Thursday, July 21st, 2011 by Simon Hackett
The National Broadband Network (NBN) is the subject of promises from the government that consumers will pay comparable prices to current day ADSL2+ and phone service bundles in order to access entry level NBN based services, and that NBN based retail pricing will be nationally uniform.
Unfortunately, a number of pressure points in the wholesale pricing model exist which will make these promises (from the government) untenable in practice, unless serious issues with the underlying pricing model are addressed by NBNCo and the ACCC.
This post elaborates on some of the relevant issues that serve to place upward pressure on NBN based retail pricing in general and even more pressure upon retail pricing in regional areas in particular.
A focus on the end-state (ten years from now) in terms of wholesale pricing without apparent consideration of how the industry needs to get ‘there’ from here’ means that NBNCo have, in my view, not yet properly addressed the full lifetime impact of the costs of participation by RSPs across the build period.
In particular the NBNCo “CVC” cost construct (when sufficient CVC capacity is installed for adequate customer service quality) will generate huge monthly ‘overhead’ costs for RSPs (ahead of sufficient customers being connected to defray those costs), for several years, cumulatively and separately in each point of interconnect.
A simple change to the pricing model (first 200 megabits of CVC included at no added cost) would solve this almost entirely.
Solving this issue is essential to ensuring an adequate participation rate by RSPs in the NBN, and to ensuring consumer pricing is not driven far higher than it would otherwise be driven during the first several years of the NBN’s build phase.
Second, the ACCC’s ’121 POI’ decision is fundamentally at odds with stated government policy in terms of consumer retail pricing outcomes, it will drive the continued market dominance of Telstra (as the only party not exposed to the resulting additional costs), and it will cause all consumers to pay more for their Internet access as a result. Hence this decision is clearly not in the Long Term Interests of End Users.
The Story In Detail
There are multiple cost challenges in the currently provided wholesale pricing model for the NBN that will be noted in turn. It is important to appreciate that these issues are additive – and hence that their cumulative effect is a greater issue than each component issue may suggest.
These are merely the highlights of a more complicated story. Everything about NBN pricing and access, at some level, is… complicated.
NBNCo’s CVC pricing model
The ‘CVC’ (Concentrating Virtual Circuit) is an arbitrary $20 per megabit per month fee in the NBNCo wholesale model, charged at the point where retail service providers (RSPs) physically connect to the NBN in a Point of Interconnect (POI).
This financial construct is a ‘collar’ that constrains the size of each aggregated access connection into the NBN based on the monthly sum of money an RSP pays for that access.
It is not a charge based on real costs; Rather, the quantum of this charge has simply been chosen to fill in an otherwise huge hole in the federal government policy requirement that the network return funds to the commonwealth at a commercial rate and in a short timeframe (relative to the expected lifetime of the network).
The most serious problem with this charge is one of the effective ‘bootstrap’ or ‘overhead’ charges that will punitively impact the expenses of RSPs over the ten year build and growth period of the network.
Here’s how it works…
Each CVC connects to a geographic area in Australia (one of the 121 that the network has now been split into by the ACCC).
To ensure that services can be delivered to customers at an acceptable quality, each of these CVC connections has to be purchased with at least 200 megabits of initial capacity.
Why 200 megabits?
Simply, so that more than one customer at once is capable of achieving the advertised 100 megabit download speeds the fibre can provide.
This results in a need to have 100 megabits to allow a any 100 megabit service to work, plus another 100 megabits of ‘burst’ capacity (for a total of 200 megabits) to allow for the transient needs of more than one customer using the network at once!
This is not merely a theoretical issue. Internode discovered the necessity of this in practice during its pioneering work in the Tasmanian NBN initial deployment. We operated with a 100 megabit CVC size at first, and rapidly ran into unacceptable performance issues with our 100 megabit connected school customers. These problems were instantly and permanently solved by upgrading our CVC in Tasmania to 200 megabits per second, a level at which it has since remained through the subsequent growth in customer base that we serve on the NBN in Tasmania.
