Updated: July 2026

Hot take: Low latency is more important than throughput when it comes to giving subscribers a superior quality of experience (QoE)🔥At Preseem, we like to say that throughput matters, but end users feel latency. When advertising their services, however, most ISPs only market plan speeds to consumers without ever mentioning low network latency as a selling point.

On one hand, this makes sense—consumers are more likely to understand what 100 Mbps download speed means, whereas promising them latency under 30ms probably won’t resonate right away. However, there was a time when consumers didn’t know what “Mbps” meant either. Educating subscribers on the importance of low latency and its impact on their experience could be a good way for regional operators to separate themselves from the competition and grow their business.

For regional operators serving rural and underserved communities, getting latency right carries particular weight. The FCC’s National Broadband Map shows that millions of locations across the country still lack access to reliable high-speed internet, meaning the ISPs building these networks are often the only option their subscribers have. In that context, delivering a connection that not only meets advertised speeds but actually feels fast is both a competitive advantage and a community obligation.

For regional ISPs, particularly those running fixed wireless networks where RF conditions, link rates, and shared AP capacity all directly affect latency, this is not a theoretical concern. A fixed wireless subscriber experiencing high latency during peak hours is not just having a bad evening. They may be on a radio that can’t support their plan speed under current signal conditions, on an AP approaching airtime saturation, or behind a backhaul link that fills up every evening.

Each of those root causes requires a completely different response, and none of them is visible without subscriber-level latency measurement. Getting latency right for these operators means knowing not just that a subscriber’s connection feels slow, but exactly where in the network that feeling is coming from.

What Causes High Latency

Before we look at the benefits of low latency, it’s important to understand what causes high latency in the first place. For network operators, packet buffering is the main culprit. More specifically, on networks that use first-in, first-out (FIFO) queues with large buffers, bandwidth-intensive interactive applications like online gaming or video calls can cause bufferbloat during peak busy times.

Excessively large buffers lead to queues with too many packets waiting to be transmitted. This results in saturated links that cause network bottlenecks and dropped packets, leading to high latency under load and making the connection feel slow for the subscriber.

For example, some physical interfaces can have queues that are thousands of packets deep. As you can imagine, this usually translates into more “my internet is slow” calls for ISP support teams to troubleshoot.

The traffic shaping algorithm your network uses determines how those queues are managed and whether subscribers feel the difference. For a detailed comparison of traffic shaping methods and why the choice between FIFO queuing and AQM has such a direct impact on subscriber QoE, see our guide to ISP traffic shaping software.

One finding worth noting for fixed wireless operators specifically: according to the 2026 ISP Network Report, the latency gap between fiber and fixed wireless subscribers largely disappears when both are idle. Under load, the difference primarily comes from the number of wireless hops: a fiber subscriber adds the latency of one wireless link (in-home Wi-Fi); a fixed wireless subscriber adds two (in-home Wi-Fi plus the last-mile FWA link). This means that for FWA operators, in-home Wi-Fi is not a minor footnote to the latency story; it’s a meaningful and often overlooked contributor to the loaded latency a subscriber actually experiences.

How to Measure Latency

The first step in reducing latency is to accurately measure it. At Preseem, we do this by measuring TCP Round Trip Time (RTT) latency using actual subscriber traffic. Here’s how it works: When ISPs use Preseem, all customer traffic passes through one of our inline devices, typically deployed in the network’s core. Each packet has an IP address that’s mapped to a subscriber and relevant network elements (e.g. access points).

We measure how long it takes for a TCP packet to go from the inline device all the way downstream—through the backhaul, access point, CPE, in-home network, and finally to the end-user device (such as a phone, TV, or laptop). TCP on this device then sends back an acknowledgement packet, and when that reaches Preseem, it calculates the RTT (which equals one latency sample). It then continues doing this for potentially thousands of latency samples for each subscriber and each 10-second logging interval.

By measuring latency in this way, Preseem enables regional operators to see where and why network latency occurs. For example, is it a congested AP, or an issue with an end user’s in-home Wi-Fi? This knowledge makes it easier for support teams to understand the causes of high latency issues and resolve them. It also allows ISPs to proactively address those issues and achieve low network latency moving forward.

What is Loaded Latency?

