Skip to content
Illustration of a speed gauge reading below the plan marker, with a laptop on Wi-Fi and a wired cable test side by side, for the TheWriter.id guide on speed test results lower than the plan.

FTTH Operators • Customer education

How to Explain Why Speed Test Results Are Lower Than the Plan

By Junun Saleh · Updated · Built on Ofcom, M-Lab, Cloudflare, ISPreview, Snapone, Bisnis.com and Telkomsel sources, each linked below

A customer pays for 300 Mbps, runs a speed test on a phone in the bedroom, sees 90 and calls support. If the agent says “it is your Wi-Fi”, the customer hears “it is your fault”. If the agent says “we will check the line”, the engineer wastes a visit. Both are guessing, because a speed test does not measure the plan. It measures the narrowest point between one device and one server at one moment. This page gives support, sales and marketing teams the explanation, the three places a speed is lost, and a repeatable test process that separates in-home limits from access-network problems.

Quick answer: A speed test result is lower than the plan because it reports the slowest link on the whole path, not the plan speed at the router. Explain it in three steps. First, Wi-Fi: a phone tests the wireless link on top of the plan, and Ofcom notes a wired connection is usually faster and more reliable, so test with a cable first. Second, the path: the home, the shared access network and the test server can each be the narrowest point, and server distance, streams and time of day change the number. Third, prove it: run the same wired test several times at peak and off-peak, then add Wi-Fi, and the step where the number falls shows which side to investigate.

In this page, plan speed is the speed the contract describes, delivered to the customer’s router. Throughput is the data rate a test actually achieves. A link rate is the negotiated Wi-Fi rate between two devices, which is always higher than the throughput you can use. Latency is the round-trip time for data, and loaded latency is that time while the connection is busy. A bottleneck is the slowest point on the path, which sets the result.

How do you explain Wi-Fi vs wired speed test results to customers?

The plan is delivered to the router. A phone on Wi-Fi then tests a second link, the wireless one, and that link is usually the weaker of the two. Ofcom’s advice is plain: a wired connection is usually faster and more reliable, and it suggests connecting a device with an Ethernet socket directly to the router when the device does not need to move (Ofcom: Get help with broadband and Wi-Fi problems). That is why a cable test is the first step in any complaint about speed.

29.8%of Ethernet speed was what UK Wi-Fi tests reached in Q1 2023, in Ookla dataISPreview, May 2023
20-40 Mbpsreal-world throughput in a good 2.4 GHz set-up (vendor rule of thumb)Snapone Araknis
150-400 Mbpsthe same rule of thumb for a good 5 GHz set-upSnapone Araknis
~1 Gbpstypical wired Ethernet link to a router, so a cable rarely limits a plan below thatISPreview, May 2023

The Ookla figure comes from user-run Speedtest.net tests, and ISPreview points out that those tests record no device, room, wall material or distance, so treat 29.8% as a sign of the gap, not a rule. The vendor ranges are for one three-stream Wi-Fi 5 access point and will differ for other equipment.

Ethernet and Wi-Fi tests on the same 300 Mbps planThree horizontal bars against a dashed line marking a 300 Mbps plan. The Ethernet bar reaches the plan line. The 5 GHz Wi-Fi bar solidly covers 150 Mbps with a lighter band to the plan line. The 2.4 GHz Wi-Fi bar covers only 20 Mbps with a lighter band to 40 Mbps.One 300 Mbps plan, three ways to measure itPlan: 300 Mbps (illustrative)Laptop on an Ethernet cableLimited by the plan, not by the home.Phone on 5 GHz Wi-FiVendor rule of thumb: 150 to 400 Mbps in a good 5 GHz set-up.Phone on 2.4 GHz Wi-FiVendor rule of thumb: 20 to 40 Mbps in a good 2.4 GHz set-up.Schematic. Light band = range, for one 3-stream Wi-Fi 5 access point (Snapone Araknis guide), not every router.The cable result is capped by the plan, so one line can look fast or slow depending on how the device is attached.Ethernet and Wi-Fi tests on the same 300 Mbps planThree horizontal bars against a dashed line marking a 300 Mbps plan. The Ethernet bar reaches the plan line. The 5 GHz Wi-Fi bar solidly covers 150 Mbps with a lighter band to the plan line. The 2.4 GHz Wi-Fi bar covers only 20 Mbps with a lighter band to 40 Mbps.One 300 Mbps plan, three testsPlan: 300 Mbps (illustrative)Laptop on an Ethernet cableLimited by the plan,not by the home.Phone on 5 GHz Wi-FiVendor rule of thumb:150 to 400 Mbps, good set-up.Phone on 2.4 GHz Wi-FiVendor rule of thumb:20 to 40 Mbps, good set-up.Schematic. Light band = range, for one 3-streamWi-Fi 5 access point (Snapone Araknis guide).A cable result is capped by the plan.

