Fibre-to-the-House-of-Lords

I've been invited to give oral evidence to the House of Lords inquiry into Superfast. This is sporting of the Committee, since one of their key questions is whether the UK government "is being ambitious enough in its plans" for Superfast. I of course believe the government is being too ambitious.

I've already submitted written evidence to the Committee, so at least they're aware they've invited a dog into the manger.

The instructions for witnesses tell them to find the committee room by turning right at the statue of Joseph Chamberlain. I'm reviewing mug shots of 19th century politicians to be sure I recognise him when I see him.

Do consumers care about HDTV?

This might seem like a silly question. In the UK retailers don't even try to sell standard definition TVs any more - all 9.5m sets sold in 2010 were HD ready. Moreover, 3.5m households, or about 14% of the UK total, were paying for HD TV subscriptions from Sky by the end of that year. On top of this, all cable households were getting HD for free, as were at least some Freeview (digital terrestrial) households. However, as we will see, ability to view is not quite the same as actually choosing to view.

Ofcom consumer research suggests that 32% of households have HD TV channels. As Ofcom says, this may be an overestimate, since not all consumers distinguish between having an HD TV and getting HD channels. However, we know that at least 28% of households receive HD channels, given the Sky HD households mentioned above, plus the 3.8m cable customers (though a few of the latter may not have HD sets). Freeview HD households are on top of this. For the purposes of discussion, let's say a round 30% of households can watch HD.

Several UK channels broadcast in both SD (standard definition) and HD. In HD households, it is a simple manner of selecting the HD version of the channel to get the higher video quality. The interesting thing is, audiences generally don't bother.

ITV1 is the largest UK commercial channel (and second only to BBC1 overall), and broadcasts in both modes. The average person watched 5 hours, 46 minutes of ITV1 (SD+HD) per week in December 2010. Of this, 14 minutes was of ITV1 HD. Of course all this HD viewing came in HD households, so that implies the average ITV1 HD viewing was 46 minutes in those households.

However, this means that in households with the option to watch ITV1 in HD, viewers chose to do so for just 46 minutes out of their 5 hours 46 minutes of ITV1 viewing in both modes, or 13% of their viewing.

This is a striking result. Even in these HD households situation, where consuming exactly the same programme in HD rather than SD is simply a matter of clicking a few buttons on the remote, 87% of the time audiences simply don't bother.

This doesn't prove HD is irrelevant, but it does suggest it has limited general value for audiences. It is, for instance, likely more valuable for films or sports than for quiz shows.

Which brings us to superfast broadband. HD is regularly cited as one of the strong arguments for superfast, and often features heavily in proposed 'app stacking' (multiple activities in a single home using one broadband connection). However, if consumers see HD as irrelevant to most of their viewing, then it's hard to justify spending large sums on superfast to bring it to them.

Moreover, the strength of the internet for TV is in on-demand programming (since broadcast works perfectly well for linear TV). But sports are very heavily consumed live. Most countries already have infrastructure to deliver HD TV versions of sports channels and programmes, be it via terrestrial, satellite or cable. Superfast therefore adds very little to HD live sports.

This leaves movies as the key remaining TV content that might have substantial value as on-demand HD, and therefore contribute to the case for superfast (though cable VOD is also capable of this). This is a limited use, and certainly not used heavily enough to contribute regularly to an 'app stack'.

Thus those building the case for superfast on the back of HD need to explain why they expect consumers both to become much more interested in high definition and on-demand services than they have been to date.



[Note that for the purposes of this simple analysis I have ignored issues of secondary sets and different channel shares in HD/non-HD households, though these are unlikely to change the outcome materially]

Chancellor offers you £17 to upgrade your broadband

In today's UK budget the Chancellor announced ten cities that will "share £100m to introduce ultra-fast broadband". The cities are Belfast, Birmingham, Bradford, Bristol, Cardiff, Edinburgh, Leeds, London, Manchester and Newcastle. You will have spotted that there are some big cities in this list - their combined population is 14.6m.

