Broadband OFCOM voluntary code of practice on broadband speed
OFCOM have been meddling in the way ISPs sell broadband and the way that consumers see the speed of broadband services for some time and for some good reason. Consumers are understandably confused and the range of technologies deployed to provide faster and better internet access is increasing. We agree consumers need to be able to make informed choices. However we are concerned over the latest version of OFCOMs code of practice on broadband speeds to the extent that we have decided not to sign up and instead we have set out our own code of practice - our commitments on broadband speed.
The previous version of the code, to which we were a signatory, essentially said we had to tell people an estimated of speed, and if we did not stick to it we had to let change to a lower speed tariff at no penalty. Well duh, our services were, and mostly still are, not priced based on speed, so we could comply by saying stay on the same tariff if ever there was an issue. There was not an issue as we quote the BT speed checker results on ordering. So customers had informed choice.
The main issue with the code is No matter how fast our lines are, and no matter how consistent speeds are for similar lines, we will always have at least 1 in 10 lines that are below a minimum guaranteed access speed. What is worse is that speeds at or below this minimum can leave with no penalty, so the more consistent speeds we have the more customers will be at or below the minimum. Fixing these slow speeds or allowing them to leave does not reduce that number, it just moves the guaranteed minimum up to ensure at least 1 in 10 lines are below the minimum.
Even OFCOMs idea on speed estimates, a simple principle of telling people what to expect, is designed to ensure 2 out of 5 customers are outside the range quoted.
Of course if we have narrow definitions of similar lines so as to give the best speed estimates, and we use BRAS rate not sync rate as the access speed as it is more honest, the problem gets way worse. You could have a situation where the majority of similar lines are all on the same BRAS rate (as it goes in 1Mb/s steps when you get higher up) meaning the 10th percentile is actual the same as the 50th percentile making most lines at or below the minimum.
The fact many technologies have a top limit on sync also causes a big problem. Baring a few exceptional faulty lines which would be less than 10%, all lines below a certain line length on ADSL1 will be 8128Kb/s access speed (7.15M BRAS). That means the 10th percentile for short ADSL1 lines will be 8128K and so all of those lines will be at or below the minimum guaranteed speed. It is crazy that OFCOM should make a set of rules that define the very best lines we have as a problem!
The other key issue is OFCOM changing contract terms without imposing any similar changes to wholesale services. OFCOM make it that at any time 10% of lines can be ceased (within first 3 months of service or more) with no penalty. Yet BT have minimum terms (12 months on FTTC) and have cease charges. We would have to absorb these even though it is most likely the reason the customer is leaving is BTs fault. The result is higher prices for everyone, or higher up-front charges. It is hard to see how that is in the interests of consumers.
The use of percentiles means that the code is only sensible where 10% of lines are faulty and customers are not complaining. This is an active discouragement for any ISPs (such as us) doing and pro-active monitoring and contacting customers with problem lines. To follow the OFCOM code we need people with slow speeds to stay like that so that the 10th percentile stays sensibly low. This alone is reason not to sign up to the new code.
The one thing that would allow poorer ISPs to be shown up is that their speed estimates are also based on their existing lines and the spread of speeds. This just means they end up giving a wider range of speeds in estimates not necessarily lower maximum speeds. For most ISPs the speed estimates simply come from carriers such as BT and so are the same regardless.
Spirit
It is worth bearing in mind that OFCOM themselves talk of following the spirit of the code not the letter. If such major issues can be dismissed as following the letter of the code, then most of the wording in the code should be deleted and replaced with simple one-line statements of objectives. The spirit of this code is clearly to make speed issues relative to the ISPs current performance for similar lines - and that alone is a problem as it means all ISPs always have to consider a chunk of their lines as a problem no matter how good or bad the ISP is, and ISPs are clearly discouraged from pro-active monitoring and repair.
Speed
The first issue if that we are focusing on speed at all. OFCOM understand that speed is a complex issue, and depends on many factors some of which are not down to the ISP. Access speeds (the speed the line syncs at) have been easy to identify and so were the main factor when comparing lines. Even now the BT speed checker is only considering access speeds. Newer technologies mean that access speeds are not that simple. Seamless rate adaption affects DSL speeds, changing them from second to second. Contended local networks as used by cable operators can mean effective speed on a shared LAN between a consumer and the exchange can be complex and variable, and hard to measure.
As an ISP we have always been keen to offer a quality internet connection. The speed is not actually the key factor - it is as fast as the line goes. We obviously make efforts to resolve faults that make lines slow, but actually our concerns have been over the back-haul network over the likes of BT and the connection to BT and our network and transit and peering. We want to ensure we are not the bottleneck. This means providing the best service possible.
