Blog 8 min *

* Publication times use Tampa, Florida local time (Eastern Time).

An overhead view of a hand holding a mid-tier Android phone with a dark screen, beside the corner of a laptop keyboard.

What Device and Network Your Customer Is Actually On

By Gabriel Lopez-Seco

Picture the visitor coming to your website right now. If you are like most business owners, you picture them on a reasonably modern phone, probably on 5G, sitting somewhere with a decent signal. You picture your own experience of the site, which you probably tested on the office WiFi.

That picture is wrong, and the gap between it and reality is where your leads disappear.

Two independent data sources: Alex Russell’s 2024 performance analysis derived from Cloudflare Radar, and Ookla’s Speedtest Intelligence for H2 2024, tell a different story about what is actually happening on the other end of your URL.

What the P75 Baseline Actually Says

Here is the number that matters most: the 75th-percentile real-world mobile user loads your page on a connection running at 7.2 Mbps down with a 94 ms round-trip time (Alex Russell / Infrequently Noted, using Cloudflare Radar data, 2024).

RTT, round-trip time, is the time it takes a signal to travel from your customer’s device to your server and back. It is the baseline delay baked into every request the browser makes before a single byte of content loads. At 94 ms, that delay is 94 milliseconds per round-trip. A modern website makes dozens of these round-trips as it loads images, scripts, fonts, and tracking pixels.

7.2 Mbps is not a slow connection by rural standards. It is simply the reality of what a real browser session looks like at the 75th percentile, meaning one in four of your visitors is on something worse than this.

Compare that to the headline figure: Ookla’s Speedtest Intelligence for H2 2024 reports T-Mobile’s 5G median download at 281.52 Mbps. The US all-tech mobile average on Verizon is 86.23 Mbps (Ookla Speedtest Intelligence, 2025, H2 2024 data). These are real numbers. They are also opt-in speed test results, numbers captured when an engaged user opens the Speedtest app specifically to test their connection. That is not your customer browsing for a plumber at 2pm on a Wednesday.

The methodology difference is the whole story. Ookla captures a dedicated speed test. Russell’s Cloudflare Radar analysis captures what a webpage actually sees during a real browsing session, a fundamentally different thing. Use the 7.2 Mbps figure when asking what your site’s performance budget actually is. That is the real budget.

The Carrier Spread Is Wide

Not all of your customers are on the same carrier, and the spread matters more than most business owners realize.

Ookla Speedtest Intelligence for H2 2024 shows the spread within US mobile alone: T-Mobile’s 5G median hits 281.52 Mbps. Verizon’s all-tech mobile median is 86.23 Mbps. That is a 3x difference on download speed before you have accounted for 5G vs 4G LTE, coverage area, or whether the customer is indoors (Ookla Speedtest Intelligence, 2025).

A customer on T-Mobile 5G in downtown St. Pete is on a materially different experience than a customer on Verizon LTE in a Hillsborough County suburb. Your one website serves both. Designing for the faster of the two is a business decision that costs you the slower half of your audience.

And the carrier spread is just one dimension. Even within a single carrier, the difference between 5G coverage and 4G LTE coverage is a 3x swing on latency and a significant swing on throughput. Suburban Tampa, commercial buildings, job sites, medical waiting rooms, these are places where your customers call from and where 4G LTE is a realistic description of what they are on.

Latency Is the Tax That Never Goes Away

Here is the figure that business owners almost never see: global mobile latency averages 27 ms compared to 9 ms on fixed broadband (Ookla Speedtest Global Index, 2024).

That 3x latency gap means that even a fast 5G connection imposes a delay on every round-trip that a home WiFi connection does not. T-Mobile’s lowest all-tech mobile latency in H2 2024 was 49 ms (Ookla Speedtest Intelligence, 2025). Compare that to the 9 ms of a wired connection. The difference is not dramatic in isolation. It compounds across dozens of round-trips as your site loads.

If your homepage requires 60 HTTP requests, not unusual for a site built on a modern CMS with plugins and tracking scripts, each request carries that latency tax. On a 9 ms wired connection, those 60 round-trips contribute roughly 540 ms of pure latency overhead. On a 49 ms mobile connection, the same 60 round-trips contribute nearly 3,000 ms, almost 3 full seconds, before a single byte of content has been processed by the browser.

