« Pray it doesn’t take off | Main | Feedburner: lessons learned »
Beyond the headlines
By jpenston | May 3, 2007 |
|
My £2.10 headline cost for 2 hours of HD TV is a simplification - call it an exec summary if you will. It ignores the key dynamic in telecoms: scale.
£2.10 irons out the granularity to come up with something that makes a point rather than painting the whole picture. A number of people have questioned the number and some have pointed out that the assumptions are questionable, so today I am writing a more detailed article on the subject. If detailed numbers isn’t your thing, come back another day when there will be something less turgid on the blog.
The £2.10 headline cost is based on the exchange being connected by 100M and not Gig-E, which can be 4 times cheaper per bps than a 100M version - assuming the pipes are both full. If you have a requirement for 3 x 100Mbps connections, you should get a 1Gbps connection as they are typically around 2.5x the price of a 100Mbps one.
If you are serious about TV, then you will need a baseline service at least equivalent to the Freeview Digital Terrestrial TV service. You need at least 100 channels to have a multichannel TV service worth the name. Carrying that using multicast streams at 2Mbps per channel adds as much as 200Mbps of multicast capacity regardless of how many subscribers you have. Add onto that a few uncontended business customers and you go over the breakeven point of 200Mbps where it is actually cheaper to buy a Gigabit link than 3 x 100Mbps ones.
On a gigabit link, the £2.10 headline number becomes £1.36. This breaks down into £1.03 for the transit (still) plus 33p for backhaul. If you are serving the content on-net (ie avoiding Transit) then you only have the 33p cost for the backhaul for that on demand, unicasted 9 GB / 2 hour 1080p file.
Transit
The Transit portion is clearly then the biggest chunk of cost. It is only relevant where the content is coming from the internet: where the consumer is getting video services from the likes of iTunes, YouTube or Joost.
This point highlights the commercial imperative to place content servers “inside” the ISPs networks. This can be achieved via a number of different commercial and / or technical solutions such as those highlighted in my Routing the Money and Routing the Packets posts a few weeks back. They key here is that content distributors and ISPs need to work together to make sure that the best possible service (resolution) can be delivered efficiently and revenue shared to pay for the investment in networks.
The alternative is Mutually Assured Destruction (MAD), where ISPs spend ever more money to block as much traffic as they can (because they have to control their costs) and content distributors spend ever more money to make sure that their services bypass these controls. No-one wins a MAD war, except perhaps the broadcast networks.
To be fair to Joost though, it should be noted that they are keen to get “inside” the networks to remove this cost from the distribution chain. It is definitely worthwhile reading my earlier analysis of how Joost delivers traffic and Colm MacCarthaigh’s piece on the Joost network as background here.
The transit portion is not relevant for the ISPs with their own IPTV services because they are already hosting the content within their own networks. Their philosophy is to control the content and the user by developing end to end products, from set top box to managing the content relationships.
This “walled garden” approach raises a lot of questions. The industry has tried walled gardens before and spectacularly failed as the innovation on the internet side of the transit port has had so much more development resource.
There may be the odd company (PCCW’s NowTV service for example) with enough of a head start and sufficiently aggressive approach may work in some markets – particularly where protection from regulators exists. However, it has struck me over recent months how defensive most telco approaches to IPTV are. I can’t help thinking that if you don’t really want a market, no matter how hard you try to stop others taking it, you will again fail as the industry did with previous attempts to build closed networks.
Anyway, time to move on. The rest of this article assumes that via whatever means, content is delivered from inside the ISPs network and that the transit cost issue is solved. The following cost analysis is therefore for backhaul only.
Multicast
One of the clear criticisms of my £2.10 number is that companies offering IP TV are not just offering video on demand. VOD is a premium service over and above the linear “broadcast” of a basic multichannel TV package.
For a number the big players this is undoubtedly true. On their networks, the requirement to deliver this basic multichannel package drives them to gigabit backhaul from day 1.
