The complete UK IPTV test: what survived a Saturday, and what quietly fell apart

The test ran for nine weeks between November 2026 and January 2027. Twelve services were chosen to cover the range a UK buyer actually meets: four large resellers advertising heavily on search, three small operations recommended by readers, two bargain services under £5 a month, two "premium" resellers above £20 a month, and our own service, tested the same way as everything else because the alternative is a report nobody should trust.

Each service was trialled or purchased for the shortest term that allowed at least two full weekends of observation. Where a cardless trial existed, the trial was used first and the paid plan second, so we could see whether the trial and the real product differed — which, for two providers, they did.

Every test stream was logged with the following data: service, channel, date, kick-off time, whether the stream loaded, seconds to first frame, buffering events per 90 minutes, average bitrate sampled three times, observed resolution, and whether the picture recovered without manual intervention. Channel changes were timed with a stopwatch because every provider's own claim about "instant zapping" turned out to be negotiable.

The hardware, because it matters

Streams were watched on a Fire TV Stick 4K Max (wired via adapter), an LG C2 with the set's own app, an Nvidia Shield for the heavier 4K work, and an iPhone for the mobile check. Identical player software was used across services wherever possible — three of the twelve only worked with their own branded player or a specific portal, which is itself a finding worth noting.

The four broadband lines deliberately included one congested evening line in a shared building, because a report written entirely on a pristine 900 Mbps connection describes a life most readers do not have. Roughly a fifth of all sessions were run on the worst line, and those sessions are identified separately in the results.

What we could not control

Three limitations, stated plainly. First, provider catalogues change weekly, so a channel that existed at the start of the test may not exist now. Second, two services refused a trial without a card, so they were bought outright for one month — which is itself part of the price comparison. Third, our own service is run by the people writing this, and we have tried to compensate by publishing the failures rather than omitting them; you are entitled to discount our own score accordingly.

How the scores work

Each service was scored out of five on seven criteria — stability, sport coverage, picture quality, EPG and catch-up, support, pricing honesty, and device flexibility. Scores are the median of three observers, not an average, which stops one extremely good or bad evening from dominating. The full criteria, and why each one earned its place, are below.

1. Stability at peak — weighted heaviest

Everything else is decoration if the picture dies at 3pm. We scored stability on dropped streams per session rather than "uptime", because uptime is a server metric and you are buying an evening. A service that claims 99.9% uptime and drops your match twice has failed, whatever the graph says. Our own target is zero drops across a full 90 minutes on a wired line at peak time, and one graceful recovery from a source-side glitch per month.

2. Sport coverage — including the matches nobody advertises

Sport is why most UK households look at IPTV at all, and it is also where rights packaging makes honest comparison genuinely difficult. A complete service needs Sky Sports and TNT Sports, the free-to-air competitions, and some way of covering the Saturday 3pm fixtures that UK television cannot show. Score it by fixtures available per season, not channels listed.

3. Picture quality — measured, not claimed

We sampled bitrate three times per stream and looked at the same three things every time: grass texture, crowd faces, and scoreboard edges on a paused frame. A 1080p feed at a healthy bitrate beats a "4K" feed at a starved one, and the label tells you nothing. More on this in the 4K section below.

4. EPG and catch-up — the quiet dealbreaker

A seven-day guide that is wrong is worse than no guide at all, because you plan around it. We checked 50 random programme slots per service and recorded whether the description matched what aired. The range was 38% to 94%, which is a scandal nobody talks about, and it is invisible until you have already paid.

5. Support — measured in minutes, not policies

We sent one pre-sales question and, later, one genuine fault report, and timed the first human reply. Anything above an hour during UK daytime is a warning sign, because the same rota will be the one handling your matchday problem at 8:05pm on a Tuesday. Nine days — the worst we recorded — arrived with an apology and an emoji.

6. Pricing honesty — the total, not the headline

Three questions decide this criterion. Is the standard rate printed next to the promotional one? Does the checkout add anything at the end? Is there auto-renewal you have to remember to stop? Two services in the test did not survive contact with the third question, and one advertised a price that became 40% higher at the payment page.

7. Device flexibility — no hostage-taking

If a service only runs in one branded app or a proprietary portal, you are renting a service inside a locked room. It also tends to mean the guide, the favourites and the search are all that provider's own design. The best providers let you bring your own player, and they tell you which one they recommend.

Saturday 3pm is the single most demanding moment in British streaming. Between 2:55 and 3:05, a country that has been waiting all week hits the same handful of feeds at once. Whatever is thin in a provider's setup becomes visible in those ten minutes — and the pattern repeats in the second half when anybody who missed kick-off catches up on delay.

Our protocol was deliberately dull. Fifteen minutes before kick-off we loaded the relevant channel and left it running. We logged time-to-first-frame, then watched the full match, noting any pause longer than two seconds, any audio desync, any resolution drop, and whether the stream recovered by itself. The same eleven fixtures were watched across services wherever rights allowed, so the comparison was as close to like-for-like as this industry permits.

