Monday, September 8, 2014

Ask the Wrong Question, Get the Wrong Answer

The latest "review" of the NBN,a "Cost Benefit Analysis" is finally out and is predictably anti-FTTP, strongly at odds with reality.

Last year, the Financial Review was very clear that NBN Co was "on track" re-enforced by Mike Quigley, in accepting the highest award in his industry, laying out NBN Co achievements and their good financial performance and only a couple of days ago a story broke that new internal reports show that the FTTP rollout was cheaper than the retrograde FTTN.

Peter Martin, Economics Editor for Fairfax in this piece, falls prey to the dual Occupational Hazards of Journalists and Economics: skimming over complex topics and believing their economics knowledge uncovers all answers, even for subtle technical areas they've not sought specialist advice on.  Martin is wrong, massively so, in this piece.
NBN cost benefit analysis finds there's such a thing as too much speed
Martin makes a fundamental attribution error in the article and never acknowledge his assumptions. His thesis conflates a single parameter, "speed" to be the sole benefit, sole consumer driver of Broadband use. This is a ludicrous proposition, it asks completely the wrong question and gets a wrong answer.
Martin doesn't ever question the "independence" of the Review, nor the validity of their data and conclusions. This article blindly accepts party political bias and puffery. To me, this is very poor, uncritical journalism. It accepts obviously concocted figures as "fact".
The actual, economic questions are:
  • What are the important classes of users, market segments, their differences in the product preferences, their "Value in Use" of the product and both the revenue and profitability from those market segments?
    • Are their additional technical metrics that some customer segments value over 'speed'? 
    • Do the offered products sell to those market segments, maximise the revenue-per-customer, and minimise "Consumer Surplus"? These are all necessary to maximise profits.
    • Are all consumers equally profitable and can the super-profitable segments be pre-identified?
    • What is the aggregate, and segment specific, price elasticity of consumers?
    • What are the Fixed and Variable costs of each delivery option versus the revenues?
    • What are the Marginal Revenue and Costs figures for each line-of-business?
  • What is the best source of data for predicting future aggregate network demand and also for new services and demand from the most profitable, and/or highest revenue market segments.
Martin starts with an outright fallacy:
No wonder Labor never wanted its National Broadband Network subjected to cost benefit analysis.
This is wrong at three levels:
  • The Abbott Coalition doesn't subscribe to Cost Benefit Analyses everywhere, in fact only for the NBN. If they did, we'd see them done as a matter of course, at least for every programme that spends more than $1 billion of taxpayer money. The NBN "CBA" is not accurate, not independent and entirely hypocritical of the Liberals.
  • CBA's are done to choose between non-profitable project options, before a project starts, not to review in-progress projects, and not to assess the financial viability of businesses, like NBN Co. Big companies don't run CBA's, they do financial analyses, with sensitivity tests and risk assessment, for multiple scenarios, to arrive at NPV, IRR, break-even and funding/cash-flow profile, all these were known and on-track or better for NBN Co. 
The most important question is about the central study commissioned for this review, the Demand Forecast. This is performed by economists, without reference to experienced Network Engineers, Traffic and Network load forecasting experts and real NBN customer data. With 100,000+ NBN customers, there is now definitive data on preferences, Willingness to Pay, consumption patterns and more.

The most erroneous and misleading assumption by the CBA, never queried in this piece, is breathtaking in the depth and breadth of its ignorance:
All broadband services are domestic and centred around entertainment.
The NBN is of most value to businesses, it will see the highest increased demand, in volume, 'speeds' & other service-options from business and will gain its highest profits from business. The "Party of Business", and this economics writer, seem completely aware that the NBN is all about business.

No successful Telco or ISP, since the web came along in 1991, has used the micro-economic bandwidth-demand modelling technique adopted here.