At 2.5 people per household (the rough UK average) this is equivalent to 5.9m households. So, the £100m works out to £17 per household in the relevant cities. Given average civils (construction) costs of £50-60 per metre, this would get the duct carrying your fibre roughly 30cm down your front path / driveway.

My point is not of course that the government should be throwing more money at superfast - they certainly shouldn't, particularly in cities where existing broadband is likely to be of a reasonable standard. Rather the fact that £100m gets you 30cm of duct per household just shows how expensive fibre-to-the-home is.

Is FTTH material to the EU's broadband target?

The European Commission has set a target that "By 2020 ... 50% or more of European households have subscriptions above 100Mbps". A bold target - how is fibre roll-out and adoption tracking against this?

The FTTH council - arch enthusiasts for superfast - publish annual statistics of fibre-to-the-home roll-out and uptake. The story so far looks like this:

Note: These figures include some countries outside the EU27, most notably Turkey

As we can see, both roll-out and adoption have been quite steady over the last four years, adding roughly 5.8m and 1.0m homes per year respectively over that period (with a slight acceleration in the rate of adoption).

As a percentage of homes passed, uptake has been in the range of 15-20% since 2007. This doesn't prove uptake has stalled - each year, adoption is diluted by the new homes passed in that year, which inevitably start with zero penetration. However, with each passing year this dilution effect gets smaller, since each year the portion of passed homes that are new gets smaller and smaller. (For instance, in 2008 57% of homes were new - in 2011 20% were). This suggests that adoption is making only very slow progress even in locations where fibre has been available for some time.

Purely from a financial perspective, this has to be concerning. The costs of fibre include a very substantial fixed element simply to pass homes. Thus the rate of uptake is critical. You lose a lot of money on the homes you pass that don't take fibre, which you have to make up from the homes that do take fibre. Other than in highly urban environments (where costs are lower), I would be surprised if many FTTH business plans show a profit with adoption rates anywhere near as low as 15-20% uptake.

Setting aside the financial challenges for individual operators, what does this rate of progress tell us about FTTH's likely contribution to the EC's target? Using a straight line projection of FTTH adoption based on the slightly higher growth of the last two years, we get a 2020 figure of 6.9% of European households taking FTTH service:

In other words, if FTTH is going to a significant contributor to the EC target, then either roll-out or adoption  will have to improve dramatically. This would be more believable if either of these metrics were on an upward trajectory. However, additional homes passed has been steady over the last four years in the range of 5.3-6.3m (and 2011 was in the lower half of this range).  For adoption, the graph above uses the slightly higher rate for 2010 and 2011, an average of 1.3m homes per year. However, the rate for 2011 was actually lower than than for 2010 (1.2m vs 1.4m), suggesting that FTTH is certainly not enjoying exponential growth.

While it is conceivable there will be a surge of adoption, perhaps driven by some currently unknown, compelling superfast application, it is worth noting that at the current rate of roll-out, FTTH will only pass 24% of European households by 2020. Even if they all took FTTH broadband (and that at 100 Mbps), FTTH would still only be contributing less than half of the EU target.

Does this mean the EU target is doomed? Not necessarily. FTTH is not the only way to deliver 100 Mbps. Cable broadband already can, and by 2020 FTTC (fibre to the cabinet) may very well be offering such speeds. (BT is launching 80 Mbps FTTC in April 2012).

However, this analysis does suggest that unless there is some radical change, FTTH is something of a sideshow for the EU's superfast target. If policy makers have serious ambition to meet that target, they will need to focus on cable or FTTC solutions.

[The underlying data for the above charts is available from here: 2008, 2009, 2010 and 2011]

'Future proof' - A flawed justification for fibre

Fiber-to-the-home (FTTH) enthusiasts often say that it is the essential choice because it is 'future proof' (see, for example, this). While this sounds good, there are a number of problems with this argument.

