2012-06-05

Sustainable pricing for broadband

I saw an interesting post on twitter about this.

I had posted about the fact that our pricing is (and always has been) based on days of the week and not whether they are a public holiday or not. The exception being that we usually do a special Christmas rate for that week. Apparently the post was condescending, and I do apologise for that. It is a tad frustrating handling some of the questions we get on this, and we really do try to make it crystal clear how the charging works. We are not trying to trick people, and there are a lot of industries and services where pricing is not affected by public holidays.

We do close sales, accounts and support over public holidays, but that is mainly for the same reasons we only work 9 to 5. Getting engineers out on public holidays is almost impossible, so taking calls outside normal working hours is not usually going to help fix a problem any quicker. Various people are working during the weekend anyway, and we do have support on irc by various staff giving their own time. We also have people ready to handle any major outages whatever the time. So it is not that we are not working.

But one comment was that our pricing is unsustainable because of IPTV. This is an interesting view, and got me thinking. It is not, as suggested, why the posting was made. To be honest, that was to try and make sure people did not make assumptions and regret it later - I was trying to help, honest.

IPTV means there is now a gradual increase in the usage of broadband lines. More and more people are starting to do some streaming and downloading of video that did not before. This is an issue in the industry, obviously.

One of the main problems is that back-haul from a customer premises to a handover point in the UK costs 10 times what connectivity from an ISP to the rest of the world costs. Yes, it seems crazy that a company that only has to link a few hundred exchanges in a small geographic area charges so much more than companies linking thousands interconnects all over the globe.

However, the upshot of this is simple, IPTV is increasing costs for ISPs. It does vary from ISP to ISP and depends how they connect to customers. Some ISPs with exchange equipment are in a better position to cope, those using BT back-haul are not so much. But either way, it means people spending more upgrading links from exchanges. That does cost money.

So that leads me to ponder what is a sustainable pricing model for an ISP. One where prices relate to costs in some way (like us), and increased usage (because of IPTV or otherwise) means both increased costs and increased revenues... Or, a business model where price competition drives prices down, and ISPs offer "all you can eat" services where IPTV increases costs.

What do I think? Well, I am not an economist, but I think that this all puts pressure on the back-haul links and their costs. I think competing on price is less sustainable, to be honest. Transit costs have plummeted over the years as usage has increased, but back-haul has reduced more slowly. What I hope is that pressure on the back-haul pricing means that it gets a lot cheaper. That means we can lower our prices to customers. Even so, I don't see us competing on price - it means compromising far too much on other aspects of the service, and means potentially running an unsustainable model. We intend to stay in business.

Eventually, back-haul links will be high enough capacity for fixed pricing models to be viable. When that happens we will offer them. Having said that, we are trialling some interesting options for businesses even now, with much more affordable, fixed price, all-you-can-eat Internet links using FTTC and Etherflows. They are not cheap enough for residential use yet, or rolled out to much of the UK yet, but we are getting there.

So, sorry if my status post upset anyone - I was trying to avoid the calls on Wednesday with people asking why their usage is so high when it was a public holiday.

I hope everyone has had a good weekend (we all went to the zoo yesterday, 10 of us).

17 comments:

  1. People seem to read tone into written communications, often altering the sense, or perceived attitude of the author.

    ReplyDelete
    Replies
    1. I think you are right - and that does make it a tad difficult at times.

      Delete
    2. I had someone read the wrong tone into "How can I help you?" on an IRC channel once. Some people manage to read the wrong tone into anything.

      Delete
    3. I am somewhat blunt at the best of times, and sarcastic the rest, so anything I say comes across badly if I am not careful :-)

      Delete
  2. I can see there could be a problem for people who want to watch a lot of IPTV between 9am & 6pm on weekdays but most people are going to want to use IPTV during off peak times when data is a lot cheaper.
    I'm not really a fan of the "all you can eat" business plan as it inevitably means some customers will be subsidising others so I prefer charging based directly on actual usage as it is a lot more fair.

    ReplyDelete
  3. Is the back-haul data pricing cheaper in the evenings? Looking at the LINX stats, the peek data usage is late evening and therefore the pricing model seems to be back to front.

    ReplyDelete
    Replies
    1. Generally the pricing is the same all day - the differential pricing for times of day allow this to be balanced. In practice, the evening/weekend usage has now exceeded the daytime business so not quite as balanced as we would like.

