Illustration of several homes sharing one fiber line through a splitter, with a busy-hour clock and a speed gauge, for the TheWriter.id guide on explaining contention ratio to customers.

FTTH Operators • Customer education

How to Explain Contention Ratio or Oversubscription to Customers

By Junun Saleh · Updated October, 2026 · Built on Nokia, Ofcom, FCC, ITU, an AAISP engineer’s analysis, Selectra, Indosat and Wikipedia sources, each linked below

A customer on a 300 Mbps plan asks why the fiber is “shared” when the plan says “dedicated line” on the sales page. An agent answers “our contention ratio is 1:32”, and the customer hears a number that sounds like a 1-in-32 chance of getting service. Contention ratio is a true and useful engineering term, and it is among the least useful things to say to a customer. This page gives sales, support and marketing teams the plain-language explanation, the peak-hour expectations that can honestly be set, and the engineering that keeps a shared network from congesting.

Quick answer: Contention ratio compares the total speed sold on a shared link with the capacity of that link. Oversubscription means that total is higher than the capacity, on the bet that customers are not all busy at once. Explain it in three steps. First, shared capacity: one GPON fiber carries 2.488 Gbit/s downstream for all the homes on it, so 32 homes sold 300 Mbps each is a 3.9:1 ratio (our arithmetic). Second, peak hours: when many homes are online together, such as 8 to 10pm in Ofcom’s definition, each home’s share can fall, though published data shows well-run fiber stays close to plan. Third, engineering: operators measure the busy hour, add capacity before links run hot, and manage how what remains is shared.

In this page, contention ratio is the total demand sold on a shared link divided by its capacity, such as 8:1. Oversubscription is selling more total speed than the link can carry at once. A PON (passive optical network) is a fiber network in which one fiber is split passively to serve many homes. The OLT is the operator’s equipment at the head of the PON and the ONT is the unit in the home. A split ratio such as 1:32 is the number of homes on one PON. The busy hour is the period of highest demand on a link. The ITU-T standards behind these networks include G.984.1 (GPON) and G.9807.1 (XGS-PON).

How do you explain shared bandwidth and contention ratio with simple analogies?

Say that the plan speed is the top speed one home can reach, and that the fiber behind it is shared by many homes who are rarely all busy at once. A PON is a shared network: the OLT sends a single stream of downstream traffic that every home on that fiber receives, and bandwidth is commonly shared among its users (Wikipedia: Passive optical network). GPON carries 2.488 Gbit/s downstream, and most deployments split it 1:64, 1:32 or smaller. XGS-PON raises the shared capacity to about 10 Gbit/s and supports up to 1:128.

2.488 Gbit/sshared downstream on one GPON fiberWikipedia summary of ITU-T G.984

1:64the split ratio at which most PONs are deployed, per the same sourceWikipedia

3.9 : 1oversubscription for 32 homes on 300 Mbps plans sharing one GPON downstreamOur arithmetic

8 : 1an ISP engineer’s example: 100 customers on 80 Mb/s sharing a 1 Gb/s fibreAAISP, 2017

Figure 1: Oversubscription is arithmetic: speeds sold added up, divided by the capacity they share. Here 32 homes on a 300 Mbps plan sit on one 2,488 Mbps GPON downstream (illustrative).

Figure 1 is the whole idea in one picture. The speeds sold add up to 9,600 Mbps, the shared capacity is 2,488 Mbps, and the ratio is 9,600 divided by 2,488, about 3.9 to 1. That is not a defect. It works because home internet use is bursty: a household streaming a film uses a fraction of its line speed on average, and the moments when many homes ask for the full plan together are short. The number you quote to customers should therefore say what the worst case would be, and what a normal busy evening looks like. Customers often fold this into a wider question, so it helps to know how reliable fiber internet really is.

Why the ratio alone misleads

An engineer at the UK ISP AAISP argues that a contention ratio is a poor measure for three reasons. It ignores link size: two 10:1 links can have very different risk. It ignores usage: a ratio is only meaningful if you know the demand as well. And it can be gamed: a provider selling slow packages looks good on the ratio because the speed side is small. His preferred measure is the average throughput per customer on a link, in Mb/s (AAISP: contention ratios).

Figure 2: The same ratio can mean very different things. The ratio ignores how big the link is and how much the customers actually use it (illustrative scenario).

Analogies that work, and where each one breaks

Customers grasp sharing through everyday pictures. Each picture also leaves a wrong impression if you stop there, so finish the analogy with the corrective sentence.