There are two powerful technical reasons for this:
  • Packet networks are two-way, variable and chaotic in nature, they must be modelled very differently, stochastically, from fixed-rate, predictable circuit-based networks, like two-way telephony and single-direction Cable TV. The CBA modelling erroneously treats all services as simple, isolated circuits where bandwidth is trivially additive, this just isn't so.
    • The simplest example of this fallacy is "No Car needs more than a 5 kilowatt engine".
    • If we take the total km commuted every day (40km) and divide by the time away from home (10 hours), then all cars only need achieve 4-5kph, which is more than adequately achieved with 2-5kW. The inherent fallacies abound:
      • Not nearly all vehicles are domestic, business & transport use is much larger.
      • Not nearly all commuters travel by car.
      • Domestic cars are used for much more than commuting.
      • Most multi-person household have more than one vehicle, used for very different purposes, commuting being a small role.
  • The defining feature of both The Internet and current consumer take-up of technology is that adoption rates, from 10% to 90% penetration, are exponential and now blindingly fast. It used to take 50-75 years to reach 90% adoption, the smartphone got there in just 7 years. Not only don't we know what applications and services people will be using, or demanding support for, in 5 years time, we know they are "unknown unknowns" and we must allow for them.
    • In 2007, the iPhone blind-sided every mobile service provider, causing them to scrabble as hard as they could building G3 & G4 mobile data networks.
    • In 1984, the year G3 fax 'broke-through' in business, international telephony almost broke. Voice call minutes for business collapsed, while the modems in faxes defeated the "voice multiplication" and carrier systems. Revenues plummeted while direct costs per minute-carried multiplied 3-10 times. No Telco had predicted the year of the fax.
    • Every prediction of Internet demand since 1995 has been low. The continued exponential growth in demand for both raw access bandwidth and aggregate download volume has been the hallmark of this industry for 20 years. Yet the CBA presents no case as to why this suddenly won't be the case, or just in Australia, known globally for its rapid and enthusiastic adoption of new technologies. At some point, we can expect Internet data demand to reach maturity, but nobody credible has posited when. We're still more than a decade away from there.
Definitive research on the increasing rate of change of technology adoption, from Horace Dediu of Asymco: A 34min video and the charts to go with it.

The Next Big Thing will blind-side everyone in business. Some will thrive, some die, depending on how quickly they can notice & respond.  This fundamental is something everyone on the Coalition side ignores, reflected in spades in the ridiculous and lamentable CBA and it's absurd Demand Forecasting. CISCO invests a lot of money in its VNI reporting precisely because Internet Demand Forecasting is hard & not easily predicted, and it is vitally important to its customers, Telcos, because of the relatively long lead-times compared to growth rates.

Amazingly, the report dismisses CISCO's exponential demand forecasts out of hand. What do these geniuses know that the technical experts at CISCO and all the Tier 1 carriers around the world don't?

There's a simple test: if this modelling was credible and better than CISCO's, it would be the hottest property in the Telco market at the moment, being sold at a massive premium. The company that did it would be so over-whelmed with customers they would be the #1 success story this year in the Australian business world.

Look to what the people who make their money from buying good-enough traffic forecasts are saying and doing: this Demand Forecast has been roundly ignored by the Networking Engineering experts. If the Industry isn't abuzz with a radically different demand forecast like this, then it is unsound and not credible.

Martin then blasts us with this inanity:
Bandwidth that isn't used is bandwidth that's wasted. If it costs a lot to get it it's money wasted.
He later quotes, as if fact, a completely fantastical estimate included in the CBA.
The Coalition scheme will cost a net $620 per person in today's dollars. Labor's would have cost $2200.
The Node to Premises connection is not the dominant cost of a whole network, or the newly modified Turnbull NBN would have a total cost of 25%-33% the FTTP NBN. Martin fails to understand that economic difference: the cost of one component doesn't equate to the full cost of the system.

