Superfast or superfit? The case for UK broadband policy reform

© Martin Geddes Consulting Ltd July 2017
Superfast or superfit?
The case for UK
broadband policy reform
Martin Geddes

2
About the author
Martin Geddes is a network performance scientist,
and a scholar of emerging telecommunications
technologies and business models. He holds an MA
in Mathematics and Computation from the
University of Oxford, and is based in London.
As founder of Martin Geddes Consulting Ltd he
advises clients globally on broadband technology,
policy, economics and service innovation.
As a co-founder of Just Right Networks Ltd, he is
presently working on the launch of the world’s first
commercial quality-assured broadband service.
To learn more visit martingedd.es.

3
Introduction and summary
This is a critical moment for UK digital infrastructure policy. The context is
one of rapid political, market and technological change. As a nation, we
face important decisions over topics like post-Brexit regulation, universal
service delivery, Openreach independence, TETRA replacement and 5G
readiness. The imperative is to reflect on whether our historic approaches
will meet our future needs. Where we anticipate a shortfall, we must act
to protect our long-term national interest.
This paper aims to educate policymakers about one specific shortfall: the
growing ‘capability gap’ between broadband demand and supply. Notably
this is not principally an issue of delivering more capacity. Furthermore,
this gap is not the result of regulatory or supplier failure.
The ‘capability gap’ is a symptom of a young industry built on immature
technology that faces novel demands. It is systemic in nature, and crosses
all geographical and political boundaries. Consequently, it must be
addressed by governments and regulators everywhere.
The good news is that we can realistically achieve a ‘better, cheaper,
faster’ broadband future for the UK. The capability gap can be
substantially narrowed within the lifetime of the present government.
Rather than being a policy reversal, we can build upon our past
achievements: a well-regulated wholesale market, with substantial retail
competition, offering a diversity and abundance of supply.
Such future success requires a policy change: to transcend the current
supply-led ‘superfast’ approach, which risks becoming an expensive
‘superfat’ and unhealthy ‘superidle’. It is not worth coming top of the
national league tables of over-investment in empty networks that still
don’t deliver the capabilities needed. The booby prize is to win
internationally on the wrong terms.
Our collective national opportunity is to redefine success around a
demand-led ‘superfit’ model. Connected vehicles, health monitoring,
flexible teleworking and smart homes are arriving and spreading fast. We
can take the lead in creating a world-class platform that is fit-for-purpose
for tomorrow's ‘Internet of Everything’, as well as today's common
applications.

4
The time to act is now. Many rural MPs are already finding their mailbags
full of angry constituents who lack adequate coverage or capacity. As a
result, they cannot access key entertainment or work applications. On our
present trajectory, all MPs will find their mailbags swelling with angry
‘superfast’ broadband users. Constituents will be unable to access vital
energy, health or transport services due to shortfalls in capability. This will
cause an outcry that cannot be ignored.
This unwanted situation is avoidable by two readily attainable changes in
our policy approach.
Firstly, our policy metrics need to reflect the readiness of broadband
infrastructure to support both present and future demand.
Secondly, the money needs to move to incentivise the right market
behaviours to create a correspondingly fit-for-purpose supply.
When these reforms are enacted together, this will help to position the UK
with a world-class infrastructure ready to attract capital and talent on a
global scale.