Three analogies for shared bandwidth, with the sentence that fixes each one
Analogy	What it explains	Where it breaks, and what to add
Water main	Each home has a tap that could run at full flow, but one main feeds the whole street. If every tap opens together, each gets less.	Water is used up; data is shared moment by moment. Add: “Most taps are closed most of the time.”
Motorway	One road for many cars: fast at 3pm, slower at 6pm when everyone drives home.	It can sound as if the ISP is choosing to slow you. Add: “We widen the road when traffic grows.”
Shared lift	A building lift is fine all day and queues only when everyone leaves at once.	A lift queue is visible and fixed in length. Add: “The wait on our network is usually a small speed dip, not a queue.”

One contention explanation, written two ways (illustrative; use your own split and plan)

TypicalOur network uses a 1:32 contention ratio.

Customer-readyYour plan is 300 Mbps, and that is the top speed you can reach. The fiber that serves your street is shared by up to 32 homes, and it carries 2.5 Gbit/s in total. Most homes are quiet at any moment, so you normally get the full 300. If all 32 homes were downloading flat out at the same instant, each would get about 78 Mbps, which almost never happens. We watch the busiest hour on every shared fiber and add capacity when it fills.

How do you explain why broadband speed varies at peak hours?

Say that more homes are online together in the evening, so a shared link can reach capacity and each home’s share can fall. On a well-planned network that dip is small, and a large one is a fault to report. Nokia puts the effect plainly: during busy hours, long-lasting high demand may prevent users from reaching the advertised bitrate on a crowded PON (Nokia: Assure peak bandwidth for broadband subscribers).

Figure 3: Demand on a shared link follows the day. It reaches the link’s capacity only in the evening peak (schematic), so the same line can test fast at 3pm and slower at 9pm.

The curve in Figure 3 is a schematic, not measured data. It shows the shape that matters: demand follows the day and touches the link’s capacity only for a few evening hours. That is why a monthly average hides the problem and why a speed test at 3pm says little about 9pm. Regulators define the window for this reason. Ofcom reports UK speeds for 8 to 10pm (Ofcom: latest home broadband performance trends), and the FCC measures US performance on weeknights from 7 to 11pm local time (FCC Measuring Broadband America, thirteenth report).

8-10pmUK residential peak window used by Ofcom for peak-time speedsOfcom, Sept 2023

95%UK peak-time average download speed as a share of average maximum speed, March 2023Ofcom

8 of 12US ISP set-ups that met or exceeded advertised download speed at peak, Sept to Oct 2022FCC

20%+year-on-year data traffic growth Indosat reported over Eid 2026 on its mobile networkIndosat statement, March 2026

Figure 4: Well-run networks usually deliver close to the plan at peak, but not always at 100%. Published data supports saying so honestly: a small dip is normal, a large one is a signal.

Two readings of Figure 4 belong in any customer explanation. First, a small dip is normal: Ofcom’s peak-time average was about 95% of the average maximum speed, and the FCC found that most subscribers saw only slight degradation from off-peak to peak. Second, the dip is smallest on fiber: Ofcom found the effect of congestion during busy periods was lowest for full-fibre lines. The data does not say that fiber never congests. Four of the FCC’s 12 set-ups reached about 86% to 90% of advertised speed at peak, and individual PONs can run hotter than the average.

Demand spikes are real and sometimes national. Indosat reported more than 20% year-on-year traffic growth during the 2026 Eid homecoming and a government projection of up to 40% growth over Ramadan and Eid, on a mobile network and by the operator’s own account (Meatech Watch on Indosat). The lesson for fiber teams is that peaks can be forecast, so they can be planned for.

What a customer sees in the evening, what it usually means, and what to say
What the customer sees	What it usually means	What to say
Slightly lower at 8 to 10pm, near the plan	Normal sharing at the busy hour	“A small evening dip is expected. You are still close to your plan.”
Well below plan every evening, fine in the afternoon, on a cable	Shared-link congestion	“That pattern points to a busy shared link. Send us the times, and we will check its load.”
Low at every hour on a cable	Not contention: look at the line, port or equipment	“A problem at all hours is not about sharing. We will test the line.”
Low only on Wi-Fi	Home wireless limits	“Your line is fine. Let us look at the Wi-Fi.”

To tell the first two patterns apart, ask for a wired test. Ofcom’s customer advice is to test more than once at different times and to use a cable where practical (Ofcom: Get help with broadband and Wi-Fi problems), and Selectra suggests a laptop on Ethernet, three tests ten minutes apart, in the afternoon and in the evening, repeated over three evenings (Selectra: Your fibre line is shared with up to 31 other homes). Selectra is a comparison site, so treat that as practical advice and not a standard.

A peak-hour statement, written two ways (illustrative)

TypicalSpeeds may vary during peak hours.

Customer-readyBetween about 7 and 11pm more homes are online together, so a small dip below your plan speed is normal. On a cable you should still be close to your plan. If your evening tests are far below your afternoon tests, that is congestion and we want to know. Please send three wired test results from two evenings, with the times.