We perfectly sensibly invest in things that aren't future proof all the time. Newly weds do not immediately buy houses to accommodate a brood of children, we don't build all airports with three runways on day one, we don't build all roads as six lane highways and so on.

There are two fundamental reasons for this - we don't know if the anticipated demand will materialise, and even if it does, there is value in delaying the investment to accommodate it. The newly weds might never have children, and even if they knew for cast iron certain they were going to, why spend the money on empty bedrooms today? The cash could be put to better use in the meantime (particularly if, say, the newlyweds were facing a global financial crisis at the time of their wedding).

In the world of superfast we can take a look at the choice between FTTH and fibre-to-the-cabinet (FTTC) in these terms. The latter is considerably cheaper, since it doesn't involve digging up the roads all the way to individual houses, but on the other hand its theoretical capacity is lower. It's possible that at some future point demand for speeds even greater than FTTC can provide (currently in the region of 80 Mbps and rising) will mean that we'll feel obliged to overbuild with FTTH.

At that point, will those crying 'future proof' in 2012 be proven right? Well, it all depends on the numbers - the costs of the two technologies, how much of the FTTC build cost would be wasted, how long before any overbuild is required, the probability of overbuild being required and the discount rate (a measure of value of delaying spend).

Below is an illustrative analysis of this. One critical assumption is the relative costs. I've assumed FTTH is three times the cost of FTTC (this may be conservative - Ofcom says 3-4x, and WIK says 5x in urban and suburban areas). Don't worry to much about the absolute dollar figures - $400 per household for FTTC and $1200 for FTTH - it's the ratio that matters for our purposes.

I've somewhat arbitrarily assumed that there's an 80% chance that overbuild will be necessary, and that it will be required in year 4. These sound generous to me, but there's room for reasonable people to disagree, certainly. However, keep in mind that the 80% is not the chance that there will be any demand for FTTH speeds, but rather the chance that there will be enough demand for incremental speeds beyond those of FTTC to justify the cost of the overbuild. Given that FTTC can provide 80 Mbps and rising, 80% feels like a high number to me.

I've assumed that 50% of the investment in FTTC is a write-off if it's overbuilt with FTTH (based on Value Partners' view). Finally, I've assumed a 10% discount rate, fairly typical for commercial investments. So how do the numbers play out? See below:



As we can see, going for FTTC now saves us $198 per household, even allowing for the fact that overbuild with FTTH later may be necessary. The fundamental reason is that there is huge value in delaying the substantial spend on FTTH. Note that even if we were 100% certain that FTTH would be required in a few years, it would still make sense to start with FTTC now, though the saving drops to $49.

All the above analysis assumes (generously) that FTTH itself won't be replaced by something better. Certainly it is today the way to deliver the highest bandwidth to the home, but the future has a way of surprising us. Want to know what was being called 'future proof' in 1995? ISDN, now a product in its twilight years - BT killed of consumer ISDN in 2007, for example.

So, the next time someone tells you we have to spend money on FTTH because it's 'future proof', keep in mind that they may be wrong, and even if they're right, it may still make sense to build FTTC now and see if we need FTTH later.

What does Onlive tell us?

Onlive has recently caught my eye. They provide cloud based gaming, and I think tell us something very interesting about broadband requirements for cloud based services in general.

All the processing for Onlive is done on their servers in the cloud - the pictures are streamed down to your PC or tablet in real time. Your computer is essentially a dumb terminal, displaying the video and sending up mouse clicks (frantic ones, perhaps, if you're desperate to blast the aliens). A technical description is here.

Onlive is a great test case for cloud based services, for a couple of reasons. Firstly, it is a demanding application from the perspective of latency. If there is material lag between the buttons you're pressing and what is happening on screen, action games (such as the first-person-shooters and racing games) will simply not work. This is the reason serious online gamers obsess about ping, even for games where the processing happens locally.