Loaded latency (also called latency under load) is the latency a subscriber experiences when their connection is actively being used—downloading a file, streaming video, or another household member running a speed test—as opposed to unloaded latency, which is measured on an idle connection. The difference between these two numbers is the most accurate indicator of bufferbloat on a network. An ISP can have excellent unloaded latency (5–10ms) but terrible loaded latency (200–500ms) if its traffic management system uses FIFO queuing with large buffers. Subscribers feel latency, even on a fast plan, as the connection “feels slow.”

Loaded vs. Unloaded Latency: Why Both Numbers Matter

An ISP can have genuinely excellent infrastructure and still deliver a poor subscriber experience, and the gap between unloaded and loaded latency is where that disconnect lives. Consider two subscribers on the same 50 Mbps plan on the same network:

Subscriber A — unloaded latency test on an idle connection:

• Download: 48 Mbps
• Upload: 22 Mbps
• Latency (unloaded): 8ms
• Result: Passes any standard speed test with flying colors

Subscriber A — loaded latency, same connection, same evening:

• One household member starts a Netflix stream
• Another joins a Zoom call
• Latency spikes to 180–220ms
• The Zoom call pixelates and drops
• Subscriber calls in to report slow internet

Nothing changed on the network between those two measurements. The access point is healthy, the CPE signal is fine, and the link rate is unchanged. What changed is that the connection is now under load, and the traffic shaping algorithm managing the queue is filling buffers indiscriminately rather than prioritizing the Zoom call over the Netflix stream.

Subscriber B — same plan, same network, AQM-based traffic shaping:

• One household member starts a Netflix stream
• Another joins a Zoom call
• Loaded latency: 12ms
• The Zoom call is clear throughout
• No support ticket generated

The difference between these two outcomes is not the network; it’s the queue management algorithm. Subscriber A’s ISP uses legacy FIFO queuing; Subscriber B’s ISP uses FQ-CoDel. The unloaded latency for both subscribers looks identical on any standard test. Only loaded latency reveals which one is actually delivering a good experience when it matters.

This is why both numbers matter, and why ISPs that benchmark only on speed test results or unloaded latency are working with an incomplete picture of their subscriber experience. The loaded latency gap is where silent churn originates, where support tickets are generated, and where the felt quality of a connection is determined every evening during peak hours.

How to Reduce Latency in Your Network

It stands to reason that if packet buffering and large queues are the main cause of high latency in networks, then finding a way to mitigate those will improve latency issues for subscribers.

Luckily, this solution already exists. Active Queue Management (AQM) is a traffic shaping method that proactively drops packets before buffers fill up. This keeps queue sizes small, meaning lower latency and a better experience for subscribers, even when multiple devices in the home are online at the same time.

Preseem uses the FQ-CoDel (Fair/Flow Queuing + Controlled Delay) AQM algorithm to separate traffic into bulk (e.g. system updates) and interactive (e.g. online gaming, VoIP) flows. This ensures the internet “feels fast” for end users, even during peak usage times. Created by Eric Dumazet and Dave Taht, FQ-CoDel enables “reduce bottleneck delays by several orders of magnitude.” As a result, this means fewer support calls and reduced churn for regional operators.

The Internet Engineering Task Force (IETF) calls FQ-CoDel “a powerful tool for fighting bufferbloat and reducing latency” due to its ability to provide isolation for low-rate traffic, keep queue lengths short, and its compatibility with a wide range of hardware.

How your bandwidth management software handles queue depth is the single biggest variable determining whether your subscribers experience low or high latency under load. For a full breakdown of what to look for when evaluating bandwidth management software, including AQM, subscriber-level enforcement, and mixed-vendor deployment, see our guide to ISP bandwidth management software.

Side note: Preseem’s Automatic AP Capacity Management feature can also help with AP congestion, packet buffering, and high latency issues. The ‘easy button’ for AP capacity management, this feature uses automatic bandwidth control to ensure a great subscriber experience at all times. Find out more here.

What’s Next for Low Latency: L4S

FQ-CoDel represents the current state of the art in AQM-based latency management, and for most regional ISPs, it remains the right tool for the job today. But the latency standards landscape is moving, and one development worth understanding is L4S — Low Latency, Low Loss, Scalable Throughput.

L4S is an IETF architecture, published as RFC 9330 in January 2023, that takes a different approach to the queuing delay problem. Where AQM techniques like FQ-CoDel manage queue depth reactively, keeping buffers from filling up, L4S addresses the root cause further upstream. Its premise is that queuing delay originates not in the queue itself, but in the congestion control algorithms that senders use to seek capacity. By pairing a new class of congestion controls at the endpoint with modified Explicit Congestion Notification (ECN) signals from the network, L4S aims to reduce queuing latency to roughly 1 millisecond under heavy load, compared to the 5–20ms average achieved by current state-of-the-art AQMs, including FQ-CoDel.

