate, building stakeholder consensus on measurement methodology would yield more trustworthy numbers.

The underlying goal is to make disclosed performance match real user experience. Current speed tests often query endpoints embedded inside ISP networks — not the routes typical traffic takes — a practice that can flatter results. Performance measurement should approximate what users actually experience, not a synthetic best case.

Network Management: Beyond Yes or No

Performance is not purely a function of capacity. ISPs deploy traffic shaping to throttle high-bandwidth destinations and prevent congestion elsewhere. Some impose bandwidth caps that slow heavy users after a threshold, particularly on mobile networks. A simple binary disclosure of whether management techniques exist is insufficient. The label should state specific policies — for instance, when a slowdown kicks in after a monthly data allowance is exhausted.

The draft label’s privacy section, if it relies solely on a link to a dense policy document, falls short of genuine transparency. A useful privacy disclosure should address:

  • Collection and retention: Whether the ISP collects data beyond what is necessary for service delivery — including browsing history and location — and how long it is kept.
  • Use of information: Whether collected data serves purposes beyond providing the connection, such as advertising.
  • Sharing: Whether data including location or browsing activity is sold or shared with third parties.
  • Opt-out options: Whether subscribers can decline data use or sharing, and whether the ISP receives compensation for sharing.
  • Security: Whether the ISP has technical safeguards — such as encryption of metadata — and breach reporting mechanisms in place.

Beyond human-readable text, the data on the label should also be available in a machine-readable format, supporting researchers and consumer tools that can parse and compare offerings systematically.

Measuring the Internet Users Actually Get

The 2016 label design provides a solid foundation for pricing transparency. But the Internet — and how people depend on it — is not the same as it was six years ago. Broader performance metrics, concrete network management disclosures, and substantive privacy details would move the label closer to reflecting the real quality of service. Without that alignment, even the best-intentioned label will describe an Internet that no longer exists.