So what exactly is the impact of putting multicast TV on the network and using gigabit instead of 100 megabit backhaul? To understand the cost of video on demand on top of this multicast service, you need to lock down the assumption on ISP size (I have selected 20% in what follows next). The following chart shows the backhaul cost of varying amounts of VOD on networks with 1) no underlying multicast TV service, 2) 100 channels of 2Mbps multicast TV service and 3) 100 channels of 10Mbps multicast TV service. The horizontal axis shows the percentage of the population you are covering with your LLU network. The bigger you are the better it is for you because an ISP with 1 subscriber has the same multicast traffic volume as one with 1,000 subscribers on the exchange. In order to show this, the horizontal axis on the next chart shows a single exchange with a variable number of users up to 10,000 users. What I am showing is how the cost per subscriber changes with 10, 40 and 100 hours of video on demand per month. The dotted lines show this cost without the multicast traffic in the background while the solid lines describe what this looks like with 100 channels of standard definition TV multicasted at 2Mbps. Note the largest exchange in the UK covers a little over 60,000 households, the second largest around 45,000 and given 50% penetration, you would need to be a very successful business to have anywhere near 10,000 subs on an exchange. Realistically, even the biggest players only have between 900 and 4,000 or so customers on the exchanges which they will have so far unbundled, although unbundling costs in against BT’s Wholesale product if you have more than 350 or so. Someone with 20% market share is likely to have over that number on around 1,790 exchanges and between them those sites (32% of total exchanges) cover 85% of the population - coincidentally or not, this is about how many the big players are saying that they are going to. It is therefore worthwhile focusing in on the small exchanges and blowing up the horizontal axis so that you can see the dynamic for exchanges with up to 1,000 users. Again, the dotted lines show the cost per subscriber without the 100 channels of multicast. Having that multicast traffic immediately justifies a gigabit connection, regardless of how much or little other traffic there is. The dotted (no multicast) line jumps when the requirement for a second 100M connection kicks in and then converges with the solid line (where the is multicast background traffic) when requirements exceed 200M as it is then most effective to go straight to gigabit rather than getting a third 100M link. You can see also the slight delay in requiring an additional gigabit links on the dotted line because of the missing 200Mbps multicast streams. Next, it is worth looking at the multicast resolution to see what difference that would make if that were increased to 1080p at 10Mbps rather than standard definition at 2Mbps. In these charts, the dotted line shows the multicast traffic as 1080p HD TV and the solid lines the multicast traffic at 2Mbps as above. The chart seems to indicate that for larger exchanges and where VOD is commonplace, the marginal cost of delivering the basic package at 1080p high definition is insignificant. This is not the case for smaller exchanges though where instead of perhaps one 100Mbps connection if you didn’t have a TV service at all, or 1Gbps connection if you were delivering at standard definition, you would need perhaps two gigabit links increasing the cost by a factor of 6.25 over what you would pay if you offered no TV service at all. Given that these channels almost certainly have to be free because they are similarly delivered on free to air digital TV, increasing your cost by this much in order to offer a better base service might be a challenge. I can’t see the finance director liking the idea very much… Scale Economics The chart below shows how a network with 10% market share would need to provision more bandwidth to allow its users to do the same as can be done within the scope of existing assets on smaller networks. This phenomenon shows the difference between fully allocated costs (total costs divided by total subscribers) and incremental costs (the cost of an additional X amount of usage). What you can see on the chart immediately below is the cost of a small amount of incremental traffic. In later charts though you can see that even though the cost of incremental capacity is lower, the fully allocated cost per subscriber is still higher on smaller networks. Before I come back to fully allocated costs, it is worthwhile looking at the incremental cost of 2 hours of VOD on top of the 100 channels of underlying 2Mbps multicast traffic. Because the multicast traffic means you will buy gigabit pipes instead of 100Mbps ones, even if you have 20% market share, 2 hours per subscriber of video on demand will not incur any incremental costs: you don’t need more capacity because your gigabit links only have 200Mbps on them so you can deliver the 2 hours of VOD without need for an upgrade. In fact, 2 hours VOD per subscriber only triggers network upgrades for networks with 50% share or more. Everything on Demand Time shifting can easily be explained. Perhaps you missed something last night and there is nothing on tonight. Instead of boring yourself with something you don’t want, you will catch up on what you did want to see. Or, perhaps your two favourite shows are on at the same time, with time-shifting, you can watch them both. Or, someone phones you up half way through your favourite show… All a bit Sky+, but time-shifting allows you to do this in retrospect rather than having to plan for it as you do with a