Thursday, 8 August 2013

Up to

Another article on "up to" speeds, and this is not the first time I have had a rant on this. It really annoys the hell out of me.

The problem!

The problem is that some end users do not quite understand what is being offered by ISPs, and feel aggrieved that the service they get is somehow not what was advertised, so they complain to ASA and/or OFCOM.

Lets take a simple case of ADSL2+, which can sync at 24Mb/s ATM rate. ISPs used to advertised as "Up to 24Mb/s".

The solution to this problem, as dictated by the likes of the ASA, is ISPs advertising "Up to 16Mb/s" or "Up to 14Mb/s" instead.

What you are measuring?

It is, perhaps, more useful to say "Up to 21Mb/s" simply because the line is typically used for IP not ATM, and the overheads of ATM and IP mean an "IP data throughput of around 21Mb/s for an ATM sync of 24Mb/s". Even 21Mb/s is not necessarily the same as the stats on a TCP file transfer or web based speed test. That said, comms lines have traditionally (and therefore useful for comparison) been quoted at the bit rate. A 56k modem could do "Up to 56kb/s" at a raw bit rate - which was different to the IP rate. An ISDN channel is quoted as 64kb/s, not the IP rate. It is perhaps sensible to use a rate that is more closely related to the speed the end user will see when doing a file transfer though to avoid disappointment. You also have confusion of Mega (1000000) and Mebi (1048576) to add to the fun, but at least we have SI units well defined.

What do you mean "Up to"?

Even so, the issue here is clearly a simple misunderstanding.

I think the issue is that "Internet" is an "Up to" thing anyway - you may download no data some of the time, a bit of data slowly some times, or a big file quickly (depends on the other end). However, the "Limit" if your line sync speed. So people think an "Up to 21Mb/s" line is one that can actually do 21Mb/s and they can use any speed they like all the way up to that speed.

It is like selling a car that can go "Up to 100mph" - it will be able to achieve that speed if you have a suitable road on which to try it. If you bought a car that does "Up to 100mph" you would not expect it to be that some of the cars only do "Up to 50mph" and some do "Up to 100mph". You expect all of the cars to be able to do "Up to 100mph". So you expect all the ADSL2+ lines to be able to do "Up to 24Mb/s" in the same way.

The problem is some people somehow do not understand that the limit of their line will be what it is depending on the line quality and length. Their service may end up being what they might consider "Up to 5Mb/s" of data transfer possible, even though sold an "Up to 21Mb/s" service.

ISPs do try to make that clear, but the message is not clear to some.

I have had people, face to face, say that the line from some ISP was not the "Up to 21Mb/s" they advertised. They did not understand my objection to their statement. As far as they were concerned the line should be able to get 21Mb/s, at least some times, if sold as an "Up to 21Mb/s" line.

The solution?

First you have to decide if you need a solution - if a few people are confused by something that is very clearly stated, and those people have not lost out (i.e. that any provider selling ADSL2+ will get roughly the same speeds), well, is there actually a problem to solve?

Assuming there is a problem to solve - it is about getting some industry consistent message / wording for this. Something like "Up to a limit that may be up to 21Mb/s depending on where you live". Or just "Up to 21Mb depending on where you live", or something simple. Maybe "Up to a speed between 250kb/s and 24Mb/s". I don't know the actual answer on how to explain it to people. It seems that this should not be beyond the wit of a marketing person to find.

If we could find a good wording, that would help remove some of the disappointment and complaints.

The wrong solution?

What does not make sense is lowering the figure. The ASA rule that the figure has to be one 10% can get is just silly. It does not change the message, and does not aid comparison between ISPs. If one ISP does ADSL2+ and so does another, on the same BT copper pair, the speeds will be basically the same (there are some tweaks some ISPs can do better than others, but that is not easy for comparison).

All this solution does is reduce complains by exactly 10%. It means that 10% of customers will no longer feel they did not get what the were sold. The rest will still feel cheated somehow. Ironically, some of the 10% would (and have, AFAIK) complain that the ISP lied to them as they are clearly getting more than the "Up to" speed quoted.

