Thursday, 11 June 2015

Slow broadband

Seems OFCOM are being daft again, see

One of the issues with the code of practice, which, last time I looked, was voluntary, was that it defines all lines at or lower than the 10th percentile of similar services as being "faulty" and so could be ceased for no charge.

This causes a load of issues.

For as start, the "no charge" is simple for any ISP to work around. Make the router you supply free as long as you take service for X months, then if you cease within that for any reason you have to pay for the router. Supply of a router is not a communications service, it is a simple sale of goods, and so not covered by the OFCOM rules or ADR. You could even invoice it on 6 months terms with an agreement of a bonus/credit once you have the service for 6 months meaning you won't have to pay the invoice. So OFCOM trying to change business models of ISPs where the install and equipment is cheap in return for a minimum term simply won't work.

However, the far bigger issue is simply defining 1 in 10 lines as faulty. It is mental. It does not encourage an ISP to do anything about slow lines, as whatever they do they will always have 1 in 10 lines defines as faulty.

Indeed, it actually encourages ISPs to try and have 10% of lines working really slowly if the customer is not actually complaining, perhaps even deliberately so, as that lowers the 10th percentile speed and removes the issue of people leaving because of slow speeds. After all, if the ISP fixes those slow lines, or the ISP loses those slow line customers, they just have a higher 10th percentile and more lines that are impossible to improve as they are working as well as the service can let them.

I agree, making estimates at point of sale as accurate as possible, and even the idea that failing to meet those is reason to leave, is sensible. But if you are on a long line and have an estimate of 1Mb/s and get 1Mb/s why should you be able to consider that a fault just because you are in the lowest 10%? You canna change the laws of physics captain - there will always be 10% of the people on the longest lines and they will not be "fixable" just because they are on long lines!

Of course for more modern links, like FTTP, you get even more crazy. OFCOM defined it as lines at or lower than 10th percentile. When all lines get exactly 80Mb/s on a fibre to the premises, that means all of them are at or lower than the 10th percentile! The better the ISPs service the worse the problem for them and so OFCOM are discouraging good service and speed.

If OFCOM start insisting on this code of practice we will start a telemetry package deliberately clamped at a very slow sync rate on DSL, and sell that to people with parking meters and signage and credit card terminals and so on. That way we can try and get the 10th percentile down to like 250kb/s and be immune from this stupidity.


  1. Note that it's not as crazy as you think; it does not apply to FTTP, for example, as those are not rate adaptive lines. Nor does it apply to Virgin Media, as the line speed is fixed, so not rate adaptive.

    Your telemetry package won't get you out of this, either; that's a separate service (as the DSL profiles in use aren't the same), so gets its own 10th percentile.

    OFCOM's goals do appear muddy, however, as this only lets you move retail providers if your rate adaptive line is in the bottom 10% of lines of that equivalence group (which appears to be based roughly on the estimates you get from BT et al). With BT's effective monopoly on FTTC, this is just going to let the slowest 10% of FTTC customers churn as fast as they like to no effect; there needs to be some mechanism in place to force network operators to move customers to fixed rate services.

    1. The other problem with most of these things Ofcom bring in is they only bind the retail CP, e.g. in this case I imagine it's not going to force BT Wholesale to let ISPs out of the 12-month term for FTTC...

    2. Indeed, I have a big concern over this new notice of transfer - we have to allow cancel of migrate in free of any charge but looks like BT charge us around £19, which is crazy.

  2. If it is provable that the ISP has been unable to provide an acceptable level of service due to problems not fixed by BT Wholesale, surely a class action by all of the affected parties (retail CPs) against BT Wholesale would be possible ?

    1. That, or action by Ofcom - BT Wholesale is supposed to be regulated too, and if it's doing things substantially detrimental to its customers like that (particularly where the risk is avoided for BT Retail) they might actually act.

  3. Let's try and be productive; what should Ofcom be requiring to get consumers a reasonable outcome, given the reality of a market containing both rate adaptive and fixed rate services? My suggestion:

    1. You are given a predicted speed estimate (PSE) at sign up; if the line speed drops below the PSE and the ISP is unable to fix it given reasonable time and access to premises, you are entitled to cease or migrate at no fee regardless of contract duration. The PSE should be the IP layer throughput that the customer will see when all the bottlenecks and congestion on the path are within the ISP's network, so BT IP Profile if the ISP is not normally the bottleneck. Further, the ISP must not mislead customers into expecting higher speeds than the PSE; customers must be told that the PSE is the speed they can expect.