There are two related issues to this, about Apples-to-Apples comparisons:
  • Both the CBA and Martin don't include the cost of deliberate wastage in the FTTN. In the pre-election Turnbull plan, there was assumed to be "50% CapEx Reuse", i.e. they always planned to throw away half the taxpayer money they'd use. On their own figures, this wasteful detour via FTTN will cost $4-$8 billion dollars, yet that cost doesn't appear in the CBA.
  • The in-premises equipment for FTTP will be supplied, maintained, replaced and even upgraded by NBN Co. But nowhere, not pre-election nor in the FTTN pilot Product Specification is this cost mentioned. The pilot study states the VDSL splitter and face-plate will be installed & owned by NBN Co, presumably at their cost. This leaves the customer to pay $250 or more for a VDSL router/firewall and to bear the entire maintenance, repair and replacement cost of that over 25 years. This will have an NPV of around $400-$500, yet this is omitted by the CBA and Martin.
    • There is also an issue of equity in this: should all NBN Co customers be treated equally and have access to the same 4-port NTD, wholly supplied, maintained and upgraded by NBN Co.?
Martin clearly doesn't know the world of Telcos, but he should be aware that every valid economic analysis has to be on a like-for-like basis. The real cost of the FTTN connections is around $1800, when deliberate wastage, costs transferred to users and on-going maintenance is considered. Can anyone argue that a saving of $350 per household is worthwhile against the insurance of future-proofing every fixed-line Telco business in Australia?

This is without considering the contribution to profitability Fibre will make.

The CBA, and Martin, blithely assume that only costs matter in a business.
Here's a NewsFlash: Ledgers have two sides! Is this a problem with Economists generally, or just these ones?

The CBA authors have failed to even consider the extra revenue, and profits, available from FTTP.

There's a lot more to it that 'speed'. Margins increase, and very quickly, with high-end customers. Download volumes are exponentially distributed across users, the top 1%, combined, download 50% more data than the bottom 50%, but the input costs to NBN Co are the same for all users. NBN Co makes many times more profit from the high-end users, but this is ignored by the CBA and Martin.

There's a potential 75-fold difference in profit to be unlocked from Fibre from high-end users.

Being able to remove the Consumer Surplus by charging many-times more for the same product while charging customers a lower per-unit price ($/Mbps), is genius economics. Because costs are the same, profits, the difference between revenue and expenses, increase many fold. This financial difference goes unremarked by these economists, why?

The ugly truth is that NBN Co could make more profit from an FTTP by only serving the top 25% of customers. The high-end users create all NBN Co profits, the rest of us get a glorious free-ride, either wholesale services at cost, or heavily subsidised. The gamers will subsidise everyone else, not the reverse. But "gamers" are far from all the high-end users, seemingly unknown to these economists.

This business fundamental that "revenues aren't profits" seems beyond the ken of both the 'independent' CBA writers and Martin. The reason Fibre can make a profit, while an FTTN will struggle is simple and elegant: by making the full access rate range available to everyone, all the high-profit users can be tapped. The problem for all Telcos is that you cannot pre-identify who your most profitable customers will be, when they'll become high-end, and where they'll be.

You need to rollout the same service everywhere, then only over time you, and your customers will discover their true demand, elasticity, price-points & economic preferences.

The marginal cost to NBN Co on FTTP service between an entry-level 12/1 Mbps service and the current top-end 1000/400 Mbps service is zero.  For an 80-fold difference, it costs NBN Co nothing extra to remove the throttling on individual services. The 32:1 multiplexing of GPON services means that NBN Co may send out a tech to repatch connections when the total demand on a single, shared GPON uplink exceeds the 2Gbps (yes, 2Gbps, not 1Gbps) available.

This is why they are deploying so many extra fibres at every level of the network. The cost difference between laying 2 fibres and 864 (6 tubes of 12 ribbons of 12 fibres) is around $2/m, or from $8/m to $10/m. The first two fibres cost you $4/m, the next ~1000 cost you $0.00232/m, or 2150 times less, each. That extra built capacity isn't quite free, but it's so close, in practice it can be considered free.

