Tuesday, 10 October 2017

One in ten UK broadband lines are faulty, says OFCOM ?

We have had this in the past and once again we seem to be facing another (voluntary, phew) broadband speed code of practice from OFCOM.

Our reply to the latest consultation is here (pdf).

But once again the big issue here is that OFCOM consider any lines where the speed is below the 10th percentile of speeds for similar lines to be "faulty".

This means :-
  1. The customer can expect the ISP to try and "fix" the line, taking up to 30 days to do so.
  2. The customer can expect to be allowed to exit with no penalty and to get a refund of upfront costs if not fixed within 30 days.
Now, if the line is actually faulty, as some will be, this is all very reasonable. But the threshold is not a "fault threshold" as determined by measuring the speed of similar lines that are not faulty. It is set to the 10th percentile of speeds of similar lines.

This means OFCOM are defining that one in ten lines are faulty, end of story... In fact, this is a moving target. If some part of those lines are faulty and fixed, all that does is push up that threshold.

In fact, assuming the ISP can get a refund from Openreach or the carrier then it is in their interests NOT TO TRY AND FIX such lines. If they do, they will end up with more and more lines that are NOT FAULTY but below the 10th percentile if they do start fixing the genuinely faulty ones. Those lines simply cannot be "fixed" and so just cause even more hassle for the ISP. An ISP will actually want a load of low speed faulty lines that are not complaining so as to reduce the 10th percentile level.

The problem is that if you are unfortunate enough to be in that bottom 10th percentile, and bear in mind that one in ten people will be, you may well have a service that is indeed doing the best it can and there is no fault whatsoever on the line that can be fixed by anyone.

It is as bad as trying to say that every school should be above average or some such. It makes no sense. Why on earth do OFCOM still insist on this nonsense in the code of practice. Why do so many large ISPs agree with OFCOM by signing up to their code of practice? Are BT plc really saying one in ten of their lines are faulty. I am glad A&A don't say that to be honest.

I wonder if any other countries in the EU or the world publish stats on broadband take up, and how many of those lines they consider faulty. The UK must be leading the way with 10% of all lines being faulty by definition of the regulator.

I have to wonder if there is any other industry in the UK, or in the world, where the regulator defines that one in ten of the things you sell are faulty, regardless of what you do, even to the extent that customers can get a refund on that basis? Imagine if OFWAT defined that the lowest 10th percentile of water pressure was a fault and water companies had 30 days to fix or else refund the customer. This is basically what OFCOM are saying about broadband.

We'd love to sign up to the CoP, as it has many good things, but until this fundamental issue is fixed I don't see how we can. We simply do not agree that one in ten UK broadband lines are faulty, sorry. When there are faults we fix them (whatever speed that means you line gets when faulty).


The fix?

Have the modem providers, e.g. Openreach for most FTTC/VDSL, and BT Wholesale or others for most ADSL, define a realistic "fault threshold" which is the lowest speed for non faulty similar lines. Use that as the reference, and have them guarantee that to the ISPs who can pass on that guarantee to the end users. Not complicated!

23 comments:

  1. I'm glad it's not just me who thinks this "all schools under average must be forced to improve to average or above" is non-sensical. It's written by someone who has no understanding of the word 'average' and how that figure might be calculated. Which is probably most people...!

    ReplyDelete
  2. Of course, it may simply be an ill-designed attempt to force telcos/ISPs to improve the worst lines on an ongoing basis.

    ReplyDelete
    Replies
    1. Except you can only really improve the faulty ones - DSL is an adaptive service, one in ten customers will be in the lowest 10th percentile of speeds, always. Indeed, it is better to have a load of not-yet-complaining faulty slow customers if you can so as to lower the 10th percentile.

      Delete
  3. Surely all this gets prevented by the solution of selling lines which are "greater than X Mbps" rather than selling lines which are "anything up to Y Mbps". Then, provided X is chosen carefully, almost every line will be far above the contracted minimum and, by definition, not only "not faulty" but also "providing an extra, free, speed increase".

    ReplyDelete
    Replies
    1. Well yes, except OFCOM decree that X is the 10th percentile speed of similar lines. That is my point! The fix would be to set X based on what is a fault and not an arbrary “one in ten lines”.