PVR. If you are sceptical, consider that 100 hours is approximately what the average UK TV viewer watches in total in a month. Consider also Tiscali Home Choice’s figures for its VOD service; their viewers already watch between 15 and 18 hours a week on demand (60 to 72 hours per month). So as the demand for on demand grows to 100 hours, what happens to your smaller provider with no incremental cost for the two hour show that we talked about above? It starts to get a bit ugly in fact - to the point where someone with 1% market share does not fit on the chart. This is a great example of how you can be misled into getting into a market to fill your network without thinking about what happens if the plan works too well. Here you can see the lines crossing where sub-optimal utilisation problems occur, for example when you need just over 1 gigabit and have to pay for two circuits instead of just one. This shows that the biggest exchanges are not necessarily the best (even without the hyper-competition that you get in the hot locations). In general however it shows how deep you can go without blowing your costbase apart, based on how big you are. If you have a 10% share you might stop when you have covered 25% of the market, while if you have 15% share you might go as far as 55% and so on. Summary Do not bet against open networks or internet based content distributors. Work with them rather than fighting them to minimise the transit burden and align revenues with costs. For the big guys, the financial impact is marginal and providing a full range of TV channels makes the service extremely attractive. This is a clear example of how scale wins in telecoms. A “proper” IPTV service will indeed include a significant element of multicast traffic which will drive networks to gigabit backhaul from day 1. This will give you a lot of free capacity to “grow into” and an attractive baseline service if you can get the distribution rights for a DTV equivalent. Do not play this game unless you are a big player or have very strong local geographic niches with a concentrated customer base. If this is you, buy wholesale from others with network, but who are struggling to fill them. Do not get misled into offering on demand services just because you have spare capacity today. Only do this if you have plans in place to grow capacity and monetise usage or you will find yourself praying for failure. Offering a High Definition base package may be good marketing but the cost implications are severe on exchanges with as many as 1,500 users. This will impact even the big players and so is unlikely until pay per view on demand services at that resolution are well established. Do not simply think that bigger is better. In general it is, but beware of step changes. Alternatives to Hara Kiri It is extremely likely that very little of the value associated with video content will remain with the ISP. This is for two main reasons: 1) the power of the content owner and 2) problems of service differentiation against other network companies. So, even if network costs for 100 hours of VOD per month can be kept to around £4 as in the last chart above, it is quite possible that such a service will become yet another zero margin hygiene service. It may be that you are required to offer it to get / keep the customer and yet find it impossible to monetise beyond the bare essentials required to buy the content for resale and cover the distribution cost. So while the TV offering may not be quite such a commodity as a POP3 email account is today – a product that hundreds of ISPs can offer – it may still be a commodity offering for those such as Virgin, Carphone, Tiscali and Orange unless they can each find a way to offer the same (BBC, ITV, Sky, Sony, Warner etc.) content in a unique way. In this environment it makes a lot of sense to focus on how to deliver the hygiene service at the lowest possible cost. What I am saying is that rather than throwing money at the problem and building ever bigger networks as we have in the past, throw innovation at it instead. This could save investors a huge amount of pain and possibly a lot of people’s jobs during the inevitable round of redundancies that follows bankruptcies. There are a few alternatives that I can see today, although the list is growing as the open source innovation resource on the internet starts to focus on the problem of how to deliver large files cost effectively. 1. Caching Caching Local routing + P2P Staggercasting The problem I see is that while a +1 channel may work as a nice little value add, the alternative (true) on demand options offering time-shifting to any point the customer requires cannot be replicated using staggercasting. Staggercasting may be a great free option, it will never generate any revenue and is likely to be beaten by the full implementation of VOD, in my view. Conclusion Finally, I would like to thank you for reading this far! This wasn’t the most fun article I have ever written because of the complexity, all the charts and the lack of a simple message around which to pivot. If you have read it all, well done. I’d love to give you some sort of prize but unfortunately the economics of running a free blog like this don’t really leave much room for making money, let alone giving it away. There’s a parallel there somewhere…
So just how important is scale and market share? This is pretty complex because costs are characterised by step changes that you can see in the above charts. On very small networks (eg. 1% market share), the incremental cost for each subscriber to get 2 hours video on demand is zero. This is because the network is seriously under-utilised and the 2 hours of VOD will not require a capacity upgrade.