In many ways it would be better if we went back to comparing ATM sync rates. That way every ADSL2+ service would be quoted as "Up to 24Mb/s" and every ADSL1 service as "Up to 8Mb/s". It would allow people to compare numbers between ISPs as that is easier than comparing "Is VDSL better than ADSL" or some such.

Right now we have a confusing mix of numbers, and someone may indeed think an ADSL2+ sold as "Up to 16Mb/s" is going to be better than and ADSL2+ sold as "Up to 14Mb/s". Is it really helpful for the ASA and OFCOM and consumers for people to make a decision based on how far an ISPs marketing department feel they can push things and get away with it?

What I fail to understand is why this wrong solution was proposed. Was there not one rational person in the discussions and meetings to say "This obviously does not address the problem". As director I am facing a marketplace where I have to work within rules made by people that cannot see the problem with this solution, or even with calling coax cable "fibre", and it makes my job somewhat difficult!

Sync is not everything

Ironically, one of the big differences between ISPs is not the advertised line sync rate. Yes, that allows one to see ADSL1, ADSL2+, FTTC, and so on as a key difference, but there are other factors. Some ISPs are good (and this is not just small or just big or just A&A even). Some ISPs make sure they run uncongested links, and monitor for faults and issues. This means the overall "experience" is much better. The problem is that this is very difficult to market as a feature, or to guarantee, or to have clear metrics to quote and compare. If there was a way to see these sorts of stats, loss, latency, congestion, and so on, then that would be a far more important and relevant factor in choosing an ISP, typically along side cost and terms and so on. A&A publish a report here.


  1. The problem with everyone advertising "up to" numbers is that they are pretty much meaningless for the end user. When I'm shopping for an internet connection then I don't care what the theoretical top speed I might get is if I lived right next door to the exchange, I care what speed _I'm_ going to get. This means that unless you can tell me what speed connection I should expect to receive, there's really no point in giving me this information at all.

    In fact, I'm inclined to say that quoting "up to" speeds isn't just useless - its detrimental. Someone who doesn't understand anything about the underlying technology is going to compare two ISPs, one of which says "Up to 14Mbps" and the other says "Up to 20Mpbs" and conclude that the "Up to 20Mbps" ISP is somehow going to be faster. Whichever one of these ISPs you choose, you'll have exactly the same speed because they will both use the same copper pair, the same DSLAM, the same backhaul, etc. The only difference is that the statistical measure of speeds across their whole customer base is different. So the quoted "up to" speed is actually misleading because it makes me think that my choice of ISP is actually going to affect the speed I get.

    Obviously there may be no sensible way to give people a realistic idea of what speeds they're going to see when advertising to them, since doing so would require extensive tailoring of the advertising to each recipient; but my argument is: if you can't give meaningful figures in marketing materials, don't give any figures at all.

    What makes the difference is the type of technology used: I know that all the ISPs using ADSL2+ are basically going to have the same sync speeds as each other, all the ISPs using FTTC are going to have the same sync speeds as each other and will generally be faster than ADSL2+, etc.

    Maybe a fairer way of doing things, which would reduce complaints, would be to charge people based on bandwidth - if someone wants a 24Mbps connection and they can actually achieve that speed, charge them more than someone who only wants 8Mbps (or can only achieve 8Mbps). If people are told "I'm sorry we can't offer you the 24Mbps product, but you can have the cheaper 8Mbps one" then they are probably less likely to feel hard done by than someone who buys a 24Mbps product and isn't compensated for the fact that they're never going to achieve anywhere close to 24Mbps. (FWIW, I'm on an "up to 24Mbps" connection and I only get a sync speed of about 9Mbps even though I'm not especially far from the exchange, so we're not talking about people falling slightly short of the advertised speed, we're talking about people falling _far_ short of what was advertised). Also I guess if BT got paid less for the crapper local loops they might be more inclined to replace them...

    And yes, I agree entirely that sync isn't everything; but unfortunately the general public aren't capable of using a large number of metrics to compare products. Raw speed is a simple number that the public can understand and make cost/value assessments of (however meaningless this assessment might actually be!), whereas the other metrics you mention are far less tangible. I guess its like digital cameras - many people shop for the most megapixels per pound, even though the number of pixels a camera has is a relatively minor point compared to quality of optics, etc. I've certainly had trouble, on occasion, convincing some *business* customers why they should replace their £5/month talk-talk ADSL with a more sensible ISP (and this is never because of speed, always because their support people are completely useless).

