Friday, 13 October 2017

Another little gem in the OFCOM CoP

There is another little gem in the OFCOM Broadband Speed Code of Practice in 2.23

When network infrastructure providers or wholesalers make available the live access line speed that is actually received on the customer's specific line, ISPs must use this as the basis for speed estimates (rather than using an access line speed range for similar lines) in circumstances where they will be using the same infrastructure and access technology to provide service. This must incorporate the measures of contention derived from the testing outlined in paragraph 2.20, and should still take the form of a range, where possible.

So, let's make sense of this. Normally the requirement is to provide a range of estimated speed that are the 20th and the 80th percentile speed of "similar customers", and set a guaranteed minimum of 10th percentile speed. As I say this makes one in ten lines faulty by definition.

But consider one of those random one in ten that are faulty, getting service. They complain. The ISP "canna change the laws of physics captain" and it gets no better, so the customer gets a refund and leaves to another ISP.

So new ISP ideally gets to see the sync speed, or gets from a carrier new speed figures based on the carrier knowing the actual sync speed. This gives a few problems :-
  1. Knowing the new sync speed it is still necessary to report a "range" ("where possible"). Well, the only range allowed is 20th and 80th percentiles, but this is a sample size of one! The 20th and 80th percentiles are the actual sync speed of that one sample. How could a range be given? What are the rules for working out that range. I can only assume it is going to be not possible, or the range will have to use some other, perhaps saner, criteria than percentiles.
  2. Assuming the ISP just makes shit up and picks a range from below the actual sync to above the actual sync in some arbitrary and undefined way, and then, of course, picks an arbitrary minimum guaranteed speed that is even lower, what then? Well now the customer migrates to a new ISP, using the same modems and the same line, and getting the same speed. All that has changed is that now they no longer has cause to complain.
This helps the customer how, exactly, OFCOM?
This helps the ISPs or gives them any incentive to change things or invest, how, exactly, OFCOM?

Maybe the existing ISP, on complaint, can offer to "migrate you to us, at not charge, here are your revised speed estimate and guarantee"? Who knows...


  1. Well it says a range must be given 'where possible'. If the sample size is 1 then it's not possible. So that's fine (if a bit strange) - you would quote a speed based purely on the current actual sync speed.

    The problem of is that an ISP like A&A who will take on knowingly faulty lines and try to improve them is hampered by this, because you would be required to quote the current sync speed as the speed you expect the line to achieve, even if you have reason to believe you can do much much better.

    But I don't see how this affects the 10% of lines issue. Irrespective of whether the estimate is given based on actual current sync speed the line is still grouped within the same set of "similar lines" and so it either is or is not still in the bottom 10% of those. Or am I missing something here?

  2. You can still give a range in this case - for example "between (known sync speed) and 0". Or maybe give a range between those extremes, and use some traffic shaping to ensure the effective line speed falls below, in and above your specified range 20%, 60%, 20% of the time respectively. Hard to argue with that; also hard to see how it would benefit anyone at all ever!

  3. The mention of contention is interesting (or opens another can of worms!) - without that, it could be taken as more rambling about sync speeds Ofcom don't quite understand - but they want "peak time" (arbitrarily defined as 12-2pm, which might come as a surprise to residential ISPs!) contention to be taken into account. So presumably a certain large ISP would have to start quoting "speeds" of zero, thanks to overselling backhaul to the point even web surfing is an exercise in futility...

    One bullet point does note that where the "range" is less than 2 Mbps you can just quote that figure instead, so at least there's a glimpse of sanity there.

    Perhaps this is all a devious ploy by Ofcom to make xDSL such a hassle to sell that Openreach get their finger out and offer FTTP properly instead?

  4. Surely there is still likely to be a range, as for most isps now a new customer will get a new modem, which may have slightly different characteristics, and of course, each time I power cycle my modem I will get a slightly different sync!

    1. Well, one could make up a range, yes, but one cannot produce a range from 20th to 80th percentile as OFCOM have said.