Once deployed, 200 megabits on a CVC can actually support quite a lot of customers, based on whatever contention ratio an RSP chooses, until the RSP ultimately upgrades beyond 200 megabits in accordance with its individual policy in terms of customer POI access point contention ratio.
Each 200 megabit CVC will cost 200 x $20 = $4000 per month (at wholesale, plus GST) to connect (for an effective rate of almost half a million dollars per month for an RSP connected to all 121 points of interconnect) – before a single paying customer is connected.
There turn out to be multiple ways in which this overhead generates quite irrational effective wholesale costs ‘per customer’; When I first detailed this issue in March 2011, I concentrated on the issue in terms of looking at how large a national provider had to be, before these overheads stopped being insane.
(That answer is: more than 250,000 customers need to be connected nationally before these overheads are no longer a barrier to entry for an RSP to using the NBN directly, meaning that smaller RSPs will be the permanent downstream customers of aggregators, and hence operating with worse operating costs than the larger players)
But it turns out that this is actually not the worst problem with these overheads! Courtesy of the ACCC’s 121 POI decision, another issue about ‘being too small’ has emerged – for NBNCo itself!
This new overhead problem is created not by the size of each ISP so much as by the fact that in each of these 121 sub-networks, the absolute size of the NBN itself will be very small indeed for a number of years; So small in fact that the overhead of these 200 megabit CVC’s (necessary for adequate initial customer performance) will be irrationally high for many years.
For instance, the first Greenfields areas are close to being switched on by NBNCo, and each of these is in a different ACCC determined geographic area (a different POI) and hence each will require a separate 200 megabit CVC to connect to it.
A year from now, we expect (generously) to have only a few hundred houses capable of being connected, in total, for all RSPs, in each of these greenfields areas.
The issue is then that even if a retail provider managed to capture, say, 10% of the entire market in these areas, that means 10% of (say) only around 200 homes at best, implying that the CVC will be serving only about 20 customers.
(Of course, if one ISP managed to become 100% dominant, it’d do far better in effective terms – and this fact illustrates the falsehood of government and NBNCo assurances that the pricing model doesn’t have volume discounts in it. It very much does have them – embedded in the capacity of a larger or dominant player to work through these overheads faster than a smaller player).
Anyway – at $4000 per month minimum, that makes this arbitrary CVC wholesale cost equate to $200 per customer per month (plus port costs!).
Unless NBNCo address this, they will have a massive financial windfall in theory, or the abandonment of participation by RSPs in practice, until there are thousands (not hundreds) of addressable premises connected downstream of each of these CVC’s (which is the point at which the 200 megabit initial CVC allocation starts to be at least a little more rational in terms of average cost per customer).
Also, the previously observed benefit for larger RSPs (a lower effective per customer cost because they can absorb the overheads sooner) also needs to be addressed in order to generate equity of participation via a genuinely level playing field on per user costs.
We have proposed various solutions to this dilemma to NBNCo and they are currently considering our propositions.
Here are our two key propositions:
1) Provide the first 200 megabits per CVC at no additional cost
The most obvious approach to allow RSPs to participate in the network without insane and unintentional overhead costs is simply to remove this unintended overhead structure from the cost model.
Thats extremely simple to do – NBNCo merely need to provide the first 200 megabits of CVC capacity free of charge, and to impose the $20 per megabit CVC charge only for all CVC expansion above that initial 200 megabit size. NBNCo have a stated intention of reducing the CVC ‘per megabit’ rate in later years of the network lifetime, and delivering 200 megabits per CVC without initial charge would simply mean that the point at which CVC per-megabit rates reduced would be delayed to a reciprocal extent.
2) Provide technical burst capacity in each CVC and use 95th percentile charging for actual utilisation
This alternative approach is to charge for CVC usage based on something called ‘95th percentile‘ charging – in effect, to charge on the average real world usage in the CVC by customers, and to provide as much CVC size as each RSP technically requires above that average for adequate performance (which winds up being… 200 megabits minimum and then at least 100 megabits of burst above the observed usage peak assessed).