    2. ISPs must provide a mechanism to verify your current line speed; this is permitted to be intrusive (e.g. blocking all other packets on the link for the duration of a speed test) if the ISP wishes; however, all it has to do is get you the IP throughput that you could achieve on your line (so could just get the Tx Rate from a Firebrick L2TP LNS).

    3. Whenever you discuss the speed of your product, you must give the median PSE you've issued at sign up and the lowest PSE you will issue for that product. You may not mention the highest theoretical speed of the product unless you also mention the highest PSE you have issued for that product. So, a VDSL2 80/20 product could be referred to as "typical 55M - should be greater than 10M", or as "up to 80M; typical 55M, real world speeds between 10M and 70M", but not as "up to 70M" or "up to 80M, typical 55M, minimum 10M".

    4. If a customer's speed exceeds the PSE given to them at sign up by more than 25% for at least 10 days in a row, then the customer must be given a choice between ending their contract without penalty (as if the speed had fallen below the PSE and the ISP had not fixed it), or accepting a new PSE. Their new PSE does not affect the speeds you're allowed to advertise.

    The first two are just good practice - promise a given speed, and provide a way to check that it's being delivered. The third is the disincentive for an ISP advertising on speed to routinely underestimate at point of sale - it reduces their advertised speed. And finally, the fourth item is the incentive for an ISP that does not advertise on speed to give a reasonable estimate - if you routinely underestimate by too much, your customers can leave anyway.

    This also avoids the issue of fixed speed links like FTTP nicely - the best PSE for a fixed speed link is 80% of the actual line speed, as that's the point at which neither limit can be breached.

  4. I'm not sure how it would improve anything either. I'm on a long line, stuck on around 3Mb/s. So I'm probably in the lowest 10% and yes I'm unhappy (with BT/Openreach lack of investment - not AA of course) so what can I do? Move to another provider who will use the exact same copper line and provide me with a service at best comparable....
    Most people stuck in the bottom 10% are not there by choice and the ISP is probably not the main bottleneck.

    It's a bit like penalising people on benefits with a spare bedroom when their council is unable to provide them with a properly sized alternative. Punishing the innocents and not providing any incentive to the real responsible of the problem.

    1. I agree on that first point: it's almost always Openreach/BT Wholesale who control the sync speed, which is the usual bottleneck on bad lines like yours. Setting a sane fault threshold, below which BT have to fix it, would achieve a lot more, IMO.

      What the ISP does control, though, is the actual throughput achieved. What use is a line that syncs reliably at 20 Mbps, when the ISP is so over-contended you only get 2 Mbps at busy times?

      Under this policy, it would appear you could take your 3 Mbps line from A&A because it's "too slow", switching to an ISP with more optimistic SNR target settings, getting a less reliable 4 Mbps sync speed but only 1 Mbps of throughput because they're so oversubscribed - but be stuck with a slower and less reliable service, because some of their customers are even worse off! Good work Ofcom.

  5. I disagree with you on this one rev, normally I agree with most of what you say, but to start you seem to have misunderstood ofcom's change here. The 10% is not 10% of an isp's lines its 10% of lines with a similar speed estimate, so its only comparing to lines of equal spec. Now of course you are right there will always be a bottom 10%, but what is wrong with bringing that bottom 10% closer to the average?

    Now of course this may just lead to lower estimates rather than actual work in the field to fix problems. ONly time will tell on this.

    1. No, sorry. For a start, it still means 10% of lines will be faulty. Move those "closer to average" and there is still 10% in the bottom 10th percentile. It is a never ending task with no success. Secondly, simply picking the slowest lines, even "of a similar spec" is not finding lines that are faulty necessarily. It is treating "speed", or possible "sync speed" as the only relevant metric, not packet loss or latency. It means that one can simply increase speed and make a very flaky line just to get it out of the lowest 10% without actually fixing lines that have a fault and are faster (when working). It is a broken way to assess "faulty lines" and not helpful to anyway.

    2. To quote what was supposedly said by an education minster "We won't stop until all schools are above average".