Meanwhile, Turnbull et al are starting with a technology that attempts to push 1925-spec copper 25,000 times past it's design limit. The physics of Radio transmission over copper for xDSL, are very well understood. You can't break the bandwidth/distance ratio: the only way to double the speed now is to halve the distance, with a quadrupling of the number of nodes needed. You can have gigabit xDSL over 1925-spec copper, all the way to your gate. And if it rains, it stops dead. High frequencies are carried on an increasingly thin 'skin' of the copper wire and will be shorted to earth at the first hint of moisture. Joint corrosion also has a highly non-linear impact as frequencies increase.

The other side to this is that the physics of data transmission in optical fibre does have limits, it does obey power/rate and noise/signal laws and these limits are being explored. The physics limiting data rates over optical fibre start to kick in at the terabit/sec or petabit/sec level. The extra power to drive more bits/second starts to affect the glass and ways to compensate are being explored. This will only ever affect the internal trunks of Telcos, never customers.

Ordinary consumers will never need to deal with the physical limits of optical fibre, apart from the current distance limits dictated by the "loss budget". The glass in optical fibre is amazingly clear, but not infinitely so. After 40-100km, you need to regenerate the signal on customer-grade systems. For over a decade, international optical fibres of 1,000+km, underwater, have been repeater-free, something never achieved with copper, not matter what specification.

Right now, off the shelf, 10Gbps multi-colour (CWDM & DWDM) optical transceivers are a commodity item, while CISCO can sell you 100Gbps DWDM transceivers supporting 96 simultaneous streams down a single fibre, for 9.6Tbps capacity. It's used by NBN Co high in their network. We know that volume production drives down the prices of Silicon parts. As demand grows, 100Gbps transceivers will become cheap commodity items, just as every mass market silicon chip has for the last 50 years. There's an iron-clad economic law, Learning Curve, if you're looking for one.

Summary: the next upgrade of the FTTP network will see massive speeds, expect 40Gbps or 100Gbps vs current 2Gpbs native rates, for the same price, or cheaper, electronics. It will be quick, cheap and simple to implement: just change out the GPON ONT's, nothing else. This low-cost upgradability is not part of the CBA, just like the serious omission of "unknown unknowns".

FTTN and broadcast wireless not only have limited technical lifetimes, they are starting off near their physical limits. They are a very, very poor investment if you need a 25 year service life. That Martin has failed to grasp this simple, fundamental of economics and electronics is truly amazing to me: there is NO seller cost for extra consumer bandwidth in FTTP. It's the central difference between the FTTN and FTTP, the zero marginal cost of additional bandwidth over a range of 80-times, versus the FTTN's guaranteed range of 2:1 (12/1 Mbps to 25/5Mbps). If we compare like-for-like, then we can only use rates guaranteed everywhere, not "best possible" available to the lucky few.

Couple to that the current upgrade path for high-end users, another 9,600 times, the known commodity price curve for volume production of Silicon chips and NextGen FTTP will cost less than now and go 20-50 times faster. xDSL, FTTN and broadcast wireless technologies can't come close, they are in a cost region where prices, absolute or per-unit, can only increase.

The problem with both the naive and blinkered Coalition CBA, and Martin's blind acceptance of it, is that it ignores the central problem of all Telco's since the 1880's:
laying out and dimensioning the network must be able to support most, not every, scenario of demand & provisioning for the end of its service life, a minimum of 25 years. What works today is simply not good enough in the Telco world. It has to keep working until it wears out.
Both the Vertigan/Ergas CBA and Martin's analysis of it fail to deal adequately with what happens in 25 and 50 years when the current Fibre will still be in service. Even for the average user, the guaranteed 25Mbps of FTTN won't be acceptable in 7 years, yet they are arguing that an FTTN with a service life of 25 years will be sufficient. Can we get our money back in 5 years when this foolish idea blows up? No, we have one chance to get it right, and that's before we build it.