Figure 1: The same plan measured three ways (schematic, using one vendor’s rule-of-thumb ranges). The test shows the weakest link between the device and the server, and on Wi-Fi that link is often the wireless one.

Why the number on the router box is not a speed

A router sold as 1,750 Mbps adds its two radios together: 450 Mbps on 2.4 GHz and 1,300 Mbps on 5 GHz. A phone uses one band at a time. Most phones have two Wi-Fi streams where the access point has three, which takes the 5 GHz figure to about 867 Mbps (our arithmetic: two thirds of 1,300). Wi-Fi is also half-duplex, so a device sends or receives but not both at the same moment, which the Snapone guide says roughly halves the usable rate. Distance, walls and neighbours then take more. Even the guide’s best-case 400 Mbps is under a quarter of the number on the box.

How a router label of 1,750 Mbps shrinks to a real Wi-Fi resultFive bars in sequence. The label reads 1,750 Mbps. One band at a time gives 1,300. A phone with two streams gives about 867. Half-duplex operation gives about 433. Walls and interference leave a vendor rule of thumb of 150 to 400 Mbps.From the router’s label to a speed test resultNumber on the box450 + 1,300 Mbps added up1,750One band at a time5 GHz, 3 streams, 80 MHz1,300Phone has 2 streams2 of 3 streams: about 867867Wi-Fi is half-duplexSend or receive, not both433Walls and interferenceVendor rule of thumb150 to 400Example: a 3×3 Wi-Fi 5 access point sold as 1,750 Mbps (Snapone Araknis guide). 867 is our arithmetic (2/3 of 1,300).Half-duplex roughly halves the rate in the guide. Schematic: real results depend on the room, phone and neighbours.How a router label of 1,750 Mbps shrinks to a real Wi-Fi resultFive bars in sequence. The label reads 1,750 Mbps. One band at a time gives 1,300. A phone with two streams gives about 867. Half-duplex operation gives about 433. Walls and interference leave a vendor rule of thumb of 150 to 400 Mbps.From the router label to a resultNumber on the box450 + 1,300 Mbps added up1,750One band at a time5 GHz, 3 streams, 80 MHz1,300Phone has 2 streams2 of 3 streams: about 867867Wi-Fi is half-duplexSend or receive, not both433Walls and interferenceVendor rule of thumb150 to 400Example: a 3×3 Wi-Fi 5 access point sold as1,750 Mbps (Snapone Araknis guide). The 867 isour arithmetic. Half-duplex roughly halves therate. Schematic; real rooms vary.

Figure 2: Router box numbers add the radios and use the most streams. A phone uses one band, fewer streams and shares the air with everything nearby, so the result sits far below the label.
What lowers a Wi-Fi speed test result, and what to tell the customer
Factor What the sources say What to say
Distance and walls Ofcom: Wi-Fi works best with fewer obstacles and walls between router and device; keep the router off the floor and roughly central “Test next to the router first, then in the room where it feels slow.”
Band: 2.4 GHz or 5 GHz Snapone: 20 to 40 Mbps on 2.4 GHz and 150 to 400 Mbps on 5 GHz in a good set-up (one access point type) “Connect to the 5 GHz network for the test.”
The device itself Snapone: most mobile devices have fewer streams than the access point. Ofcom: if one device is much slower than the others, the problem might be that device “Try a second device. If only one is slow, the line is not the cause.”
Interference Ofcom lists microwaves, baby monitors, cordless phones and dimmer switches as possible sources “Move the router away from those and test again.”
Other traffic and old devices Snapone: more devices mean longer waits to transmit, and a slow legacy device holds the air longer for everyone “Pause downloads and streams while you test.”