    So I guess to summarise - ISPs should stop marketing based on meaningless "up to" sync speed metrics and figure out a way to market based on the performance of the underlying technology, and less tangible benefits.

    1. Some of these points are spot on - and hence my comment that if you want "numbers" which people can compare, then the same ATM rate numbers, e.g. 8M for ADSL1, 24M for ADSL2+, 40 or 80M for FTTC, and so on, do make a number people can compare as a proxy for comparing the technology. The current mess where the same technology may be sold as 14M, 16M, 21M, or 24M for exactly the same thing is crazy and directly caused by the ASA ruling. They have made matters WORSE.

    2. ASA ruling or not, I still don't think that the "up to" speeds are a sensible thing to publish, because it doesn't make it at all clear that the ISP A offering "up to 24Mbps" and ISP B offering "up to 24Mbps" are actually using exactly the same equipment and therefore switching from one to the other won't make a difference.

      As it stands, if a not-so-technical customer buys an "up to 24Mbps" internet connection and only gets a sync speed of 5Mbps, they will feel that the ISP is to blame, bad-mouth them to all their friends and switch to another "up to 24Mbps" ISP, expecting them to be different. There is no visibility to the customers that the underlying shared infrastructure is to blame rather than the ISP themselves and therefore (without taking expert advice) the customer isn't to know that they actually need to switch _technology_ in order to achieve better sync speeds.

      Maybe, rather than quoting raw "up to" speeds, ISPs need to agree on a classification system that will allow people to compare different technologies without focussing on absolute numbers (which can never be individualised). For example, ADSL1 could be labelled "Class 1 broadband", ADSL2+ is "Class 2 broadband", FTTC is "Class 3 broadband", etc. So the customer knows that if they aren't happy with the sync speed they are getting from their current "class 2" connection they need to switch to a "class 3" connection.

      Of course, this gets messy in situations where you can't make much guarantee that a lower class would actually be slower - for example, how do you rank a 60Mbps Virgin Media connection against an 80Mbps FTTC connection? I imagine a good proportion of the FTTC connections actually sync lower than the Virgin connections, but similarly a good proportion of the FTTC connections are probably faster than Virgin - they're using completely different infrastructure so it would be very hard to make a generalised comparison.

      Hard questions with no trivial answers, but I'm pretty convinced the current marketing is confusing and detrimental to everyone who doesn't have the technical ken to know what to ignore and how to read between the lines :)

  2. By definition, a service sold as "up to 16Mb/s" that actually does 24 is not "as described". I think, if they can find a way to word it, it might be better to say "the maximum speed possible with this service is 24Mb/s, however the actual speed available on any particular line may be less than this." - perhaps even coming up with an approximate figure for a customer. If an ADSL2+ service only gets, say, 7Mb/s for a particular customer, ADSL1 will be much worse. FTTC has the DSLAM much closer to the end user, so speeds closer to maximum are more likely.

    Really, they need to say "we cannot guarantee that you will ever get the maximum possible speed" as well.

  3. “… but we actually have 5% extra headroom on the BT side to ensure we are managing the traffic rather than BT.” – so, without that headroom, you'd be managing BT? ☺

    Anyway. Advertising is intended to confuse and distract, and they're doing exactly what's required to ensure that it becomes more so and are succeeding brilliantly at it. I suspect that a B Ark is the best way to sort this out…

  4. I don't know whether A&A does this, but the big ISPs all include small print about distance and line quality (I think they're obliged to), and the sales process begins with an individual estimate of what each customer can expect, so they can't accuse the ISP of mis-selling.

    What's interesting is that Sky only recently re-introduced an "up-to" speed figure into its marketing because it needed a way to compare ADSL with FTTC, beyond using nebulous terms like "superfast".

    Recombu Digital (which I edit) has done summaries of the Ofcom ADSL2+ and superfast data, including things like latency for the big ISPs:

    1. Quite - A&A do not generally quote speeds in the main marketing anyway - we do have a "technical statement" which explains how different technologies work to different possible speeds, but the "marketing statement" is that speed depends on your line and use the checked. Yes, we also do speed estimates as part of the signup even though not signatories to OFCOMs crazy code of practice on this.