The results were brutal. Nine of the twelve services dropped at least one stream during a Saturday afternoon session; four of those dropped three or more. The failures clustered in two patterns. Pattern one: a stream that loaded perfectly and then degraded — gradual resolution steps down, audio drifting ahead, then a hard stop somewhere around the 60th minute. Pattern two: a channel that simply would not open, spinning until we switched, which is usually a capacity or routing problem rather than a broken source.

Only three services completed every Saturday session without a manual reload: ours, Provider C, and Provider B (which dropped four streams elsewhere in the test, midweek, which is why it did not top the table). The practical lesson for a buyer is to run this exact test on a 24-hour trial before paying for anything. Providers that know their infrastructure will encourage it. Providers that do not will offer you a discount instead.

One more observation worth the ink: the Saturday failures did not correlate with the number of channels advertised. The service with 45,000 channels dropped more than the service with 6,000. Catalogue size is bought cheaply and weighted for search engines; capacity is bought with money and discipline.

A programme guide is the least glamorous part of television and the one you use most. Ours was tested by pulling 50 random channel slots per service and checking the listed programme against what was actually airing at that minute. No technical skill required — just patience and a notepad.

The spread was astonishing. Two services scored above 90% accuracy. Four sat between 70% and 90%, which in practice means one wrong programme in five and a guide you stop trusting by week two. Five scored under 60%, and the worst managed 38%, meaning most of what was displayed was wrong, outdated, or replaced by a placeholder reading "No information".

Why does this happen? Because a playlist and a guide are different products. Anyone can resell a channel list; populating seven days of accurate metadata for thousands of channels requires supplier relationships and, crucially, somebody whose job is to notice when it breaks. Providers who skip that step often have no favourite lists, no catch-up, and a search that cannot find a channel you can see on screen — the same underlying neglect showing up in three places.

What good looks like

Seven days of listings on every channel, populated within an hour of a schedule change, with catch-up markers where a replay exists. Favourites that persist across devices. A search that finds "Sky Sports F1" whether you type "f1", "formula" or "sky". None of that is exotic; it is simply the difference between maintaining a product and reselling a file.

Support is the part of a subscription you buy and hope never to use, and it is also the part that reveals how a provider is staffed. Two messages were sent to each service: a pre-sales question at 8pm on a weekday, and a genuine fault report during a live match the following weekend.

The pre-sales results divide neatly. Seven services answered within fifteen minutes, all of them small or medium operations with a visible human on WhatsApp or Telegram. Three took between forty minutes and two hours, and two never answered at all — one of which is still advertising heavily today.

The fault report is the revealing one. A matchday message sent at 8:05pm on a Tuesday was answered within ten minutes by three services, within an hour by three more, early the next morning by four, and never by two. Think about that in practical terms: at 8:05pm, your match is twenty minutes old and the reply arrives tomorrow morning. That is not a support process, it is a mailbox.

What "good" sounds like

A reply that names a cause, not a policy. "That feed is down, here is a backup on channel 412" is worth more than an apology and a ticket number. If the first response you ever get is generic, expect the same at the moment you care most. And if a provider cannot answer a pre-sales question promptly, the honest interpretation is that they have already been paid by enough people who did not ask.

1. "Free trial" that requires card details — the trial is a payment funnel with a clock on it.
2. Channel counts in the tens of thousands with no quality statement, no feed list and no way to search the catalogue before paying.
3. No published server locations. If capacity is a secret, capacity is a problem.
4. Support only via email, answered in days, with no acknowledgement of the fault itself.
5. Prices in crypto only, or prices quoted in another currency with "approximately" attached.
6. No EPG, or a guide that shows "No information" on most channels — usually a resold file nobody maintains.
7. Locked to one app or portal, so your player, search and favourites all belong to the provider.
8. A countdown timer that resets every time you reload the page, forever.
9. "Lifetime" terms. Not a business model that can be honoured, merely one that can be sold.
10. No stated company, address or trading history anywhere on the site — check the footer before the price list.
11. Pressure to pay outside a tracked channel — bank transfer or gift card requests are a hard stop.

There is a twelfth, subtler one. If a provider cannot answer "what happens when a route to your server congests?" with a sentence about another route, they have never thought about it. That single question predicted the stability ranking of this entire test better than any advert did.

Two instructions that make the log trustworthy. First, always test the same fixture slot on both services — a Saturday 3pm on one and a Wednesday noon on the other is not a comparison. Second, write the entry while you are still annoyed, not a week later when the memory has softened. Complaints are data; nostalgia is marketing.

What we would test next, given the time

Three gaps we would close in a second round. Longevity: a twelve-month follow-up on the four best services, because a good November says nothing about the following February. Scale: how each service behaves when four devices in one house stream at once, which is the reality in shared student and family homes. And churn: how quickly a provider responds when you ask to stop paying, which is where the friendliest sales channel often turns cold.

We would also like to test a genuinely reproducible 4K ladder — the same thirty seconds of a UHD broadcast sampled on every device and every service — because "real 4K" is currently a claim made by everybody and verified by almost nobody. If you want to run that yourself, the paused-frame method in our 4K guide is the whole procedure; you just need to repeat it consistently and write it down.