A cable test tells you…

  • Whether the plan reaches the router.
  • What the line does at different hours.
  • Whether the access network deserves a fault ticket.

A Wi-Fi test tells you…

  • What one device gets in one spot.
  • Whether the router position, band or device is the limit.
  • Where Wi-Fi coverage needs help, not the line.

One speed explanation, written two ways (illustrative)

TypicalWi-Fi speed depends on many factors, so results may vary from your package speed.
Customer-readyYour plan is 300 Mbps at the router. A phone on Wi-Fi measures the wireless link on top of that, and walls, the band and the phone all lower it. To see the plan itself, plug a laptop into the router with a cable and test again. If the cable test is close to 300 and the phone is not, the line is fine and we will help with the Wi-Fi.

How do you explain end-to-end speed bottlenecks?

A speed test reports the narrowest point on the whole path between the device and the test server. That point can be in the home, in the shared access network, or at the test server. Cloudflare describes its own result as reflecting “the combined performance of all components along the path to the content”, including local factors such as the browser and Wi-Fi (Cloudflare: How does Cloudflare’s speed test really work?). Customers tend to hear a single number as a verdict on the ISP. It is a verdict on the path. Much of that path is in-home Wi-Fi, which is why the customer experience ends in every room.

The path between a phone and a speed test serverSix boxes in a row: phone or laptop, home Wi-Fi, router and ONT, ISP access network, ISP core and peering, and test server. Brackets group the first three as customer premises, the next two as the ISP network and the last as the test provider. Small pulses travel along the links. In turn, the home Wi-Fi, the shared access network and the test server are highlighted with a note about how each can cap the result.The path a speed test crossesCustomer premisesISP networkTest providerPhoneor laptopHomeWi-FiRouterand ONTISP accessnetworkISP coreand peeringTestserverHome Wi-FiWalls, noise and the phone’sradio often set the ceiling.Shared accessBusy hours leave lesscapacity per customer.Test serverFar, busy, or one-streamtests report less.A speed test reports the narrowest point on the whole path. Schematic: the links are not to scale.Where the server sits changes what the test can see: M-Lab places its servers outside ISP last-mile networks.The path between a phone and a speed test serverSix boxes in a row: phone or laptop, home Wi-Fi, router and ONT, ISP access network, ISP core and peering, and test server. Brackets group the first three as customer premises, the next two as the ISP network and the last as the test provider. Small pulses travel along the links. In turn, the home Wi-Fi, the shared access network and the test server are highlighted with a note about how each can cap the result.The path a speed test crossesPhone or laptopCustomer premisesHome Wi-FiCustomer premisesRouter and ONTCustomer premisesISP access networkISP networkISP core and peeringISP networkTest serverTest providerHome Wi-FiWalls, noise and thephone radio often capthe speed.Shared accessBusy hours leave lesscapacity per customer.Test serverFar, busy, or one-streamtests report less.A speed test reports the narrowest point on thewhole path. Schematic, not to scale. M-Lab placesits servers outside ISP last-mile networks.

Figure 3: A speed test measures the narrowest part of the path, and any of three places can be the narrowest: the home Wi-Fi, a busy shared access segment or the test server itself.

In our view, fiber has a lot of headroom, so on a healthy FTTH line the narrowest point is more often in the home or beyond the access network than on the fiber itself. We have no measured share to quote, which is why the test process in the next section matters.

Why two tests on the same line disagree

Test design changes the number. Cloudflare says that, even with the same method, results differ noticeably with the distance to the server, and that how a test ramps up data, how many streams it opens and how it aggregates samples all change outcomes. M-Lab’s NDT uses a single stream because it measures bulk transport capacity more accurately, while a multiple-stream test is less sensitive to spurious packet loss and may mirror the best of its streams (M-Lab: Speed tests, accuracy, NDT and M-Lab). M-Lab also points out that its servers sit outside ISP last-mile networks, so a test crosses the points where ISPs connect to the wider internet, and that, anecdotally, tests hosted inside an ISP network usually read higher.

