Friday, 30 September 2011

SHOCK! half the country below average...

OK, what the hell is it with some people.

We have quotes like "A recent study by Ofcom found that many services marketed as up to 20Mbps actually achieved an average of just 6.8Mbps."

Well, yes, and? What is the problem? Only if the speed was more than 20Mbps would the advert have been wrong. That is what up to means. The ISP was not saying "average". Indeed, an average (mean) is not really what helps. What matters when buying broadband is "what speed can I get" and that is something every ISP has had a checker to tell you. The headline speed is key for picking between different technology like ADSL1, ADSL2+ and FTTC, and allows that choice where there is one. The new rules will not help any consumers, will make it more complex, and may even steer customers the wrong way for what is best for them.

People making such quotes are exactly the sort of people who are horrified when a study shows "that 50% of people are below average" on some metric. E.g. below average wage, etc. With a lot of distributions it is absolutely expected that 50% are below the mean. Yet politicians will be horrified at such findings. Broadband is not quite that, but naturally the average speed of an up to 20Mb/s service will be somewhere below 20Mb/s.

Of course, the new rules the ASA has come up with are themselves full of holes. For a start they say that any speed has to be achievable by 10% of customers. But this clearly has to be "for that specific service" else it is meanless. i.e. something like 95% of our customers could get a gigabit if they were prepared to pay for an individual fibre dig - does that mean we can say "up to 1Gb/s Internet" and sell them 1Mb/s ADSL1?

So if it is for the specific service, then ADSL1 can be sold as "up to 7.15Mb/s" as now, because we actually have over 25% of such lines getting that speed. So no change there.

Of course, if we wanted, we could have a special 20Mb/s ADSL2+ service that is only available where you can get 20Mb/s. Because more than 10% of the customers on this special 20Mb/s service can get 20Mb/s we can advertise it as 20Mb/s (100% of them would get 20Mb/s). All we would be doing is restricting who can get that service, and anyone that cannot will be offered our alternative "ADSL2+ variable rate service" for which we would not be quoting any headline speed at all. The advert would be headline "20Mb/s broadband" when other ISPs are saying 15Mb/s or 13Mb/s, etc. I may suggest BT and Be have a profile option on the DSLAMs for a fixed rate 20Mb/s service just to allow this to be advertised.

Of course, if it was not "per service" then one just makes a brand or separate company only selling that service and problem solved.

Also, (based on comments made by Enta), if it is across the ISP then you can get unrealistic differentiated adverts. An ISP mainly only doing cities can advertise higher speeds but one specialising in rural areas cannot. So someone in the rural area would pick the ISP that specialises in cities even though the service is in fact exactly the same from both ISPs and the one specialising in rural areas will have services and staff better suited to those rural customers.

Why don't they think about the consequences of these stupid rules?

Anyway, we have removed speeds and refer people to the line checker - simples.


  1. this "50% lower than average" has always bothered me... even if you bring the lowere speeds up, the average also goes up and thus there is still 50% below average... morons.

  2. Reminds me of OFCOMs daft ideas on speeds where below a certain percentile is considered a fault - well, that's a problem that will never go away if you define it like that!

  3. Just seen a BT advert on the TV 'You get a personalised speed estimate up to 20Meg'

    Well, that pretty much solves the ASA issue!

  4. I understand where they're coming from - they're trying to get rules in place to ensure that if (for example) RipOff Broadband advertises 24M broadband, people who see the advert can actually get 24M or close to it. The reasoning is to be fair to honest ISPs - you don't want people going for 24 month contracts with RipOff Broadband (who claim 24M, neglecting to mention that it's only available if you're on the NSBOG exchange - everyone else gets BT 20CN), rather than AAISP (who give you an honest speed estimate) solely on the basis that AAISP is quoting 1M, while RipOff is quoting 24M.

    Note that the ASA requires 10% of people to get the throughput quoted, not just the sync speed - in other words, you also face pain if you're heavily traffic managed and congested.

    It's a start - I can see it remaining tricky until we get FTTP and fixed speed products everywhere.