  5. Unfortunately there is plenty of industries/services where A to B comparison (for the same product) is impossible. Energy is a typical example, whichever company you choose, you get the same electricity or gas coming using the same pipe (even if the source of energy used to replace the one you use will be different), but the same can (loosely) be said about internet transit traffic, not every ISP using and managing the backbone.
    So unless OFCOM comes in and force every ISP to publish a report like yours there never will be meaningful comparison (a bit like the train performance report - don't get me started how "retiming" and increased travel time makes it easier to beat the stats).

    But I'm personally not sure that the person not able to grasp the "Up to" (ATM Sync) will be more enlighten with 100 sec windows with more than 0,5% packet loss stats!
    What is really annoying and blatantly lying is the "Fibre" connection which is really a connection with a bit of fibre at some point. I don't really see at which point those technologies becomes fibre or not fibre: fibre to the exchange, fibre to the cabinet, fibre to the DOCSIS router. Anything but FTTH is copper with a bit of fibre.

    The same go for "really unlimited" which is "we don't apply any throttling policy" but does that mean without any congestion (even if BT is good at managing that)? I suspect, in motorway term, we don't use variable speed limit, but if you get stuck in traffic that's life.

    Could you imagine the commercial for DOCSIS operator: "Partially-fibre unlimited internet connection up to 100Mbits, on your first 100Gb during off-peak hours, 20Mbits after, only on non-congested hours", it probably describe the product more realistically, still doesn't give enough details on performance to compare with other providers.

    It's really hard, even the old contention rate is a difficult stats to handle. I should get a {real} fibre connection soon(ish) (hopefully with A&A) using Gigaclear FTTH product. Going direct, they got a quite substantial list of tariffs (enough to scare most users in my opinion), but I would hesitate between 1:25 or 1:100 contention rate on 1Gb/s connection (that mean in real term guaranteed 10Mbps or 40Mbps), but I think (and that's where I probably need to be corrected), that with such high connection speed, most users will have a burst of transfer but in real term will not have a sustained transfer above their contention speed, so the chance of getting close to burst speed is more likely, which in turn reduce the risk of congestion for others.
    It's not very clear, but let's say we have 25% of peoples watching a stream at 6 Mbps.
    On a contended 8Mbps, congestion will occurs if contention is anything above 1:(100/25*8/6) or roughly 1:5.33...
    On a contended 1000Mbs the contention rate for the same stream need to be 1:(100/25*1000/6) or 1:666 (or 1:333 if everyone switch to the 12Mbps HD stream)
    Of course the usage will vary with the pipe size, and increase with time (4K movies download etc...) but I don't think I'll more than triple my current usage in the next 3 years (from 100GB/Month to 300GB/Month in 36 months).

    Do you know any public statistic of the average usage volume per user (and ideally per hour/day) for the country? (I don't think A&A is representative due to the nature of its users base and the day/night time units). It would be interesting to know how much (as a nation) we use the internet and how much it progress.

    1. The only stats I have seen bandied about are something like the peak usage in a day for some high usage ISPs working out at an average of around 100 to 120kb/s per customer, but that was, I think, a verbal comment, and we (A&A) are somewhat higher than that already.

      I do like the "I suspect, in motorway term, we don't use variable speed limit, but if you get stuck in traffic that's life."

    2. "Anything but FTTH is copper with a bit of fibre"

      You could reasonably argue that if you're 2km from the exchange and 100m from the cabinet, then there's a lot more fibre than copper in an FTTC connection, so it's reasonable to describe it as "fibre".

      We've toyed with using the term "hybrid fibre" on Recombu Digital for both FTTC and cable, but in the end I doubted that consumers would be less confused, since no-one has ever used the term "copper" for DSL.

    3. So, lets see, you are 2km from exchange, and 100m from the cab, and say 5000km from a web server in the US. How much is fibre? 5002km and copper 100m. Now, if you are on ADSL, 5000km fibre and 2.1km copper... Even if you talk of a web server in London, you are looking at a *lot* more fibre than copper even on normal ADSL. So Fibre the the Exchange (i.e. normal ADSL which is copper from you to exchange then fibre) is as much "Fibre broadband" as any of the other solutions.