500+servers M-Lab ran its NDT test on worldwide in 2019, outside ISP last-mile networksM-Lab, March 2019
~100 msidle-to-loaded latency gap on a wired link that Cloudflare says points toward the ISPCloudflare blog
8-10pmresidential peak time in Ofcom’s 2019 broadband speed codesComms Ombudsman
51.40 MbpsIndonesia’s fixed broadband median download in Ookla’s August 2026 indexBisnis.com, 14 Oct 2026

Speed test results for the same line under different test set-upsTwo panels of three bars against a dashed plan line. In the first panel, a near server with many streams comes closest to the plan, a far server with many streams is lower, and a far server with a single stream is lowest. In the second panel, mid-morning and mid-afternoon tests come close to the plan, while an evening peak test is clearly lower.Same line, same plan, different resultsServer distance and streamsNear server, many streamsFar server, many streamsFar server, one streamplanTime of dayMid-morningMid-afternoonEvening peak, 8 to 10pmplanSchematic, not measured data. Server distance, stream count and test design change results (Cloudflare, M-Lab).Ofcom’s codes put residential peak at 8 to 10pm, and Ofcom advises testing more than once. One test is a sample.Speed test results for the same line under different test set-upsTwo panels of three bars against a dashed plan line. In the first panel, a near server with many streams comes closest to the plan, a far server with many streams is lower, and a far server with a single stream is lowest. In the second panel, mid-morning and mid-afternoon tests come close to the plan, while an evening peak test is clearly lower.Same line, different resultsServer distance and streamsNear server, many streamsFar server, many streamsFar server, one streamplanTime of dayMid-morningMid-afternoonEvening peak, 8 to 10pmplanSchematic, not measured data. Server distance andstream count change results (Cloudflare, M-Lab).Ofcom defines residential peak as 8 to 10pm.

Figure 4: The same line can return very different numbers depending on the server, the number of streams and the hour (schematic). Run several tests before drawing a conclusion.

Congestion and the time of day

M-Lab lists three possible causes for poor performance: Wi-Fi signal problems, congestion inside the ISP’s network, and congestion between ISPs. Ofcom’s codes measure speed at peak times, defined as 8 to 10pm for residential customers, and Ofcom advises customers that speed varies by day and time, so a fuller picture needs more than one test. Telkomsel’s guide for customers says the same in Indonesian: test several times and record results at different times, because morning and night results can differ (Telkomsel: how to test internet speed).

A national figure is not any customer’s plan. Ookla’s August 2026 index puts Indonesia’s fixed broadband median at 51.40 Mbps download, 36.23 Mbps upload and 8 ms latency, 113th globally. Komdigi’s 2026 target is a 45 Mbps average for fixed broadband, up from 37 Mbps in 2025 (Bisnis.com, January 2026). The two figures come from different sources and periods, so do not compare them with each other or with a household’s result.

Where the limit can sit, what you see, and who can fix it
Where What you tend to see Who can usually fix it
Phone or laptop One device slow, others fine Customer: update, restart, try another device
Home Wi-Fi Cable near plan, Wi-Fi far below, worse in far rooms Customer with ISP advice: position, band, router
Router, ONT and drop Low on cable at every hour, on every server ISP: line and port tests, technician visit
Shared access network Cable result falls in the evening peak ISP: capacity review
Upstream path and test server Low on one server, fine on another Test provider or ISP peering team

How do you prove ISP network performance with a standard troubleshooting process?

Use the same wired-first test ladder every time, and write the results down. The step where the number falls tells you whether to look inside the home or at the access network. Ofcom’s customer advice follows the same logic: run a speed test before changing anything, try a few checkers, test at different times, use a cable where practical, check whether one device is slower than the others, and reverse recent changes to the home network.

