Malaysia holds a provider to a percentage of the speed you subscribed to, and collects the results every quarter. Thailand sets a comparable percentage against the speed you were advertised, then leaves that one indicator out of the quarterly reporting that covers everything around it.
Those are the two countries in the region whose speed rules attach to your own plan. The others regulate something adjacent, or nothing. So whether a slow line is a complaint or only a disappointment depends on which country you are sitting in.
What each regulator actually requires
| Country | The obligation | Tied to your plan? |
|---|---|---|
| Malaysia | Throughput at a percentage of the subscribed speed | Yes |
| Singapore | Publish a typical speed beside the maximum | No |
| Philippines | State a minimum speed and meet it 80% of the time | No |
| Indonesia | An absolute 4G floor; nothing at all for fixed lines | No |
| Thailand | 70% of the advertised speed, checked only on complaint | Yes |
Malaysia measures what it requires
A commission determination in force since August 2021 sets throughput at not less than 70% of the subscribed broadband speed for 90% of the time on DSL, and not less than 90% of the subscribed speed for 90% of the time on fibre.
Those figures are expressed against what the customer bought rather than against an absolute, and — unlike Thailand's — they are collected. The same determination sets latency ceilings of 50 milliseconds on fibre and 85 on DSL at 95% of the time, packet loss at half a per cent and one per cent, and network utilisation at 70%. Operators report quarterly, across at least thirty locations, under a declaration signed by the chief executive.
By contrast, the wireless determination that followed in 2024 contains no reference to subscribed or advertised speed at all. Its floors are absolute — a download of at least 10 megabits per second per location from January 2025, and a state average of at least 50.
Malaysia's general consumer code, remade in 2022, carries a full chapter on advertising and no speed provision whatsoever; the word appears once in the document, describing the speedy handling of complaints.
What Singapore makes providers publish
Since April 2012 Singaporean providers have had to publish a typical speed for each plan alongside the theoretical maximum, prominently, on websites and brochures and in advertising, at every point of sale, re-measured at least quarterly, with the methodology published. Providers above a market-share threshold must gather at least two hundred data points per plan per quarter, at peak and off-peak.
The paper introducing that requirement addresses "up to" advertising in its first paragraph. It did not ban the phrase. It required a second number beside it.
The document is candid about a central gap. The regulator does not prescribe how the typical figure is computed, so a provider may publish a mean, a median, or a range. Two plans with the same "typical speed" might be describing different statistics, making the numbers honest per provider but not comparable between them.
Singapore's separate quality-of-service framework, which does carry penalties of up to fifty thousand dollars per standard per month, governs availability above 99.9%, local latency at or below 30 milliseconds, international latency at or below 300, and utilisation below 90%. It sets no speed standard at all, and the quarterly figures the regulator publishes per operator are availability and latency — never measured throughput.
The Philippine floor, and the eighty per cent that is not what it sounds like
A 2011 memorandum order requires providers to specify the minimum speed, the service reliability and the rates in advertisements, flyers, brochures and service agreements, with minimum service reliability of 80%.
The 80% is often read as delivering 80% of the advertised speed. It is measured against the provider's own declared minimum — a number the provider chooses. Declare a low enough floor and the reliability obligation is easy to meet, which is why the rule is better described as a disclosure requirement with a consistency test attached than as a performance guarantee.
A 2015 circular added measurement by the regulator itself, twice weekly at peak and off-peak from known and unknown locations, published within five days of month end. It sets no numeric threshold; it produces data rather than a bar to clear.
Indonesia's only speed floor is mobile
Indonesia's arrangement is the most layered of the five. The ministerial service-quality regulations of 2011 to 2014 were revoked wholesale in 2019, and the content was moved down a level to Director-General regulations — which is why looking for an enacted ministerial standard on internet access finds nothing.
Those Director-General rules do govern internet access. A 2021 instrument sets packet loss at or below 5%, requires at least 90% of latency samples to come in under 250 milliseconds, and puts fixed network availability at 99% or better. It also sets a download success rate of at least 80% — which sounds like a speed standard and is not one. The instrument defines it as the percentage of download attempts that succeed, measured with a test file of at least ten megabytes. It records whether the transfer completed, never how fast.
A binding speed figure does exist, introduced in October 2024 and in force since roughly the following January. It requires 4G download throughput of at least 3 megabits per second across 128 named districts and cities, and at least 500 kilobits elsewhere, counted as met when half the monthly samples clear it on a median method.
The consequences for a subscriber are that the floor is mobile-only — there is no throughput standard for fixed broadband — and that it is absolute rather than relative. At 3 megabits it sits so far below advertised plan speeds that it works as a coverage backstop, not a check on the speed sold.
So a subscriber on a 100 megabit package receiving eight has no quality instrument to point at. The provider's obligations are satisfied as long as packet loss, latency and availability hold — all of which a heavily oversubscribed line can meet comfortably.
The national consumer protection body identified this gap. In October 2020 it formally urged the ministry to regulate the accuracy of internet speeds delivered to consumers. Six years on, no instrument implementing that has been enacted.
Thailand's speed rule is not collected
Thailand's quality-of-service notification, in force since January 2023, governs fixed and mobile services in one instrument. Its fixed-broadband section requires the average transfer speed to be not less than 70% of the service speed the licensee has advertised or notified to the subscriber, assessed against the busiest 95% of usage — with the busy period defined as five in the afternoon to nine in the evening.
The denominators differ. Malaysia measures against what you subscribed to; Thailand against what was advertised to you. In a market where the headline figure and the contracted figure can differ, that is not the same test.