That math is why pages built for desktop WiFi feel broken on mobile, even on a relatively fast 4G connection.

The 4.5x Gap: Your Testing Environment vs Your Customer’s Reality

There is a specific version of this problem that shows up in almost every SMB performance conversation.

You tested your website at your desk. Your office WiFi is running on a fixed broadband connection. Ookla’s Speedtest Global Index reports US fixed broadband download at 389.80 Mbps (Ookla, 2025–2026). Verizon’s all-tech mobile median is 86.23 Mbps.

That is a 4.5x speed difference between your testing environment and your median Verizon mobile customer, before page weight, before latency, before any consideration of device processing speed (Ookla Speedtest Global Index + Speedtest Intelligence H2 2024, 2025–2026).

Your site felt fast because you were testing it on the fastest connection in the building. Your customer experienced something fundamentally different.

What Analytics Is Not Telling You

Most business owners who check their analytics have the device breakdown: X% mobile, Y% desktop. That is useful, but incomplete in a specific way.

Seeing device type without seeing connection quality is like knowing your restaurant’s customer count without knowing how long the wait was. The device distribution tells you how people arrived. It tells you nothing about what they experienced when they got there.

Your analytics tool shows you mobile vs desktop. It does not show you whether those mobile sessions came from a 5G T-Mobile connection or a Verizon 4G LTE connection in a building with weak signal. It does not show you how many of those sessions had a 94 ms RTT vs 27 ms. That data exists in Chrome User Experience Report (CrUX) field data, the real-user measurement data Google collects from actual Chrome browser sessions on actual devices and connections.

CrUX field data shows connection-quality-influenced performance metrics like LCP (Largest Contentful Paint, how long your biggest visible element takes to appear) and INP (Interaction to Next Paint, how long the page takes to respond to a tap or click). These are what Google uses for search ranking signals, and they reflect what your actual visitors experienced, not what a simulated test on a desktop browser showed.

A Worked Example: The Dental Practice

Take a dental practice in the Tampa Bay area booking appointments at $200 per visit. They get roughly 80 web-initiated appointment inquiries per month. Their site loads on a P75 mobile connection: 7.2 Mbps, 94 ms RTT, in 6.2 seconds. Their competitor’s site loads in 2.8 seconds.

If losing one booking per week to that faster competitor is a reasonable estimate of the behavioral gap, a patient in a waiting room who opened both sites and went with the one that responded, that is 4 lost bookings per month. At $200 per visit, that is $800 per month, or $9,600 per year, in lost revenue attributable to the connection-quality gap between testing environment and real customer experience.

The practice owner thinks the site is fast. They tested it from the office. The problem is invisible from the office.

The Fix Starts With Accurate Diagnosis

The reason this gap persists is not that business owners do not care. It is that the testing environment available to most business owners: Lighthouse, PageSpeed Insights, a quick load on a desktop browser, is categorically different from what CrUX field data shows about real mobile sessions on real connections.

Knowing your P75 network baseline, your LCP on mobile CrUX data, and your carrier distribution is not a developer concern. It is a revenue concern. It tells you whether the customers who are actually calling you are experiencing a site that works or a site that sends them to a competitor who loads first.

Gabriel Lopez-Seco at iServU runs a free assessment that pulls your actual CrUX field data, the real-user measurements from your real visitors, not a simulated lab test, and maps them against the P75 baseline your customers are actually on. No account manager. No handoffs. A direct read of what your visitors experience, in terms you can act on.

Schedule a free assessment at iServU.

Sources

  1. Alex Russell / Infrequently Noted (2024). “The Performance Inequality Gap, 2024.” Using Cloudflare Radar P75 data. https://infrequently.org/2024/01/performance-inequality-gap-2024/
  2. Ookla Speedtest Intelligence (2025, H2 2024 data). United States Speedtest Connectivity Report H2 2024. https://www.ookla.com/research/reports/united-states-speedtest-connectivity-report-h2-2024
  3. Ookla Speedtest Global Index (2024). 5G Page Load Speed Q1 2024 / Global Index. https://www.ookla.com/articles/5g-page-load-speed-q1-2024
  4. Ookla / Speedtest Global Index (2025–2026). United States fixed broadband and mobile median download speeds. https://www.speedtest.net/global-index/united-states