  5. I cant see throughput claims being policable, it is testing to a point outside the ISPs control in most cases (speed test sites) and if it is end users testing then many factors will affect the results from local wiring, wifi interference, to type of TCP stack.

    What pisses me off is that there are less subjective tests you can do - like loss and latency - but nobody is interested in doing such tests.

    I don't think the new rules will help - they will confuse the public way more.

    "RipOff broadband" can still sell 24M as long as they only sell to those people that can get it. That way 100% of their customers get that speed. As for back haul - all they have to do is ensure traffic management prioritises traffic from speed test sites as that is what is the test.

    I am all in favour of clear an honest advertising. I never found "up to" to be unclear or dishonest as it is a maximum, not a typical, and is the same maximum for all the ISPs offering the same technology to their customers so the customers have to differenciate based on things like customer reviews.

    What would be a way way better thing to do, in my view, is ensure you cannot tie someone in to more than a month contract. That would mean anyone making any exaggerated claims would simply lose out. Then, trying to police the claims being made, would not be needed as they would simply lose out to market forces.

  6. What I think should be measured, and reported, in these studies/surveys is not a comparison of the sync speed (which is outside the ISP's control) but a comparison of the sync and actual download speeds obtainable by the individual customers. So if, for ISP A customers who sync at 16Mbps can only get 2Mbps download during peak times that ISP would rate lower than ISP B where customers who sync at 12Mpbs get 10Mpbs download during peak periods.

    I know that this would be harder to measure but would give a much more accurate and useful (to a prospective customer) indication of the quality of the ISP. I am sure that A&A would rate highly using such a measure.

  7. The trouble with 1 month contracts is that you then can't subsidise install from monthly revenue, which is considered a public benefit.

    How about a settling in time, with proper fee-passing in place? You get 14 days with your ISP, and if you don't like it, you can migrate to another ISP paying the new install charges - the difference between install and migrate charges is then paid back to the original ISP. This lets you discover that RipOff Broadband aren't as good as you expected at free install, 24 month contract, £10/month, and that you might want to pay the extra for AAISP.

    Migrations, I can't see the need for a long term contract - you simply let people migrate out.

    And equipment is simple - it is owned by the ISP until they choose to give it to you, so if you get a "free" router with your 12 month contract, the ISP can ask you to return it in reasonable condition - letting them loan it out to a new customer instead of you.

  8. If they are going to police throughput speeds, it would only be reasonable if there was a single OfCom-run speed test server, which everyone can use. And that ain't going to happen!

    Cheers, Howard

  9. That would at least be relatively level playing field, as long as it had modern TCP (selective ACKs, sensible window scaling, etc) and peered at LINX.

  10. I must admit that I would like to see ISPs charging by the speeds they can *actually* deliver (and by extension, BT charging the ISPs for the speeds they can deliver).

    I don't live in a rural area, but my "up to 24Mbps" 21CN ADSL2+ only manages about 9Mbps (actually, I've switched back to 8Mbps G.DMT now because ADSL2+ was simply too unreliable). There isn't a lot of incentive for BT to upgrade the infrastructure here because they are getting exactly the same income from me as they are from someone 20 metres from the DSLAM. In fact, it is probably beneficial to them to not upgrade the lines because lower sync rates reduce the peak bandwidth usage on the whole network.

    If prices were tied to sync rates, there would be incentive for them to upgrade as they would recoup the costs of the upgrades through being able to charge more.

    If course, then you'd need to give people the option of sticking with a lower speed (lower priced) product rather than the current trend of upgrading the DSLAMs and migrating all the end users "for free". And if you're simply using sync speeds as a metric, it doesn't cator for situations where there isn't enough backhaul capacity. But I certainly think it'd be an improvement.

  11. Well, in general, sync speed is irrelevant to cost. What matters is how much people actually use. The issue is when sync speed is silly high (like proposed 300Mb/s FTTP) then the ISP has to allow for that headroom for just one customer to try and fill their line even for only a few seconds. That is expensive!