Contention is the hardest line on the plan page. It deserves one careful version.

“Shared”, “best effort”, “up to” and “peak hours” appear on the plan page, in the terms, on the sales call and in the support script, written by different people at different times. That is how a plan page ends up promising a dedicated line while the engineering describes a shared one. TheWriter.id writes this layer for FTTH operators and ISPs: contention and peak-hour explainers, plan-page notes, help-centre articles and support scripts, built from interviews with your network planners and checked by them before publication. The method is the one in How Fiber Providers Can Turn Technical Expertise Into a Lead Engine: publish the sourced explanation your engineers already give, in words customers can follow.

See TheWriter.id services

How do capacity planning and traffic management help ISPs control congestion?

ISPs control congestion by measuring the busy hour on every shared link, adding capacity before a link runs hot, spreading customers across links, and managing how the remaining capacity is shared. The AAISP engineer’s point is that there is no single contention ratio for an ISP, because sharing happens at several layers, from the customer line to the cabinet, backhaul, carrier links, edge routers and transit. Each layer needs its own check.

Where sharing happens on an FTTH network, and what the operator can do
Layer	Shared by	What the operator can do
PON (fiber and splitter)	The homes on one fiber, up to the split ratio	Move customers to a quieter port, lower the split, upgrade the PON generation
OLT uplink and aggregation	Every PON on that OLT	Add or upgrade uplink capacity
Backhaul and core	Whole regions of customers	Add capacity on busy routes; monitor loads
Transit and peering	The ISP’s customers and the wider internet	Add interconnect capacity; choose routes

Figure 5: Capacity planning is a loop, not a one-off: measure the busy hour, forecast, act before the PON runs hot, then verify with a peak-time test.

Two levers that change the worst case

Capacity per home depends on two things an operator controls: how many homes share a PON, and which PON generation carries them. Divide the shared capacity by the homes and the worst case, every home at full demand in the same instant, moves a long way. Figure 6 uses the GPON and XGS-PON rates and split ratios described above. Only XGS-PON at 1:32 covers a 300 Mbps plan even in the worst case, which is one reason operators plan around the busy hour and not the theoretical peak. Selectra reports that Openreach is deploying XGS-PON in the UK, and reads the move as raising shared capacity from 2.5 to 10 Gb/s; that is a comparison-site report and we have not independently verified the rollout figures.

Figure 6: Two levers change the worst case: fewer homes per PON and a faster PON generation. Even the worst case is rarely reached, which is why operators plan around the busy hour.

The trade-offs operators face

Nokia describes the trade-off directly. Competitors push advertised rates up, and the traditional response is to raise the overbooking factor so subscribers can reach higher rates without adding capacity, which increases congestion on the PON and reduces peak-rate availability. The opposite fix, fewer subscribers per PON, underuses the network and needs field work to move customers. Nokia’s own answer is a software app that uses machine learning on high-frequency traffic data to improve peak-rate availability on loaded PONs. The article is a vendor post and reports no measured results, so treat it as a description of the problem, not proof that the product works (Nokia blog, November 2022).

Traffic management then decides how what remains is shared. In GPON the OLT hands out upstream transmission opportunities to each customer unit according to its activity and its traffic contract, in priority order from fixed to assured to non-assured to best-effort (IP Infusion: DBA, the GPON upstream bandwidth maestro). The page has no named author or date, so use it as a plain explainer of the mechanism. Customers do not need the mechanism; they need to know that capacity is shared by rule, not by accident.

Operators also report using monitoring to see trouble early. Indosat said it used real-time monitoring and predictive analytics to detect congestion more than 60% faster during Eid 2026, on its mobile network and by its own account. Network planners compare these practices before they take a sales call, and they check what is published. What Network Planners and CTOs Actually Want Before They Take a Sales Call makes that point for fiber vendors generally: evidence first, claims second.

Say this about your engineering…

What you measure: busy-hour load on each shared link.
What you do when a link fills: move, add, upgrade.
How customers can check: a wired test, in the evening.

Avoid saying…

“Dedicated” for a shared PON.
“No congestion”, which no shared network can guarantee.
A bare ratio such as “1:32” with no capacity or plan speed.

An engineering claim, written two ways (illustrative; describe your own process)

TypicalWe continuously optimise our network to deliver a consistent experience.

Customer-readyWe watch the busiest hour on every shared fiber. When one gets crowded we move some customers to a quieter port, add capacity or upgrade it. If your cable tests in the evening are well below your plan, tell us the times and we will check the load on your fiber first.

What should a contention ratio explanation include before you publish it?

It should say what is shared, what the worst case is, what a normal evening looks like, and what the customer can do. Use this list to review a plan page, a help article, a support script or a sales answer.