If you don't design it in, it won't show up: Network Upgrades hold you solidly to this rule, which is exactly why we have been having a decade-long public debate over Broadband. Since 2004, it's been blindingly obvious that Australia needed better Customer Broadband services, coming with a substantial lead-time, yet nobody was building them. Planning is everything for Telcos who survive.

Because Telstra didn't invest and build a fibre network when it could, one that was designed for the future that we now inhabit, we now have a major headache. This legacy of inaction and failure to anticipate was created by the Howard Government. Frank Blount, Telstra's first CEO in 1992, is on-record as wanting to replace all customer copper with Fibre by 2010 The project never went ahead even though it would've been internally funded. That was a direct failure of the Government of the day: John Howard, Tony Abbott, Malcolm Turnbull and most of the cabinet.

Why isn't Martin reporting the fundamentals? I've always thought he was clear-headed and insightful, but apparently, not on this topic.

More on refuting the reseasons for an NBN Co Cost Benefit Analysis.
  • The NBN CBA is entirely political posturing, unrelated to financial or economic causes. If the Abbott Coalition actually believed in proper, unbiased, detailed economic analyses, we'd have seen them done and published, as a matter of course, for every single programme that spends over $1 billion of taxpayer money.
    • The Howard government was notorious for ignoring advice and never commissioning Cost Benefit Analyses. The Abbott/Hockey government are following this example.
    • The relative impact of the NBN is many-fold higher for non-urban and regional/remote population, while the build costs, due to distance, are higher. The business, health, education and administrative benefits to "the bush" are massive, but ignored.
      • As one of the most urban-dwelling countries, there are at least two different and important sides of the "infrastructure" debates:
        • Who gets what in Urban areas, and
        • What we're prepared to pay to benefit those in The Bush.
    • The Abbott "signature policy", Paid Parental Leave (really a $50,000 'wealthy women of substance' baby bonus) has seen no CBA.
    • The Mining Tax, the Carbon Pricing System, the RET and more policies they've axed had no CBA done by Abbott & Hockey. Why? "Just Because?"
    • Abbott as "Infrastructure Prime Minister" has ignored completely CBA's for the roads and freeways projects he wishes to back, and refused to look at the evidence that more strongly supports alternate public projects, such as commuter rail, long-distance rail freight and teleworking.
      • The only people that benefit from large, public investments in expensive freeways are those that can afford to commute in cars: typical the upper quartile of income.
  • David Braue has done a very good job of refuting the Libs mantra "the NBN needs a CBA" saying it was both mere political puffery and factually incorrect.
    • The Expert Review Panel of the FTTN tenders was a CBA for the Libs current policy. The panel had just one 'technical' person. On purely economic and business grounds, by 2008, with a LNP-created vertically integrated private monopoly, Telstra, cutting copper you didn't own and inserting complex, expensive electronics you planned to throw away before they'd paid for themselves, was completely insane on any basis.
    • The only reason Labor had to invent & fund NBN Co, was because Abbott, Hockey and Turnbull, personally, had approved the sale of an uncontrolled and unseparated Telstra.
      • Not only did Howard oversee the destruction of Telstra's business,
      • he didn't invest in necessary projects to secure the future of Telstra,
      • he deliberately didn't force structural separation, and
      • deliberately created an uncontrolled and uncontrollable monopoly which his own expert party members described in print as "A Monster".
    • The only reason we don't have good broadband, especially in our most profitable urban areas, is because of a profound Market Failure in Telecommunications, overseen, fostered and created by the Coalition itself.
      • We know this started in the early 1990's with the Optus/Telstra Cable TV overbuilds.
      • Overbuilding extended to GSM, G3 and G4 mobile networks, now mobile data networks, when the initial wholesale AMPS network had delivered solid financial results.
      • Telstra used RIM's to lock out line competition, including delivering not broadband to customers.
      • The ADSL1 DSLAM's of Telstra were forced open by the ACCC. After that, every Telso/ISP had to 
    • The Howard Government was pitched, in 2005, a very cheap, very timely FTTN NBN by Sol Trujillo, adequately backed by financial models.
      • ADSL2+ and VDSL1/2 were good initial technologies in 2005, but not now.
      • The Government controlled Telstra in 2005, and still controlled the copper.
      • Howard & Costello were awash with cash, they had a $300 Billion windfall from the Chinese Mining Boom.
    • The Abbott Opposition was unmatched in fomenting unrest, in phrasing three-word slogans, in dividing the community, in fostering and furthering vitriolic, vicious and personal attacks and propagating many incorrect and unjustified positions, including the fiction of a "Cost Benefit Analysis".
      • The passion and vitriol surrounding the NBN, a simple and necessary technical upgrade for our 1925's network, and not that expensive at $2,250 per premises, has solely been manufactured by the Libs for a political purpose.
  • CBA's are done to choose between options, before a project start, not to review in-progress projects, and not to assess the financial viability of businesses, like NBN Co.
    • NBN Co is a business, not simply on-budget expenditure that has no identifiable and direct returns, like roads.
      • NBN Co is an investment, it returns all the money given it, plus a handsome profit.
      • It's not just throwing taxpayer money away, but making money, every year after it breaks-even.
      • The NBN project always had good business justifications, with public, independent, detailed and highly credible Business Plan and published data on project progress.
    • CBA's are economic analyses, not financial, looking for non-financial benefits of projects which don't make financial returns. You only revert to a CBA if you want to justify losing money, not, like NBN Co, when you'll be making $50-$100 billion in profits.
    • If CBA's, not NPV/IRR financial analyses, were the best tool to assess business projects, then we'd see every single Big Business doing them and publishing them in their Annual Reports. The ASX and ASIC would demand CBA's for business, but they don't.
    • The standard financial analyses of different scenarios were well played out in the NBN Co network design.
      • The Abbott Opposition even focussed on the changing design as NBN Co did the detailed work