We look closely at the causes of slow speed - not just the sync rate but also packet loss and latency which can be caused by faults and congestion. The OFCOM code of practice does not address packet loss and latency at all.
OFCOM also make some definitions for speed, but then fail to use them everywhere they need to be.
- Headline speed is what the ISP advertises. We don't even advertise a headline speed any more. We do, buried in the technical details, specify the speeds the different technologies can manage in ideal connections, but we don't any more have a headline speed. At the end of the day all of the technologies have a more than adequate maximum (i.e. good enough to stream iPlayer, etc.) so quoting them is not really helpful when what actually matters is the speed an individual is likely to get.
- Access speed (e.g. DSL sync speed). This is confusing for a start as the sync is at an ATM layer usually, and consumers are more interested in IP throughput or more likely TCP throughput which is lower due to protocol overheads. OFCOM do not discuss this at all and don't define what speed they are measuring. Access line speeds are usually something you can test, although SRA is a potential issue.
- Actual throughput speed. This is the worst metric technically but what the consumers are looking at. Throughput from some website is affected by all sorts of factors outside the scope of the service the ISP offers or controls. It can also be affected by packet loss, latency, congestion, shaping and all sorts of things the ISP may control. But you cannot easily separate the ISP controlled factors from those that are not.
- Average speed. This makes no sense at all. Quoting this in any circumstances is likely to confuse consumers. It is probably still true that average usage of the internet is only around 64Kb/s per line. This is not a reflection of what the ISP is offering but of what people happen to actually use given unlimited resource. It is not clear how such a metric is useful to anyone. On a per ISP basis it will hugely depend on the mix of business and residential users and the number of torrent users. Its usefulness is only relevant when compared to ISP network capacity. It makes no sense to measure, let alone quote or compare such a metric. For example, as we charge for usage we will have a lower average throughput, which is good as we have capacity for more and are less congested. However, an ISP with huge bottlenecks and traffic shaping and loads of torrent users filling the pipes all the time will also have a low average throughput, which is bad. The metric alone tells you nothing. Oddly, having defined it, OFCOM don't seem to reference this metric.
- BRAS rate. Not mentioned by OFCOM. Following the spirit of the code this should be used rather than the line sync speed as the access speed. The end user access is immediately restricted to the BRAS rate at the BRAS in BT back-haul. However, OFCOMs definition of access speed is modem to exchange only which is silly.
Principle 1: Training
Makes perfect sense to have staff trained to understand the services they sell, the commitments we make, and offer the best information and advice for the individual customer.
Principle 2: Information
Makes perfect sense that we provide the best information we have at the point of sale. We don't believe speed estimates are by any means the only factor at the point of sale, and that detailing our commitments to the quality of the service are just as important. We expect that the speed estimates based on the BT/Be checkers will be basically the same for all ISPs (at least those using BT copper) and so not that helpful in differenciating services from one ISP to another - they are more of a fact of life than a different service offering. Needless to say we provide the details as we get from the availability checkers.
The talk in this and previous section about contracts being renewed is odd to us. We run ongoing contracts. They are renewed every month. An ongoing services carries on working on an ongoing basis and we are not going to faff about with telling customers their line still goes at the same rate it always did every month.
We are not sure why we need to underline (confusing on a web page as looks like a link) or embolden the access speed. Also, as we typically quote 20CN, 21CN, FTTC and Be access speeds. Which are we making bold? We do try to make things clear, obviously.
This is then where it gets really silly. The speed is to be a range for 20th to 80th percentiles (and a 10th percentile as guaranteed minimum) of our customers that have the same line characteristics. So this falls apart quite quickly. Apart from the fact that consumers do not understand percentiles at the best of times, the logic is flawed.
- If customers below 10th percentile can complain and leave, then the first 10% of consumers go. So the 10th percentile moves up. This guaranteed minimum speed (10th percentile) is nonsense as the guarantee then means stripping the 10% and moving the goal posts up indefinitely.
- The 10th percentile is done as at or below so even when you have only lines all at the same speed left they are all at the 10th percentile and all fail then. Why not below
- The speed estimates come from BT. We could only do speed estimates for 20th/80th percentiles if we had details of the line characteristics so as to group similar lines. BT do not provide details of all line characteristics, so we could not if we wanted to.
- The line characteristics are not defined. I could characterise a line by what speed it gets. Then your line is one that has, say, the characteristic of being a line that gets a sync of 5732Kb/s. In speed terms you are the 0th, 10th, 20th, 80th, and 100th percentile of lines that are of the same characteristic. That breaks the logic in the code completely.
- Even if the line characteristics were resistance, capacitance, length, insertion loss at a range of frequencies, material (copper/aluminium) and colour of plastic around the pair, then you get the issue of what constitutes a similar line. If it is all characteristics the same (a set of one) or all lines the same colour, or same insertion loss and 1.23586MHz, or what?