Real-world deployment is already underway. Comcast was among the first ISPs to trial L4S in production networks, Apple has incorporated L4S support into its operating systems since 2023, and T-Mobile announced network-level L4S support in July 2025. The architecture is designed for incremental deployment, meaning L4S-capable flows can coexist with conventional traffic on the same network without degrading the experience for subscribers whose devices haven’t yet adopted it.

For regional ISPs, L4S is not yet a near-term operational priority. Adoption depends on endpoint support in subscriber devices and on AQM implementations that can handle the DualQ scheduling L4S requires. But it is the direction the industry is moving, and understanding it matters for two reasons. First, as subscriber devices gain L4S support, networks that are already AQM-capable will be better positioned to leverage it. Second, the underlying principle—that latency is a first-class network quality metric, not an afterthought—is the same principle that makes FQ-CoDel worth deploying today.

To put L4S in context against where the industry stands today: our 2026 ISP Network Report found that most access points deliver service with less than 100ms of latency during peak times. Modern equipment, particularly newer Cambium and Tarana platforms, approaches latency figures competitive with fiber.

Older technologies, including 802.11n and 4G LTE, fall above that 100ms threshold, which is why equipment generation matters when evaluating a network’s readiness for L4S. The architecture is designed to deliver sub-1ms queuing latency at the network level, but that improvement requires both AQM-capable network equipment and endpoint devices that support the new congestion control protocols. For operators already running modern APs and AQM-based traffic shaping, the path to L4S readiness is shorter than for those still operating legacy equipment.

For a deeper discussion of L4S and its implications for ISP networks, the Preseem podcast episode with Bjørn Ivar Teigen, Head of Research at Domos and an active contributor to IETF latency standardization work, covers the topic in detail.

How Subscribers Can Improve Internet Latency at Home

So far, we’ve looked at ways that ISPs can reduce network latency, but what about the end users themselves? Is there anything they can do within the home to improve internet latency? Spoiler alert: Yes, there is 🙂

If you’ve got online gamers in the house or you’re working remotely and doing a lot of video meetings, then internet latency should be an important priority. Laggy games, slow system updates, and dropped or jittery video calls are extremely frustrating, to say the least. Below are a few things to think about if you’re experiencing any of these issues.

Optimize Your Router Placement

Conventional wisdom has it that a central location in the home is best for your router so that signal strength is distributed evenly throughout. For example, if you place your router next to an outer wall, some of that signal is absorbed by the wall. If your signal is poor and you have active users inside or outside the house, it will negatively affect the speeds each user can achieve.

Also, make sure the router is elevated and in a “low-tech” area where other devices are unlikely to interfere. For example, kitchens are not ideal because they have large metal appliances that can block access to other parts of the house. As a rule of thumb, try to avoid going through more than two walls to reach an access point. When in doubt, contact your internet service provider and ask for advice—most offer consultation services to help you figure out the ideal router placement.

Check Your Equipment

To help ensure high performance and low latency, it’s best to check you have a router that can handle what you need. Perhaps your router is a little older and needs a firmware update, or it should be replaced with a newer model. For gamers, there are routers designed specifically for gaming that you might want to consider. As well, some routers now come with smart queue management built right in.

If you’re using a router/modem combo provided by your ISP, check in with them to make sure you have the proper model based on your needs and internet usage. Bonus: if they’re using Preseem, they’ll be able to tell you whether any issues you’re experiencing are network-wide or isolated to your in-home equipment.

Run a Speed Test

End users can easily measure their network latency by running a speed test. CloudFlare is a recommended option, as it shows not only upload and download speeds, but also measurements for latency under load, jitter, and packet loss. It’ll even rate your current network quality for online gaming and video streaming.

Netflix’s fast.com is another good option that shows both loaded latency and internet speed.

Loaded latency (or latency under load) is the measurement of latency when the network is busy (e.g. during peak time or when multiple users in the home are online at the same time). High latency is an indicator of bufferbloat.

Ask Your Provider About Latency

This is an overlooked tip for improving internet latency at home, but it’s important. Choosing an internet service provider that consistently delivers low latency is more important than download/upload speeds, because latency directly affects how the internet feels for users.