More detail on the Economics of the Business of NBN Co:
  • What are the important classes of users, the differences in the product preferences, their "Value in Use" of the product and both the revenue and profitability from those market segments?
    • Are their additional technical metrics that some customer segments value over? The usual network variables are:
      • Predictable or Uniform Service Availability across regions.
      • Guaranteed minimum service characteristics, including absence of source congestion due to under-dimensioned Node uplinks.
      • Mean Time between Failures and Mean Time to Repair. Failure rates related to external factors, like weather or power interference.
      • Latency and Jitter, both fundamental to Video Quality and especially interactive services.
      • Congestion control, especially during Peak Demand. The availability of premium priced, higher Quality of Service.
      • Availability and Reliability. Are better service levels supported or available?
      • Additional services, like Multicast, peer-peer support,  symmetric links or customer network hosting.
      • Upgrade paths, upgrade availability  and incremental service pricing. Will a small business be able to upgrade their service in-place as their service requirement increase.
      • When, not if, new Internet products and services are available, will all customers be able to access them? Is the network design intended to create un-exploited "pent-up demand" like the current ADSL shambles?
    • Do the offered products recognise those market segments, maximise the price-per-customer, and minimise "Consumer Surplus"? These are all necessary to maximise profits.
    • Are there demand anomalies where some segments may be high revenue but zero or low margin, while small segments, potentially invisible, are financially most significant.
    • Are there limiting secondary factors at play, such as the inability to distinguish high-profit subscribers beforehand, to adequately forecast aggregate demand and demand peaks, or to properly predict demand by area?
    • What is the aggregate, and segment specific, price elasticity of consumers?
    • What are the Fixed and Variable costs of each delivery option?
      • How can market pricing maximise both revenues and profits from both the current and forecast demand?
    • What are the Marginal Revenue and Costs figures for each line-of-business?
      • Differentiating customers by access rates is not good business.
      • Usage patterns (slow & constant vs real-time vs 'bursty') and perceived Service Quality radically segment the market, even for identical access rate customers.