    4. Taking it to extremes, you could say that, but this market is about the connection between the customer and the exchange, or its equivalent in a cable provider's system. After that, the consumer's really paying for QoS features such as unlimited downloads, low contention, no traffic shaping, filtering (or not), etc, and the transport medium doesn't matter.

      My point is that I don't think it's unreasonable to talk about FTTC as a fibre broadband technology, but there's a dogma among FTTP purists that if you have any copper between the exchange and the customer, even if it's one per cent of the distance, then you can't call it fibre. I don't think it's constructive.

    5. Really?! This is not simply a matter of "FTTP purists" here. Some of the adverts have gone to great lengths to explain how copper slows things down and how they provide fibre broadband when in fact it is coax (i.e. copper). It is impressive what you can do with copper even when as long as a few hundred metres, I have to agree, but it is nothing compared to the capacity of a fibre. There is a huge difference between 0m of copper and 100m of copper, and given that you may actually be a lot more than 100m from a cab or node there is still the huge uncertainty and variability of copper services.

    6. Alex, I don't agree with you, the weakest link determine the performance, so the weakest link should determine the branding.
      I can correct "Anything but FTTH is copper with a bit of fibre" to "Anything but FTTH is a mix of fibre and copper, (severly) limited by the copper portion" if that makes you happy, but the fact is that the vast majority of subscriber to Virgin believe they subscribed to Fibre internet, and don't realise they got copper.

  6. Well now, there's a massive difference between Virgin's coax copper and BT's twisted pair copper, in terms of both bandwidth and attenuation, but that's a completely different discussion (after all, coax has the potential for Gigabit speeds).

    If we're talking about twisted pair and fibre hybrids, and I appreciate that there's a factor of 100 between the best capacity of a hybrid TPC-fibre compared to pure fibre, but I still find the dogma obstructive. As you say, the marketing isn't wrong about the difference between pure copper and FTTC, it's just missing the bigger picture in comparison to the huge leap you get with FTTP.

    Should the UK be investing billions in FTTP networks instead of millions in FTTC? Yes, but the ISPs didn't make that choice and they're selling what they've got, so it's hardly fair to blame them for they way they choose to market FTTC.

    Well, that's my ha'penny worth anyway.

  7. The problem with Virgin marketing "fibre optic broadband" is that they are just using meaningless hair-splitting jargon to muddy the waters and make their product sound fancier than it really is. They seem to think that they qualify as "fibre optic broadband" and ADSL doesn't because their fibre termination is *marginally* closer to the CPE.

    IMHO it's disingenuous to say "we use fibre, they use copper, and thats why ours is better" when *both* companies are using both fibre and copper. Its even more disingenuous to say this given that is was painfully obvious to everyone with some technical background that FTTP was always going to happen. None of this is about what technology has actually been used, its just about the marketing people making things sound cooler by twisting the truth.

    I half think that BT should start marketing FTTP as being better than Virgin's because FTTP "uses frickin' layzor beams" instead of just that old-hat fibre optic stuff that Virgin have. Technically its true - fibre optic communication does use lasers to drive the fibres, doesn't matter that virgin also use lasers for the same purpose - BT install them in your home! And after all, frickin' layzor beams sounds cooler than fibre optics, especially if they give every subscriber a free plush shark :)

    FWIW, in the US, AT&T seem to have similar problems since they have been marketing HSPA+ as "4G" for a few years - it just creates confusion in the market place because now you've got carriers who are *actually* offering 4G connectivity trying to figure out how on earth to market a new product that everyone thinks has been around for years because the old product's marketers essentially lied to make it sound better than it really was.