Does it say the plan speed is the top speed one home can reach, not a guaranteed constant?
Does it name what is shared, and give the capacity and the number of homes, not only a ratio?
Does it explain oversubscription as a normal practice, and say when it becomes a problem?
Does it use one analogy, and add the sentence that stops the analogy misleading?
Does it name the busy window and say what a normal dip looks like?
Does it tell the customer how to check: a wired test in the afternoon and the evening?
Does it say what you measure and what you do when a shared link fills?
Does it avoid “dedicated” and “no congestion” unless they are true?
Does every number carry a date and a source, and who checks it before it goes live?

Contention ratio and oversubscription: frequently asked questions

What is contention ratio?

Contention ratio compares the total speed sold on a shared link with the capacity of that link, written as a ratio such as 8:1. One ISP engineer’s example is a 1 Gb/s fibre serving 100 customers on 80 Mb/s connections: 8 Gb/s of possible demand on 1 Gb/s of capacity, or 8:1. It describes sharing, not the speed you will get.

What is oversubscription, and is it deceptive?

Oversubscription means selling more total speed than a shared link can carry at once, on the assumption that customers are not all busy at the same moment. It is a normal way to share capacity. It becomes a problem when busy-hour demand regularly exceeds capacity and customers are not told what to expect.

Is a lower contention ratio always better?

No. The ratio ignores link size and usage. An ISP engineer notes that 10 homes on a 10 Mb/s link and 1,000 homes on a 1 Gb/s link are both 10:1, but the chance of slowdown is high on the first and low on the second. A better measure is average throughput per customer on a link, checked at the busy hour.

What contention ratio is normal for FTTH?

There is no single standard, because the ratio depends on where it is measured. On the fibre itself, GPON deployments mostly use 1:64, 1:32 or smaller splits, and XGS-PON supports up to 1:128. One ISP, PCN, says typical fibre ratios are about 10:1 and offers lower or dedicated 1:1 services. Ask which link a quoted ratio refers to.

How much bandwidth does each home get on a PON?

On GPON, 2.488 Gbit/s is shared downstream. Divided by 32 homes that is about 78 Mbit/s each if every home were busy at the same instant, and about 39 Mbit/s each on a 1:64 split. This worst case is rare, because most homes are idle at any moment. XGS-PON raises shared capacity to about 10 Gbit/s.

Why is my internet slower in the evening?

More homes on the same shared links are online at once. Ofcom measures UK residential peak time as 8 to 10pm, and the FCC measures US peak performance on weeknights from 7 to 11pm. Nokia notes that during busy hours, long-lasting high demand may prevent users from reaching the advertised bitrate on a crowded PON.

Does fiber slow down at peak hours?

Usually only a little on a well-planned network. Ofcom found UK peak-time average download speed was about 95% of the average maximum in March 2023, and that the effect of congestion was lowest on full-fibre lines. In the FCC’s 2022 data, eight of 12 ISP set-ups met or exceeded advertised download speed at peak and the other four reached roughly 86% to 90%.

How can I tell whether a slowdown is congestion?

Test with a laptop on an Ethernet cable, once in the afternoon and again in the evening. Selectra suggests running three tests ten minutes apart and repeating over three evenings. A line that is fine in the afternoon but poor in the evening points to shared infrastructure. A slowdown at all hours points more to your own equipment.

How do ISPs reduce congestion on shared fiber?

They measure the busy hour on each shared link, move customers to quieter ports, add PON ports or uplinks, upgrade to a faster PON generation, and manage how remaining capacity is shared. Nokia notes that raising the overbooking factor increases congestion, while cutting subscribers per PON underuses the network and needs field work.

Sources

Figures 1 to 6, the analogies, the scripts, the checklist and the pattern table are our illustrations, not published standards. Figure 3 is a schematic, not measured data, and the five active homes in Figure 2 are our illustration. The 3.9:1 ratio and the per-home shares in Figures 1 and 6 are our arithmetic (for example 9,600 ÷ 2,488, and 2,488 ÷ 32); XGS-PON is treated as about 10 Gbit/s. GPON rates and split ratios come from Wikipedia’s summary of the ITU-T standards, which we used because the ITU pages we could open give titles and status but not rates; the ITU pages are cited for the existence and title of each Recommendation. The AAISP post is a 2017 engineer’s argument, not a standard. Nokia’s post is a 2022 vendor article without measured results. The IP Infusion page has no named author or date. Selectra is a commercial comparison site and its Openreach rollout figures are unverified. The Ofcom and FCC figures are from different countries, years and methods and should be compared as patterns only. Indosat’s traffic and congestion figures are the company’s own statements about a mobile network, relayed by Meatech Watch. PCN is an ISP and its ratios describe its own services. We found no published Indonesian rule or national figure on contention ratios for residential FTTH.