All this does not mean that VOD is free of course! What the above charts do is look simply at adding 2 hours of VOD per subscriber per month. To understand the dynamic properly you have to look at what happens when those 2 hours grows to 5, 10, 20 and eventually up to 100 hours where video on demand replaces standard TV. I’m not saying that people will start watching something unique to them, although niche behaviour will undoubtedly happen. What I’m talking about is when time shift behaviour takes over from watching whatever is on.
£2.10 is a simplification – I hope that all this detail has made it clear why a simple headline number was used!
The problem is circular – you need to offer low prices to get the traffic to get the low cost to justify the low prices – and this is a standard market entry strategy, but it is one that very few can afford to indulge in. It also suggests that the market will be price led from day 1, which is a bad thing when so much investment needs to go into building the networks.
2. Local routing + P2P
3. Staggercasting or Near Video on Demand (NVOD)
Looking at the cost of bandwidth vs cost of storage you can then say that if you expect the same file more than 12 times you should cache it and distribute it from within your network, on your side of the transit port (£1.40 per GB stored vs (£1.03)/9 per GB transferred). Looking at doing so at the exchange level and assuming that you only get the marginal benefit from the backhaul and not transit + backhaul, you would need 38 downloads of the same file to make caching pay off even against a Gig-E backhaul link.
Peer to peer as a technology is here to stay. We are now seeing version 2 which is much more content aware than Kazzaa was in version 1. What version 3 will no doubt have that v2 doesn’t today is network awareness built in so that it actually helps rather than hinders content distribution over the internet’s networks. For this to work though, the software developers and the network architects need to get together and work out how to distribute the routing function so that as much traffic as possible only uses the dedicated, fixed price bandwidth on the local loops.
While not being truly on demand, staggercasting gives the service provider the chance to offer an on demand-like service but with a controlled costbase. Consumers are familiar with this due to the +1 TV channels already on the Sky platform so education will not be an issue.
There is a problem, but we are aware of it. Some may bet on bandwidth being so cheap that innovation is the expensive option, but to me that feels like a never ending chase to find the pot of gold at the bottom of the rainbow. In my view, however much bandwidth is deployed, applications will find a way to use it so it must surely be a good idea to make the network’s use of the precious resources as efficient as possible.
Copyright The IP Development Network - www.ipdev.net
Phone +44 1763 81 00 20
Email jpenston@ipdev.net
Topics: Uncategorized |
May 3rd, 2007 at 3:50 pm
Would you please enable full-text feeds? They’re much more convenient for those of us who subscribe to large numbers of feeds.
Many thanks!
May 3rd, 2007 at 9:39 pm
I’ll do this, but let me explain why I have had it set up like this as maybe someone out there can help me out.
My thinking was that if I only enable partial feeds, I’ll be able to see the number of people who actually want to read the article rather than simply scan the headline and then discard because it is not right for them. That will help me write stuff that people want to read.
I know I can count feeds using a feedburner type of service but is there anything better out there that anyone can recommend?
May 10th, 2007 at 8:16 am
Bill Norton gave a very good presentation on the costs of internet traffic and video over the internet.
http://netseminar.stanford.edu/sessions/Internet%20Video%200.1.ppt