      Delete
    2. But when ISP A advertises "(at least) 10Mbps" along side ISP B that advertises "(up to) 80Mbps", a large proportion of the population will conclude that ISP B is offering a better service, when in fact it may well be exactly the same service. So the incentive is for ISPs to use "up to" figures because the numbers are bigger and the customers aren't clueful enough to understand and choose the more honest ISP.

      We saw exactly this when getting quotes for a new leased line for a customer - the ISP we got a quote from said "we can't do a 1Gbps bearer in that location", so the customer went against our advice and opted for a different ISP who said "1Gbps bearer, subject to survey". The second ISP, of course, did the survey and found they couldn't offer a 1Gbps bearer in that location, but by which time the order was well underway so rather than cancelling the order, the "dishonest" ISP got the business. (It should be noted that the winning ISP is a complete nightmare to work with and has a terribly unreliable service).

      Delete
    3. I agree advertised speeds are another issue. It is relatively sensible to have a "guaranteed minimum" speed in many ways, if that is set at a level nobody with a non-faulty line should ever sync below. What makes no sense is picking a level that you know one in ten lines will sync below. You want only faulty lines to be below it, not 10% of all lines!

      Delete
  4. Or you need to make sure that the worst 10% of your lines are owned by you...

    ReplyDelete
    Replies
    1. I'm reminded of the old (apocryphal?) story about IBM using a new Japanese chip supplier, and having a contract specifying no more than 10 faulty components per thousand - so the first delivery arrived, one thousand chips, plus a little bag taped on with 10 faulty ones and a note asking what IBM wanted them for...

      The daft thing here is Openreach/BT Wholesale already have a more sensible scheme in place, the "downstream speed handback threshold": set a speed the line *should* achieve if it's not genuinely faulty, depending on the line length. So, if my 79Mbs VDSL line drops below about 52, they'll admit that's a fault and try to fix it, because there would genuinely be something wrong - meanwhile, my grandfather's damp string in the back of beyond is doing well to deliver a dialtone never mind sync, and that's not a fault.

      Show that to customers during ordering, and you're most of the way there anyway. No need for stupid nonsense about declaring "faults" based on flawed statistical assumptions!

      Delete
    2. Oddly, we (a) see very few if any lines below the hardback threshold, and (b) BT define that as a 10th percentile which is what OFCOM are asking for. These two facts don't quite make sense to me as we should be seeing one in ten lines below the 10th percentile, i.e. below the hardback threshold!

      Delete
    3. There should, however, be something in place to incentivise Openreach to upgrade the "damp string" type lines (but anything that penalises ISPs rather than Openreach themselves isn't going to achieve this).

      Delete
    4. The problem with the handback threshold is that it just enables you to walk away and get your money back. I don't want to go back to ADSL speeds, I'd like Openreach to 'upgrade' my line to twisted pair (I'm currently syncing 20% below the handback speed)

      Delete
    5. The official line at least (not sure of the reality, something Adrian might be able to tell us?) is that if it's below that threshold, they'll accept it as a fault report, and try to fix it rather than just cancel service (since they don't get paid then). The option to revert without penalty is important for corner cases, though: if you're really unlucky and VDSL gives a lower sync speed than ADSL did, you'd otherwise find yourself stuck paying extra for a year of reduced speed!

      Delete
    6. Sorry but that is not relevant - if you are a long way from the cabinet you will be in a group of "similar lines" that are a long way from the cabinet and have a 10% chance of being below the threshold like anyone else. It is nothing to do with absolute slowness, and you can be slower than ADSL speeds but still be above the hardback threshold with no problem.

      Delete
    7. Yes, but the handback threshold gives at least some safety net, that you know before ordering you won't be stuck at less than X Mbps - if that's higher than your ADSL speed, you're guaranteed no drop, if it's lower at least you've been warned, and there's a limit to how much drop you risk without the option of reverting free of charge.

      Is BT genuinely using the bottom decile as the cutoff for that? Published explanations seem to imply that, but we're all in agreement that this would be a stupid way to set it, and it sounds as if A&A's lines suggest otherwise as well.