  8. Wow this is a touchy one, this is one of the few blog posts I will disagree with you. I fully understand what the isp's are saying, that the product sold will allow some customers to get that speed, and that merits the marketed speed. However I look at it from a sales and fit for purpose reasoning. If a company sells a product where only a low single digit amount of customers can achieve the product spec, that to me means the product is missold. The fact is hardly anyone can get 24mbit/sec on adsl2+. If I was an isp I would probably sell it at up to 12mbit/sec. I agree the ASA ruling was bad but for a different reason, I simpyl felt it was too soft and a let of to isp's. In reality they should have required a 51% able to achieve the speed as a bare minimum. The fact isp's would charge a fixed rate for what was a lottery of what your line could do really annoyed me with adsl, it is also a problem with FTTC, but not quite so bad. Since 40/10 services sold are typically cheaper than the 80/20 variants and the percentages of people able to get those speeds are more favourable than adsl2+. This is why your car analogy is wrongm every new car sold unless faulty can do 100mph if the road is right and no law existed with a lower speed limit, the same cannot be said for adsl lines. The very wording o "up to" suggests to a typical customer that they would at least some of the time be able to hit that speed so for that reason its misleading, even if the isp's dont see it that way, so to me its the isp's that dont understand they are misselling not the other way round. What really would have helped here is if there pricing based on burst speed like in the year 2000. So if your line couldnt handle 8mbit sec then just downgrade to a cheaper speed. The estimated speed given at point of sale is almost a good thing tho, except isp's are not legally obliged to honour that estimated speed which is another let off for them.

    1. Not I am not 100% sure you are disagreeing in all areas. The fact you say "The very wording of "up to" suggests to a typical customer that they would at least some of the time be able to hit that speed so for that reason its misleading". This is ABSOLUTELY not what ISPs are saying or trying to say, and is the whole problem. What they are trying to say is "a maximum speed depending on your line and location which may be as much as 24Mb/s in some locations", that is the very message ISPs are actually trying to convey and we need better wording for, but to say "typical" when "up to" is used is almost deliberately missing the point. The term "Up to" is very much stating the extreme - the maximum - the limit and absolutely not even trying to state "typical" in any way. I have no idea why people think a maximum (or a minimum) is any indication of "typical".

      My point is that ISPs do state the speed estimates for a line when you say the line or postcode, so the "up to" figures are only any use for making a simple numeric comparison of the technology. So saying ADSL2+ is "Up to 24Mb/s" is fine if everyone says that, and it is true if one person gets an ATM sync of 24Mb/s in reality (and many do). Changing that to 21, 16, or 14, does not help, and actually hinders as people thing "up to 16M ADSL2+" is somehow better than "Up to 14M ADSL2+" when they are in fact the same.

    2. On that point I agree with you, you are calling for consistency across adsl isp's. I am ok with that and it should be the case. Just I think telling customers a product is up to 24mbit/sec when barely anyone can achieve that is misleading, I would like to see a obligation to advertise "typical" speeds, and the speed should be decided by ofcom so all isp's advertise the same speed. This would still only take into account sync speeds so if a isp was bad at throughput maybe due to severe overselling then that wouldnt be evident on the marketed speed but this would be better than whats in place now. You are then free to say in the small print or over the phone that the product isnt capped at the typical speed but thats just what you expect the average customer to get.

    3. That is, indeed, one way it could be done. I suspect it would be fun getting anyone to agree with OFCOMs typical speeds. A 50th percentile of, say, all BT ADSL2+ sync speeds over the whole UK may be a good reference. Even the current 90th percentile ASA rule is open to careful usage - e.g. I could offer two services on ADSL1, one that is 8.128Mb/s sync, and sold as such, and one that is "typically 5Mb/s" or some such. The 8.128Mb/s being only available where lines are short enough. Both would be "Subject to availability, use line checker". In some cases the number of people able to get the top speed is higher than you think. ADSL1 has a large proportion on the top speed, and FTTC capped at 40M is able to get 40M in a large proportion of cases.

  9. I heard an advert on the radio the other day - I think it was for toothpaste - that said something like: "...up to 100% fewer cavities" ! The lunatics really have taken over - I give up...


    1. I know, that really gets me - it means NOTHING... And hard to prove I have fewer cavities "than I would have had"!

    2. But even the news reports are happy to quote mindless statistics... I heard on the news this evening that as I drink more than five cups of coffee a day, I'm 50% more likely to die. a) I'm certain to die, so how I can be 50% more likely to die is beyond me. b) Even if they meant 50% more likely to die within X years, unless they say how likely I am to die in the first place, the fact that I am 50% more likely to die is completely meaningless.