5
Understanding the ‘capability gap’
People matter more than packets
The first step to understand the 'capability gap' is to grasp what it means for
broadband to be fit-for-purpose in the eyes of end users. In talking to people around
the country, this author hears stories of the positive difference good broadband
makes, and the problems caused by its absence or failure. These go beyond the simple
matter of whether broadband (superfast or not) is available where needed.
Consider for a moment a couple who are both deaf, one profoundly so. They can get
broadband, but the service where they live is insufficient for their needs. Applications
like Skype fail to deliver acceptable video sign language at their home. They require a
rock-steady video frame rate with perfectly synchronised lip and hand movement, as
there are no audio cues to cover for glitches.
For them, the current offer is not fit for purpose, even if they can satisfy their love of
cricket via online streaming video. It has to have sufficient performance to meet a wider
range of needs.
Another example is an elderly lady who cares for her invalid husband. She has taken to
using online grocery shopping, and indeed has become dependent upon it. The
performance and availability of this application are not guaranteed by today’s
infrastructure. Should her shopping experience seriously decline due to growing
streaming video use at peak times—not a hypothetical scenario—then her life will be
significantly impacted.
The occasional successful availability of the application out of peak hours is not
enough to make it fit-for-purpose. It has to be dependable, too.
Then again, there is the story of a family who are using satellite broadband in their
rural home, as it is the only available option. All too easily the children use up the
whole allowance in just a few days if left to access streaming video. The monthly
overage charges, if left unchecked, can easily reach a thousand pounds or more.
A fit-for-purpose infrastructure has to be affordable. That can only be achieved by
supporting the maximum number of uses over the minimum duplication of
infrastructure, so it can have the greatest reach at the lowest cost.
This also means we can’t afford to keep building separate access networks for energy,
healthcare, industrial or emergency use. For example, when a farm tractor breaks
down or tumbles into a ditch, the sensors and communications systems need to be
able to access a common ‘information grid’. Fitness-for-purpose is about supporting as
wide a range of uses as possible on a truly general-purpose national infrastructure.

6
These parameters of availability, performance, dependability and affordability define
the nature of application demand. Success in a ‘superfit’ policy regime requires us to
construct a matching information exchange supply.
The third wave of broadband supply
As the digital demands of the population have evolved, the UK’s policy and regulatory
bodies have responded to help drive supply forward. It would be churlish to dismiss the
successes of past and present policy approaches.
In the 1990s, before broadband, Oftel delivered unmetered access to the dial-up
Internet in response to consumer demand. Then in the early 2000s, the Access to
Broadband Campaign pushed hard for rapid roll-out of the broadband availability. The
political establishment responded. Ofcom was created, and the result was a broad
success to the credit of the industry. This ‘first wave’ of broadband was all about
coverage, with users valuing the always-on nature of early broadband.
Over time, people came to use broadband for more uses, with more concurrent users,
and with more demanding applications. The second broadband wave, our present
‘superfast’ era, has been about delivering ever more capacity. We have just passed our
milestone for making superfast connections available to 90% of homes. This legitimate
achievement took just six years from commitment to delivery, only one more than
originally planned.
We are now at a transition point again. More capacity alone does not solve the
emerging and future problems that users face, as we shall soon see. The third wave of
broadband is about fitness-for-purpose, and the capabilities of the network to enable
that.
We earlier examined what fitness-for-purpose might mean to people. Those needs
eventually have to be translated into delivery requirements for packets. What does ‘fit
for purpose’ mean for the capabilities the broadband service must offer? To answer
this, let us digress for a moment away from broadband to another utility service,
namely electricity.