A five-step ladder for testing a slow connectionFive boxes in sequence. Step 1: connect a laptop by cable to the router and pause other apps. Step 2: run three tests on two different servers and note the times. Step 3: repeat in the evening peak and in the morning. Step 4: test on Wi-Fi beside the router and then far away. Step 5: find the step where the number falls and record it. A green bar under steps 1 to 3 marks the ISP side and an amber bar under step 4 marks the home side. The boxes highlight in order.A five-step test ladder: wired first, Wi-Fi second1 Wire itLaptop on a cableto the router.Pause other apps.2 RepeatRun three tests,two differentservers. Note time.3 PeakRepeat in theevening peak andin the morning.4 Add Wi-FiThen test on Wi-Fibeside the router,then far away.5 CompareFind the stepwhere the numberfalls. Record it.Steps 1 to 3 test the ISP side: no Wi-Fi in the wayStep 4: the home sideSame device, room and servers for every run. Change one thing at a time and note the time of each test.Ofcom: test before changing anything, try a few checkers, test at different times, use a cable where practical.A five-step ladder for testing a slow connectionFive boxes in sequence. Step 1: connect a laptop by cable to the router and pause other apps. Step 2: run three tests on two different servers and note the times. Step 3: repeat in the evening peak and in the morning. Step 4: test on Wi-Fi beside the router and then far away. Step 5: find the step where the number falls and record it. A green bar under steps 1 to 3 marks the ISP side and an amber bar under step 4 marks the home side. The boxes highlight in order.A five-step test ladder1 Wire itISP sideLaptop on a cable to the router.Pause other apps and downloads.2 RepeatISP sideRun three tests on two different servers.Note the time of each one.3 PeakISP sideRepeat in the evening peak and the morning.Same laptop, same cable.4 Add Wi-FiHome sideTest on Wi-Fi beside the router, then inthe far room. Same device.5 CompareDecisionFind the step where the number falls.Record it with the time.Same device, room and servers for every run.Change one thing at a time; note the time.Ofcom: test before changing anything, try a fewcheckers, test at different times, use a cable.

Figure 5: Wired steps come first because they take the home Wi-Fi out of the measurement. The step where the number falls tells you which side to investigate.
  1. Wire it. Connect one laptop to the router by Ethernet cable and pause other downloads and streams. A cable that cannot carry the plan speed will mislead the test, so use a cable rated for it, as Ofcom advises.
  2. Repeat it. Run three tests on two different servers and note the time. A single test is a sample.
  3. Test at peak and off-peak. Repeat in the evening, between 8 and 10pm, and again in the morning. Keep the device, the cable and the servers the same.
  4. Add Wi-Fi. Test on Wi-Fi next to the router on 5 GHz, then in the room where the problem is felt. Change one thing at a time.
  5. Compare and record. Find the step where the number falls. Record date, device, connection, server, download, upload and latency.
6 weeksin the UK before a customer can use a free dispute scheme if the provider’s complaint process has not solved the problemOfcom, updated Aug 2026
30 daysa UK provider may have to improve speeds before the customer gets a penalty-free exit, under the 2019 codesComms Ombudsman
45 MbpsKomdigi’s 2026 target for average fixed broadband speed, from 37 Mbps in 2025Bisnis.com, Jan 2026
92.4%precision reported by a 2024 master’s thesis for classifying whether a Wi-Fi test reflects a broadband bottleneckUniversity of Notre Dame

Reading the pattern

The results fall into a few recognisable patterns. A cable result near the plan with a low Wi-Fi result points to the home. A cable result that is low at every hour and on every server points to the line or the access side, and justifies a fault ticket. A cable result that falls only in the evening points to congestion. A result that is low on one server and fine on another points to the path or the server. Cloudflare adds one more signal: a gap between idle and loaded latency approaching 100 ms or more over a wired connection suggests the ISP is at fault.

Four test patterns and what each one suggestsFour rows, each with a pattern, a likely side and a first step. Cable near plan but Wi-Fi well below points to a home Wi-Fi limit: reposition, change band or upgrade the router. Cable low at every hour and server points to a line or access fault: open a fault ticket with the logged results. Cable low only in the evening points to congestion at peak: escalate for a capacity review. Cable low on one test server only points to a path or server issue: retest on another server. Each row highlights in turn.Reading the pattern: where to look firstPATTERN YOU SEELIKELY SIDEFIRST STEPCable near plan, Wi-Fi well belowHome Wi-Fi limitReposition, change band,or upgrade the router.Cable low at every hour and serverLine or access faultOpen a fault ticket withthe logged results.Cable low only in the eveningCongestion at peakEscalate for a capacityreview, with the times.Cable low on one test server onlyPath or server issueRetest on another server;check the upstream path.These patterns point the investigation; they do not prove a cause. Confirm with the ISP’s own line tests and the device logs.If one device is much slower than the others, Ofcom says the problem might be with that device rather than the broadband.Four test patterns and what each one suggestsFour rows, each with a pattern, a likely side and a first step. Cable near plan but Wi-Fi well below points to a home Wi-Fi limit: reposition, change band or upgrade the router. Cable low at every hour and server points to a line or access fault: open a fault ticket with the logged results. Cable low only in the evening points to congestion at peak: escalate for a capacity review. Cable low on one test server only points to a path or server issue: retest on another server. Each row highlights in turn.Reading the patternPATTERNCable near plan, Wi-Fi well belowLIKELY SIDEHome Wi-Fi limitFIRST STEPReposition, change band,or upgrade the router.PATTERNCable low at every hour and serverLIKELY SIDELine or access faultFIRST STEPOpen a fault ticket withthe logged results.PATTERNCable low only in the eveningLIKELY SIDECongestion at peakFIRST STEPEscalate for a capacityreview, with the times.PATTERNCable low on one test server onlyLIKELY SIDEPath or server issueFIRST STEPRetest on another server;check the upstream path.These patterns point the investigation; they do notprove a cause. Confirm with the ISP’s line tests.