But the rule's value is limited by a crucial detail. That speed indicator, along with the one for round-trip time, is expressly excluded from the quarterly reporting duty that applies to the rest of the annex. Both are reference indicators, applied when the regulator inspects a particular case or resolves a complaint.
The indicators around them are reported quarterly — installation within ten working days, faults below five per hundred subscribers, repair within twenty-four hours, billing complaints under 0.3%, a call centre answering within sixty seconds. The speed rule is the one the regulator does not routinely collect.
So a Thai subscriber has a number to cite and no data stream behind it, where a Malaysian subscriber has both. One further caution: the regulator's own page lists a second amendment to this notification whose text could not be retrieved, so the 70% figure is the 2022 baseline and may since have moved.
The delivered gap is smaller than the grievance
The best public data on delivered speed does not show providers systematically falling short. In the American regulator's fixed broadband programme, eight of twelve provider-and-technology combinations delivered 100% or more of advertised download speed during peak hours, and the remaining four delivered between 86 and 90%. On a stricter test — the minimum percentage achieved by at least 80% of subscribers at least 80% of peak time — DSL was the weakest at 63 to 72%.
Date that carefully before leaning on it. Those measurements were taken in September and October 2022 and published in 2024, and the programme has since stopped collecting: the regulator's own page says so, and its partnership with the measurement contractor ended in August 2023. It is the best public dataset of its kind and it is also the last one.
Read the methodology before carrying that across, though. Those figures come from a dedicated measuring device at the subscriber's router talking to a nearby test server, which isolates the access line and excludes the long path an actual download takes. It measures what the provider controls, which is the fair thing to hold them to and not the same thing as what you experience.
Why your own speed test disagrees with itself
One connection can produce several true numbers at once. Measuring a single machine on our own network gave a negotiated radio rate of 206 megabits per second, 152.8 across eight concurrent transfers, 116.5 from a fit across payload sizes, and 78.1 on one steady stream.
All four describe the same link. The radio figure is the hop to the router, not the line. The concurrency figure is why consumer speed tests open several streams — and why a single-stream tool reporting a shortfall is weak evidence of one.
The arithmetic of data transfer is consistent. Eight bits to the byte, so a gigabit line tops out at 125 megabytes per second. Ethernet framing then takes about five per cent: the standard test framework counts 1,538 bytes on the wire for every 1,500 of payload, which caps a gigabit link at roughly 949 megabits. Britain's regulator gives the same figure independently, putting the practical ceiling of a gigabit interface at about 940.
Another apparent loss comes from mislabeled units. A download manager that computes in mebibytes but prints the label "MB" will show a number about 4.6% lower than the true decimal figure, because a mebibyte is the larger unit. Multiply that displayed figure back up by eight and a gigabit line appears to be delivering around 905 megabits — an apparent shortfall of about nine and a half per cent, of which roughly half is framing and roughly half is a mislabelled unit.
Much of the speed people believe they are missing was never there to begin with. We have worked through the rest of that arithmetic in our guides on latency versus bandwidth and how much speed you actually need.
What is worth checking on your own bill
In Malaysia, the subscribed speed is the reference point, so persistent delivery below 90% of it on fibre is a measurable departure from a published standard rather than a complaint about the weather. In Thailand the equivalent number is 70% of what was advertised — worth quoting in a complaint, though nobody is collecting the evidence for you.
In Singapore, the typical speed is the number to look for, and the question to ask a provider is which statistic it is — because the regulator does not say.
Elsewhere the honest answer is that the number on the advertisement is the number you have, with a declared floor in the Philippines and, in Indonesia, a rule about how quickly someone picks up the phone.
Instruments and measurements behind this
The Malaysian obligations are commission determination number 1 of 2021 on wired broadband, in force from 1 August 2021, and determination number 2 of 2023 on wireless, in operation from 1 April 2024, both read as published determinations; the general consumer code is its second edition, registered in November 2022. The Singapore requirement is the publication requirement for internet service providers, issued in January 2012 and effective that April, with the separate quality-of-service framework for retail fibre broadband. The Thai provisions are the national broadcasting and telecommunications commission's notification on the standard of quality of telecommunications service, published in the Royal Gazette of 26 December 2022 and in force from the following January, together with its measurement guideline of 29 December 2022. The Philippine instruments are memorandum order 07-07-2011 and memorandum circular 07-08-2015. The Indonesian position rests on the enacted service-quality regulations listed in the ministry's own legal database. The delivered-speed figures are from the American regulator's thirteenth fixed broadband measurement report, and the unit arithmetic is computed here from standard frame sizes. All were read on 20 September 2026.
The Thai documents need their provenance stated. Every route to the regulator's own site and to the Royal Gazette is behind an automated challenge that the research declined to work around, so the notification and its annex were read from archived captures of the regulator's own files — copies of the primary instrument rather than commentary about it, which is a weaker position than reading it live and a stronger one than paraphrasing a law firm. The regulator's page also lists a second amendment whose text is not retrievable by either route, so the 70% figure is the 2022 baseline.
One further limit is that there is essentially no current regulator-measured speed data for the region. Singapore publishes availability and latency but not speed. The most recent Malaysian per-operator throughput report concentrated deliberately on locations where subscribers had complained, so it cannot be read as a national average. The Philippine measurement portal would not serve its own pages. Figures circulating for regional median speeds come from a commercial crowdsourced index — a legitimate source provided it is labelled as one, and not a regulator's measurement.