      Delete
    2. Is evening & weekend usage higher because you have more residential users than business users, generally?

      Actually when I say more, I suppose I mean "more usage from", regardless of the number of circuits sold to businesses versus home users.

      Because business (assuming office use) isn't going to be so included to use streaming video, or downloading stuff.

      Maybe it's time to change those time boundaries and amount of data per 'unit'

      Delete
  4. I was thinking about your original post.

    I think the confusion comes from the fact a lot of people use "Mon-Fri" and "Business Days" interchangeable. You're obviously not, but it's a common enough usage there's ambiguity.

    ReplyDelete
  5. I always took the view that your pricing structure passes the decision about contention on to me as a consumer. I can pay more and have an uncontented connection, or shape traffic heavily and have a connection that performs like the "all you can eat cheapest on the market" ISPs.

    The other advantage of having a pricing structure that has a fixed and usage based component is that you have a strong economic incentive to keep things running as fast as possible, which simply doesn't exist for the "look at our headline speeds/prices" crowd. If more ISPs adopted similar pricing structures you can bet there would be more pressure on the wholesalers to push faster connections out further rather than the current "looks alright from here" mentality.

    ReplyDelete
  6. For a bit of insight into pricing, on raw bandwidth costs (ignoring hardware and circuit costs) my place-o-work can buy in internet bandwidth at as low as £2.50/Mbps, although because we use a couple of different providers it averages out to around £20/Mbps.

    In comparison, 21CN network bandwidth costs us £50/Mbps and 20CN £150/Mbps. Favourite Telco is frankly living back in the 90's when bandwidth costs across the board were around those figures. That or their gear is gold plated and they buy all the cabling from Denon.

    ReplyDelete
    Replies
    1. if if is, that probably explains why people keep nicking it. The more metal that gets replaced with glass, the less will get stolen (eventually, after the thieves learn...which may take a while)

      Delete
  7. The issue here IMHO is that A&A offer a first class service even for applications where a lower level of service would be perfectly acceptable. How about an experiment where customers can assign a lower QoS priority to certain IPs (e.g. kids computers, set top boxes)? Assuming A&A only get charged based on peak usage (97% percentile or similar), then A&A could then drop packets to these low priority IPs at peak times to ensure the interconnects with BT don't hit new peaks. I'm not sure how long the true peaks last but it might provide a way to allow customers to "opt-in" to higher contension and lower throughput. Otherwise there's little I can do other than tell my family not to stream video during the holidays until after 6pm.

    ReplyDelete
    Replies
    1. It is tricky - the problem is that these services work where there are enough customers to make things average. We could not sensible do a small scale trial and see any advantage. Its 100th percentile, by the way.

      Delete
    2. This comment has been removed by the author.

      Delete
    3. This begs the question of the sample/billing period. Is that the 100th percentile of each hour/day/week or month?

      This feels like an area where some innovation is required and where the FB6000s may give A&A a unique opportunity.

      What about a simple web service where I can query the current utilization (averaged over 60 seconds), as a percentage of A&As commit or true peak over the past month? I could query this from a cron and use the information to implement local shaping or to schedule downloads. This would help me to make more informed choices with finer granularity than 9am-6pm periods and would essentially provide a feedback loop to control utilization. A&A benefit from reduced peaks for a very low implementation cost as I'm sure quite a few customers would be only to happy to reduce A&As overheads.

      The hard bit would be to find a way to incentive customers to be "good citizens" beyond the altruistic desire to help A&A. One way would be to redefine "peak" to be those periods where the web service returns > 80%.

      It's rather sad that today I have the fastest broadband ever at 40/10Mbit that I'm rather too scared (or too poor), to really use in creative and lifestyle changing ways.

      Delete
    4. Whereas the rest of the ISP world works 95th percentile with 5 minute samples on a monthly basis; BTW works 100th percentile with 15 minute samples on a monthly basis. And you buy a committed rate and have a burst allowance (5% to 50%) from that committed rate, with burst being charged at a vastly higher rates than the committed. (around £180/meg for burst IPStream, for example)

      Delete

Comments are moderated purely to filter out obvious spam, but it means they may not show immediately.

Missing unix/linux/posix file open option

What I would like is a file open option for "create replacement file". The idea is that this makes a new inode in the same mount p...