Figure 6: Four patterns and the first place to look for each. This is a triage guide: it separates in-home limits from access-network problems, and it does not replace the ISP’s own line tests.

These patterns are a triage guide, not proof. The Notre Dame thesis shows researchers are working on exactly this question, telling a Wi-Fi limit from a broadband bottleneck, and it reports a classifier with 92.4% precision. The abstract gives no sample size and the work is a master’s thesis, so we cite it as a sign the problem is real, not as a method to copy. For a customer conversation, the wired ladder is simpler and easier to explain.

What to do when the pattern points to the ISP

Ofcom says the provider is responsible for fixing faults even if another company owns the network, and asks customers to say when the problem started, whether it is constant or intermittent, and what they have tried. Give support the same fields so the ticket arrives complete. In the UK, voluntary speed codes also set a minimum guaranteed speed and a right to exit if speed regularly falls below it, with measurement at peak times (Comms Ombudsman on the 2019 codes). We did not find a published Indonesian rule on how a residential speed complaint must be tested, so write down your own method and publish it.

A good support script holds to one question, one number and one next step, then points to a reference page for detail. How to Educate Complex Buyers Without Overwhelming Them describes that method for technical buyers, and it works the same way for a customer on the phone.

A support reply, written two ways (illustrative)

TypicalPlease restart your router and run the test again. If the issue continues, a technician will be scheduled.
Customer-readyLet’s test in order. First plug a laptop into the router with a cable, close other apps and run the test three times. Then run it again tonight between 8 and 10pm. Send us both sets of numbers and the times. If the cable results are well below your plan at every hour, we will open a fault with your results attached. If they are near your plan, we will look at your Wi-Fi next.

What should a speed test explanation include before you publish it?

It should say what a test measures, what it cannot measure, and what the customer does next. Use this list to review a help article, a support script, a plan page note or a chatbot answer.

  • Does it say the test measures the slowest link on the path, not the plan speed at the router?
  • Does it tell the customer to test with a cable first, and say how?
  • Does it explain why Wi-Fi results are lower, with distance, band and device as reasons?
  • Does it say that server, streams and time of day change the number, and ask for more than one test?
  • Does it name the peak window and ask for a test inside and outside it?
  • Does it show what the support team will do with each result pattern?
  • Does it list what to send to support: time, device, connection, server and results?
  • Does every number carry a date and a source, and is there a review date?
  • Who writes it, and which network and support owners check it before it goes live?

Speed test results lower than the plan: frequently asked questions

Why is my speed test result lower than my plan speed?

A speed test measures the slowest point between your device and the test server, not the speed written in your contract. The usual causes are Wi-Fi limits such as distance, walls, band and the phone itself, the test server or test design, and busy periods on shared networks. A wired test from a laptop to the router removes Wi-Fi first.

Why is Wi-Fi slower than a cable in a speed test?

Ofcom says a wired connection is usually faster and more reliable. Wi-Fi is shared, sends or receives but not both at once, and weakens with distance and obstacles, and phones often use fewer streams than the router has. One vendor guide puts good real-world throughput at 150 to 400 Mbps on 5 GHz and 20 to 40 Mbps on 2.4 GHz for a three-stream access point.

Which speed test result should I trust?

Trust a pattern, not a single number. Ofcom advises trying a few checkers and testing more than once at different times. Use the same device, cable and server for each run, note the time, and compare wired results before Wi-Fi results.