- Almost all (more than 905) of the very good lines will have speeds at the limit of the technology, e.g. 8128Kb/s for short ADSL1 lines and so that will be the 10th percentile and so all our really good lines are defined as a problem by OFCOM. Crazy!
Making a record of the estimates quoted is sensible, and we have done that for a long time. Sadly we don't have a 'My Account' section on the web site so we would fall foul of that unless we rename it :-) This is obviously one of the cases where following the spirit not the letter of the code makes a lot of sense.
The idea that customers can leave if they do not get the access speed estimated is an interesting one. This only really makes sense if BT let us out of out contract with them if the estimate they quoted is not achieved. They are not forced to do that by OFCOM, shame. Thankfully BT also do not have more than a month min term or a min notice, so we have said we will try for 30 days to resolve, and after that you can leave (migrate) with no penalty. So adhering in principle and as far as the wholesale contract allows. After all, we do not do the speed estimator. Of course FTTC is an issue as that is 12 month min term. Not sure how best to tackle that. But we could sign up to the code and say FTTC is a "business" product and be exempt from all of these provisions?!?! In practice the best thing for any ISP to do for FTTC is say "your estimated speed is 15M to 40M" and that them meets the requirements.
Customers wanting out of a contract can play the system easily to get out of it by making their equipment sync at a lower rate (by configuration settings on some routers, by creating interference or dodgy wiring, etc). This then allows them to use the 10th percentile get out clause with no penalty. One could argue that they have to have local issues (wiring, etc) changed, but you either believe them when they say they have, or send an engineer at huge cost (and they could then simply put the problem back when he leaves).
Providing information on actual throughput is tricky too - the definition includes so much outside the ISPs control it is impossible to say. We perhaps would have to say that actual throughput will be the same as access speed in normal circumstances. Whilst I think true for us, I cannot see any come back in the code for saying that and it not being what happens. Who defines normal?
At the end of the day we agree we should provide the best information at point of sale. It is "matter of fact" data on what a line is likely to get for each technology and the same as all other ISPs using the same copper/technology. It is useful, but not a factor between most ISPs, so we are not sure why OFCOM are so hyper about it.
3rd principle: Accuracy of information
We have to make sure speed estimates are accurate. But we depend on BT to make estimates. Hands tied! At least OFCOM seem to understand that...
4th Principle: Managing speed problems
Well, it talks of access speeds lower than headline speeds and then talks of managing the problem. The headline speed (as OFCOM seem to understand) is an "up to" speed. So there is in fact only a problem if someone is getting more than we said. If they are getting less than the headline speed then that is the very meaning of "up to" and how is that a problem.
We have the whole thing of 10th percentile of speeds for similar lines being a guaranteed minimum speed. It is nonsense. The concept only works if 10% of all customers do not report they have speeds that are too slow. Is that what OFCOM want? 1 in 10 customers suffering with slow broadband
Oh, and the customer can then leave "within a three month period of the start of the contract or longer". Yes, they say that. "within 3 months or longer". It makes no sense. WTF is that?
Also, you have the whole line characteristics issue. If we say "your line clearly has different characteristics than we first expected, i.e. it is characterised now as 'crap'. We find that now you are actually over the 10th percentile and so we don't need to do shit for you"... It all makes no sense to me, sorry.
OFCOM do suggest we have a means to find the actual cause of problems and try and fix them - this is what we do. BT make this increasingly difficult, but maybe this code can be used to motivate them.
5th principle: Presentation of broadband information on the website
Lots on fair usage and limits. It is worth pointing out that AAISP have neither. You can use as much as you can afford - there is no limit. So pretty much all of this would not apply anyway.
Some on traffic shaping - we do not do protocol specific shaping to manage customer traffic. We do some stuff under customer control to help VoIP and the like and prioritise premium rate customers, but aim not to be the bottleneck anyway so not an issue. Not sure how that fits in with OFCOMs requirements. They make a lot of assumptions about how ISPs work.
6th Principle: Timescales
FFS this is not a principle. They are abusing the term. It is just that signatories have 12 months. We don't plan to sign. It is silly. We are implementing our code of practice now and expect the one new thing (stats on congestion of lines to carriers) end of this or next month.
7th Principle: Monitoring
Again not a principle. Good idea. Not relevant for non-signatories.
8th Principle: Awareness
Again not a principle. Good idea. We intend to make people aware of our code and maybe get some other ISPs to sign up to it.
Our aim is the best link you can get so we are not the bottleneck. Other ISPs with a similar view talk to me - we can make a published code that says that and all agree.