      Delete
  5. > Why do so many large ISPs agree with OFCOM by signing up to their code of practice?

    Probably because they know that OFCOM won't actually enforce it...? /snark

    ReplyDelete
    Replies
    1. Fixing faults is an important thing, but this covers only faults that may a low speed below 10th percentile. It does not address drops in speed that are above that, or packet loss, or latency, or frequent drops in sync. It is not effective as a fault repair CoP in any way.

      Delete
    2. I think you are looking at this too literally, and just with an engineer's view. From a regulatory point of view, pushing on one particular pain point can cause improvements to many others by causing ISPs to change their priorities and make the desired business decisions. And if it doesn't work, they can consider changing the rules later.

      Delete
    3. I am not sure I see your point - what is the "desired business decision" they are after - are they, as they seem to be, trying to get businesses to actually guarantee that all lines are in the top 90% of lines? How is "happening to be in the bottom 10th percentile" (which one in ten people will be) a "pain point". Surely the "pain point" is actual faulty lines? Happening to be in the bottom 10th percentile is not. Indeed, bear in mind, this is for "similar lines" so this is not "lines that are slow", it is "slow for where you are". The 10th percentile for some places may be 74Mb/s in some cases (one line I just checked). Why is saying someone that gets 73Mb/s a "pain point"?

      Delete
    4. Sorry, I wasn't clear. I don't think the goal is primarily faulty lines. The business decisions I am referring to are ones around investment (including CapEx), not just break-fix actions.

      What I meant is, don't just look at the particular metric and penalty defined in the measure, but look at how that will drive longer term business decisions made by you and your competitors, in areas not directly related to fault handling.

      Delete
    5. That still does not make sense, sorry. See next blog for more concrete examples. This is picking an almost arbitrary one in ten people for special treatment with no logical reason to pick them, and no reason to think the special treatment will help anyone. They are not even picking on slow lines for this special treatment. I hope the next blog post clarifies my concerns here.

      Delete
  6. We need to understand what are the goals here, and do the proposed measures drive towards the goals? As I am not an ISP I haven't looked at the consultation -- does it make clear the regulatory goals?

    As a user, I think reasonable regulatory goals could be to drive investment (both CapEx and OpEx) by ISPs (and wholesalers) in improving (i) speeds, (ii) predictability, and (iii) fairness. I have a perception (rightly or wrongly) that the ISP industry as a whole (not A&A, of course) are happy to take my money and are not investing sufficiently in those three goals.

    This CoP feels like it might change the business priorities sufficiently to drive towards those goals. Of course a target of beating the 10th percentile can't go on indefinitely, but it may well be useful for several years. Particularly if some of the problems are not just "break/fix" but may be caused, at least partly, by insufficient investment (for example, more engineers to respond to faults faster, replacement of overhead lines with underground, replacing modems in a cabinet with newer models, adding another cabinet in a rural area, or even providing some FTTH connections). We need a CoP with incentives to drive all sorts of investment.

    Of course, there is an issue with making sure that the costs get fairly split between the entities who are involved in the problem. Clearly the retail ISP has to feel some pain (so that they provide sufficient staff to manage the issue, and have an incentive to try to keep the customer) and the wholesale ISP has to feel a greater pain (as they are the main area where investment in everything from engineers to major CapEx are required).

    I think it will be a shame if A&A does not sign up to the new CoP.

    ReplyDelete
  7. The fact that Ofcom do not seem to consider latency important is worrying, and shows, as usual, an ignorance of what they are dealing with. Virgin Media have a very nasty issue with their Superhub 3 firmware that ruins the connection for real time applications (I've managed to get a ping of 7ms across 3m of copper to my router... I can get from our managed CPEs to our core network in less time for some of our customers! On a good connection I'd expect to be able to ping London in around 10ms from my location).

    Stating 10% of lines are faulty by definition is insanity.

    What I do think this will benefit is cases like Virgin Media where some areas are far too oversubscribed. It'll also be useful as FTTP rollout increases, where oversubscription could also come into play, particularly depending the splitter ratio Openreach end up using for developments.

    ReplyDelete