Does the speed test server location matter?

Yes. Cloudflare says results differ noticeably with the distance to the server, and test design such as the number of streams and how samples are aggregated also changes outcomes. M-Lab notes that many popular tests are hosted inside ISP networks and anecdotally report higher results than its servers, which sit outside ISP last-mile networks.

Why are my speeds lower in the evening?

Shared parts of the network and the wider internet path are busier at peak. Ofcom’s 2019 broadband speed codes define residential peak time as 8 to 10pm, and M-Lab lists congestion inside an ISP network or between ISPs as a possible cause of poor performance. Test in the evening and the morning and compare.

How do I tell whether the problem is my Wi-Fi or my ISP?

Test with a laptop on an Ethernet cable plugged into the router while other devices are idle, then repeat on Wi-Fi in the same spot. If the cable result is near the plan and Wi-Fi is far below it, look at Wi-Fi. If the cable result is low at all hours and on several servers, report a possible fault to the provider with your logged results.

Is it normal to get less than the plan speed in a test?

Some gap between a test result and the plan is normal because of test design, server distance and shared capacity. How large a gap is acceptable depends on the contract, which may state an up to, typical or minimum speed. In the UK, voluntary codes give customers a right to exit if speed regularly falls below a minimum guaranteed level. Check your own terms.

What is latency under load?

Loaded latency is the round-trip time while your connection is busy, compared with idle latency measured when nothing is transferring. Cloudflare’s test records both. Cloudflare says a gap approaching 100 milliseconds or more over a wired connection suggests the ISP is at fault, which makes it a useful extra number for video call and gaming complaints.

What should support staff collect before escalating a slow speed complaint?

Date and time, the device and how it was connected (cable or Wi-Fi band), the test tool and server, download, upload and latency results, results at peak and off-peak, and what has already been tried. Ofcom’s advice to customers is to say when the problem started, whether it is constant or intermittent, and what they have tried.

Sources

  1. Ofcom: Get help with broadband and Wi-Fi problems (published 13 October 2023, updated 25 August 2026)
  2. Measurement Lab: How fast is my Internet? Speed tests, accuracy, NDT and M-Lab (26 March 2019)
  3. Cloudflare blog: How does Cloudflare’s speed test really work?
  4. ISPreview: Ookla study finds UK home Wi-Fi speeds lag wired Ethernet (17 May 2023)
  5. Snapone Araknis help: Understanding real-world Wi-Fi throughput
  6. Communications Ombudsman: Ofcom update on broadband speeds from 1 March 2019
  7. Bisnis.com: Target Komdigi 2026, rerata kecepatan seluler 60 Mbps, fixed broadband 45 Mbps (5 January 2026)
  8. Bisnis.com: Peringkat kecepatan internet RI merangkak di global Agustus 2026 (14 October 2026)
  9. Telkomsel Jelajah: Cara tes kecepatan internet WiFi dan paket data dengan mudah
  10. University of Notre Dame: Robust determination of WiFi throughput tests being indicative of broadband bottlenecks (master’s thesis, 2024)

Figures 1 to 6, the scripts, the checklist and the pattern table are our illustrations, not published standards; the bars in Figures 1 and 4 are schematic. The Snapone ranges describe one three-stream Wi-Fi 5 access point from a vendor guide, and the 867 Mbps figure is our arithmetic (two thirds of 1,300). The 29.8% Ookla figure is from ISPreview’s report of Q1 2023 Speedtest.net results, and ISPreview itself notes the tests record no device, room or distance; we could not open Ookla’s own methodology page, so we do not describe it. The M-Lab post is from 2019 and M-Lab’s tools may have changed since; it is cited for the single-stream versus multi-stream principle and for server placement. The Cloudflare 100 ms point is Cloudflare’s own guidance. The Ofcom codes of 2019 are voluntary and signed by a limited list of providers, and are used as an example of structure, not as an Indonesian rule. The Notre Dame item is a master’s thesis whose abstract gives no sample size. The Komdigi target and the Ookla index come from different sources and periods and are not comparable. Telkomsel’s guide is general consumer advice and does not discuss wired testing. We found no published Indonesian rule on how a residential speed complaint must be tested. The in-our-view statement about where limits most often sit is not a measured share.