A 100Mbps plan may be more than most users will ever need; however, if your provider’s network has latency issues caused by large buffers, the internet will still feel slow during heavy usage. So, internet users: do your homework and choose an ISP with low latency! And ISPs: make sure to educate your audience on the benefits! It’s a true win-win situation.

For more insights on network latency, the impact of packet buffering, and the changing face of the internet, check out our webinar on the subject or read the recap blog here.

Frequently Asked Questions

What is latency in networking?

Latency is the time it takes for a packet of data to travel from a sender to a receiver and back again, measured as Round Trip Time (RTT). For ISPs, latency is one of the most important indicators of subscriber experience—more so than throughput in many cases, because latency is what subscribers feel directly. A connection with high latency feels slow and unresponsive even if it’s delivering advertised download speeds, whereas a connection with low latency feels fast and immediate even on a modest plan.

What is loaded latency and why does it matter?

Loaded latency, also called latency under load, is the latency a subscriber experiences when their connection is actively being used, such as when household members are streaming video, downloading files, or running a speed test simultaneously. Unloaded latency, measured on an idle connection, is typically much lower and is what most speed tests report by default. The gap between these two numbers reveals bufferbloat: if a subscriber’s unloaded latency is 8ms but their loaded latency spikes to 200ms, the traffic shaping algorithm managing their connection is filling buffers indiscriminately rather than keeping queue depth low. That gap is where dropped Zoom calls, laggy games, and “slow internet” complaints originate.

What causes high latency for ISP subscribers?

The primary cause of high latency on ISP networks is packet buffering, specifically, networks using FIFO (first-in, first-out) queues with large buffers that fill up under load. When a link or virtual subscriber link approaches capacity, packets queue up waiting to be transmitted. The longer that queue, the higher the latency for every packet behind it. This problem occurs at every congestion point in the network: AP and backhaul links under heavy usage, and within individual subscriber virtual links when a subscriber approaches their plan ceiling. The solution is Active Queue Management, which actively manages queue depth to keep latency low even under congestion.

What is bufferbloat and how does it affect subscribers?

Bufferbloat is the excessive latency that occurs when a network device uses large buffers to absorb traffic bursts rather than actively managing queue depth. Under normal conditions, large buffers seem beneficial because they prevent packet loss. But under load, they cause packets to queue for hundreds of milliseconds before being transmitted, making the connection feel slow and unresponsive even though throughput is technically available. Subscribers experience bufferbloat as Zoom calls that drop when someone else in the household starts streaming, games that lag during peak hours, and connections that feel fast on a speed test but slow in actual use.

What is AQM and how does it reduce latency?

Active Queue Management (AQM) is a traffic shaping approach that proactively manages queue depth rather than allowing buffers to fill. Instead of dropping packets only when a buffer is completely full, AQM drops or marks packets earlier, signaling to senders to reduce their transmission rate before congestion causes latency to spike. FQ-CoDel, the specific AQM algorithm used by Preseem, goes further by separating traffic into individual flows and managing each flow’s queue independently. This keeps interactive traffic like Zoom calls, gaming, and VoIP responsive even while bulk traffic, such as downloads, system updates, and streaming, is using the same connection.

What is L4S and will it replace FQ-CoDel?

L4S (Low Latency, Low Loss, Scalable Throughput) is an IETF architecture published as RFC 9330 in January 2023 that aims to reduce queuing latency to roughly 1 millisecond under load, significantly lower than what FQ-CoDel achieves today. Rather than replacing FQ-CoDel, L4S works differently: it addresses congestion at the sender level using modified congestion control algorithms and Explicit Congestion Notification signals, rather than managing queues at the network level. Real-world deployment is already underway: Apple added L4S support to its operating systems in 2023 and T-Mobile announced network-level support in July 2025. For most regional ISPs, FQ-CoDel remains the right tool today, but networks that are already AQM-capable will be better positioned to support L4S-enabled subscriber devices as adoption grows.

How can subscribers test their latency at home?

Subscribers can measure both loaded and unloaded latency using Cloudflare’s speed test at speed.cloudflare.com, which reports latency under load, jitter, and packet loss alongside upload and download speeds. Netflix’s fast.com also shows loaded latency. A large gap between unloaded and loaded latency results (for example, 10ms unloaded but 180ms under load) is a reliable indicator of bufferbloat, and is worth raising with the ISP’s support team.