There are other approaches too, but by far the simplest is the provision of the first 200 megabits per CVC free of charge (on an ongoing basis) by NBNCo.
This eliminates the punitive overhead (per CVC and per POI) imposed on RSPs by the current model.
This then delivers a genuine level playing field in terms of effective per-user wholesale access cost for RSPs by removing the punitive long-term cost overhead that will otherwise exist.
And it does this without making things more complicated for anyone concerned – its the same cost structure, but with a specific change to make it financially rational during the initial decade of the NBN build.
It would remove a windfall cost component that has a plethora of negative impacts on RSP participation and that will place substantial uppward pressure on retail prices from providers.
… and its got to be fixed, if we are to expect rational retail service provider participation in new development areas (greenfields) as well as in the major areas of existing housing (brownfields) that the NBN will spend the coming ten years overbuilding with fibre.
The ACCC ’121 Points Of Interconnect’ Decision
This decision, to force NBNCo away from operating 14 (7 pairs) of redundant major points of interconnect, and instead to have to break the national network up into 121 regionally distributed sub-networks, has deep implications that just seem to get worse the longer we look at them.
Here are just some of the highlights of this decision and its impact:
Backhaul costs from 121 points of interconnect back to the metro core of each RSP will cost a lot of money
This money will necessarily be added on to retail service costs once NBNCo starts to force its RSP’s to move away from the current ‘temporary’ points of interconnect (in metro capital cities) and makes them re-connect at the geographically distributed POI locations the ACCC have determined.
The pricing initially released by Internode for NBN commercial services does not take these additional backhaul costs into account.
Thats because we simply don’t know what they will cost, other than being quite certain that the costs will be non-trivial. Hence once those costs are clear, we expect them to drive increases in our subsequent retail pricing.
The costs to reach POI’s that are more distant will be higher than the costs to reach POI’s distributed within a major capital city area.
That means that it’ll cost more to service a regional customer.
That in turn is what will break the federal government promise of uniform national retail pricing. Because the underlying source pricing will not be uniform across all geographic areas (as it would have been in the original 14 POI model), the higher regional costs will naturally flow through to regionally specific higher retail pricing for consumers and/or to the abandonment of servicing those higher cost regional areas at all.
In a competitive industry, and since there will be no fixed line alternative to the NBN to use instead, pressures on retail pricing in metro areas will make it unreasonable to expect retail service providers to drive their metro pricing yet higher in order to cross subsidise the bush.
This cross subsidy was inherent in, and automatic by design, in the 14 POI model. Moving away from that model moves the NBN away from national uniform retail pricing.
Unless the government decides to take the initiative here and restores the 14 POI access model, regional consumers are likely to pay more for their Internet access than they otherwise could.
The overheads created per CVC that have been previously noted are made substantially worse by this decision
CVC’s will be far less likely to be efficiently ‘filled’ by RSPs because the 121 POI decision slices up the country geographically in a manner that works against that outcome. By splitting up the customer base in this manner, it makes it more likely that ‘part filled’ CVC’s will remain ‘part filled’ and hence will cost more to operate.
As one other technical side effect of this model, the network will also be less reliable. Thats because the 121 POI’s are separate, and if the single attachment point to one POI fails (e.g. due to a natural disaster), there is no way for services to be transparently failed-over via another POI. Back in the 14 POI model, there was the capability to have one of the two metro core POI’s fail, and to have all customer services transparently switched to the backup POI. This all disappears in the name of ‘progress’ in the 121 POI model.
Its important to note (and we’ll elaborate on this in a future blog post) that Telstra gain a huge economic advantage on a permanent basis from the 121 POI model. Thats simply because they own fibre to every POI already, and almost every POI is likely to be inside, or right beside, a Telstra owned exchange building.