Secondly, the video needs to be high quality and bandwidth intense. When you're using a word processor, not much changes on the screen at any one time. Even if the entire screen was being streamed from a remote server, compression could keep the bandwidth requirement relatively low. It's a very different case for action games, where the whole screen can change rapidly, as the player turns around or an explosion erupts.

In other words, given Onlive's need for low latency and rich video, if your internet connection can handle Onlive, its going to be able to handle the great majority of cloud services.

So what does Onlive require? "5 Mbps for TVs 40 inches and larger" - and no mention of any minimum upstream requirement.  This compares to an H1 2011 average broadband speed of 7.5 Mbps in the UK for example (a figure which excludes the relatively small number of superfast connections in that country). In other words, even with someone else gaming in the house, you'd still have enough bandwidth to be watching a typical 2 Mbps streamed TV programme from the BBC's iPlayer (standard definition - 3 Mbps for HD).

The moral of the story is that today's infrastructure will already deliver us even quite demanding cloud-based services with bandwidth to spare, for the majority of users. The cloud is exciting, and has huge potential (though some risks), but whatever it is it isn't a ready justification for throwing money at fibre broadband.

The story so far

My interest in superfast started with a project for a client who asked the question whether government funds were better put to work in extending broadband coverage or speed. When I accepted the project, I thought it would be relatively easy - primarily a modelling exercise based on existing research into the benefits of each. However, when we started looking at the research into superfast, it turned out to be relatively thin. Moreover, given the costs involved, we found we had to make heroic assumptions about the benefits of superfast - and in particular FTTH - to believe that it was a better return on societal investment than FTTC or extended coverage of basic broadband. This got me wondering if the emperor was in fact fully dressed. (The results of this project are here).

As it happens, my brother Charles, a development economist, has done work in the same area. After some discussion, we decided we should write a joint paper taking a hard look at the case for superfast, and in particular its purported societal benefits. He focussed on the economics (since he understands them) and I focussed on the applications and technical side. Working on this paper made me think that even if the emperor did in fact have some clothes, he certainly didn't wear them very often. Time and again, advocates for subsidising superfast to the home made one of three errors, basing their case on applications that:

• Could run on superfast, but equally could run on more basic broadband
• Had little societal benefit, and therefore did not seem to justify government intervention
• Did not require connectivity to home, but rather to businesses

Moreover, evidence seemed to be being used very casually and frequently misleadingly (see The hall of shame for some of the more shocking examples).

Charles' and my paper was published in November 2010 (with an academic version in the journal info in 2011). It attracted a fair bit of attention, not least in Australia where feelings run high on broadband policy. Some people thought the paper was full of good sense, others thought we were idiots. (Our reply to the latter is here, on p6).

Since then I've continued write and present on the topic. Fibre-caution remains a minority view (at least in public - it's interesting what people in pro-fibre organisations sometimes say behind closed doors), but I'm sticking to it for the time being, until the evidence suggests otherwise.

What's this blog for then?

This blog is primarily a place to present a somewhat sceptical perspective on the merits of superfast broadband.

There's no shortage of enthusiasts for superfast. Governments around the world are throwing money at it, equipment manufacturers are singing its praises, incumbents see the opportunity for weakened regulation in exchange for rolling it out and conversely new entrants hope that a replacement of legacy infrastructure will create opportunities for them.

However, there are few people making the contrary case: that there might be diminishing returns from ever greater bandwidth to the home, that the applications enabled by superfast broadband are likely to be primarily entertainment-related, and that therefore there is little reason for the government to be intervening to suppport superfast broadband. This blog will seek to set out that case.

Those seeking government support for superfast propose spending large sums of tax payers' money and making major interventions in the market. In markets such as Australia this includes a renationalisation of a large part of the telecoms market. It seems to me that this puts a significant burden of proof on those who would support such steps. However, I recognise that my own position is a minority one, and that suggests a higher burden of proof for me too. I will therefore strive to be evidence-driven wherever possible.

Superfast will be the focus of this blog, but I have other interests in telecoms more generally and also in media. I may post on these topics as the mood takes me.