7
Broadband is not yet a true utility
Where is the standard ‘socket’ for broadband?
One of the great unsung fit-for-purpose innovations in British society, one we are all
familiar with, is the BS1363 13 ampere power plug and socket. This is superior to other
plugs by virtue of its solid construction and safe design.
Firstly, the three square prongs make for excellent electrical contact. It is practically
impossible to wobble the plug to cause sparks or intermittent connectivity. The
‘success mode’ of clean, continuous power is fully covered off. But that’s not all.
When the earth prong goes into the socket, it opens up shutters than reveal the live
power. Small children can’t put sticky fingers in the socket, to the occasional regret of
a frustrated parent of a screaming toddler. Yanking on the cord also does not easily
apply undue force to the electrical components causing a dangerous fracture. Another
great thing about a British plug is the fuse. If there is too much demand, then it cuts
out, rather than going on fire. So as a design, the ‘failure modes’ are also well covered
off.
When you stand in the store to buy an electrical appliance, it is easy to tell what the
rated demand is in terms of volts and amps. The capability of the supply is also clear,
both for the whole dwelling, as well as in-building distribution like multi-socket power
strips. You know your cooker needs a special supply, and that you can’t power your
tumble dryer off an AA battery soldered onto a socket.
In summary, a fit-for-purpose interface between supply and demand does three
things: it enables 'success' for specific uses; it sufficiently limits ‘failure’ for those uses;
and it clearly communicates what uses it is suitable for to the buyer.
Broadband: ‘bare wires’ or ‘hard-wired’?
What is missing in broadband is the conceptual equivalent of the standardised plug
and socket. The interface between demand and supply is defined at an electrical level,
but the overall service of information exchange is (mostly) undefined. As a result, we
are left with two less than satisfactory approaches to service delivery.
One technical approach is how we use ‘over the top’ applications like iPlayer today. It is
as if we leave an unshielded live information ‘virtual cable’ exposed directly to end
users. ‘Success modes’ are enabled, since many applications work some of the time,
but the constraint on their ‘failure modes’ is weak.
In this model, users are not sufficiently ‘insulated’ from one another. Performance
‘brown outs’ from overload are common, as our example with video sign language
demonstrates. As your children come home from school and go online, the
performance of your important work application tangibly plummets.

8
Alternatively, we have vertically integrated network services, more like how traditional
landline phone calls or cable TV work. The information ‘virtual cable’ from the
appliance is ‘hard-wired’ into the wall, and it can’t be switched over.
Whilst performance is predictable, and the service is usually fit-for-purpose, it is a
highly inflexible approach. The price of constraining the ‘failure mode’ is a severe limit
on the number of ‘success modes’. Vertical integration reduces consumer choice, with
a high cost for any services delivered.
The need to ‘insulate’ the application from other uses may even result in a complete
parallel infrastructure, as we have created for smart meters at a cost of billions of
pounds. The resources spent on special-purpose smart meter connectivity could have
delivered an enormous improvement in the general-purpose infrastructure useful for
transport, healthcare and emergency services.
We certainly can’t afford to build duplicate infrastructures for every industry and
application whose needs diverge even slightly from basic Internet access.
We must transcend ‘purpose-for-fitness’
As this thought experiment about power sockets shows us, the kind of fitness-for-
purpose we take for granted with electricity is missing for broadband.
Indeed, what we have with ISP services is the exact reverse: a ‘purpose-for-fitness’
model. You buy the service and then find out what it might (however transiently) be
useful for by trying out different devices and applications! The consumer broadband
experience is more Arthur Daley than John Lewis in terms of its service promise: you
got whatever your got, without any recourse if it is unfit-for-purpose.
What we would like is an infrastructure that supports a large variety of applications
that are available everywhere. Performance and dependability should come at an
affordable price. For a “never knowingly undersupplied” infrastructure to work, we
need to define what the broadband “socket” supplies from the “information grid”. To
answer this question, we must further unpack what service the information “socket”
offers and the device “plug” receives.
How broadband supply is built
With some poetic license for a general audience, the supply of any
telecommunications service can be characterised by four factors: coverage, continuity,
capacity and consistency. Each has multiple facets, and fitness-for-purpose is the result
of them existing collectively.
Coverage about the existence of connectivity. Most obviously, it requires the physical
infrastructure to be within reach of your device's network cable or radio. Less
obviously, users must have the permission to be able to access it. For instance,
multiple ISP services allow you to ‘roam’ onto the WiFi of other customers or affiliated

9
networks. Coverage includes the idea of being able to access a diversity of services,
possibly concurrently, and being able to secure them from intruders or attack.
Continuity is a way of defining resilience. The most obvious issue is one of service
reliability and time to identify, isolate and fix faults. Yet this factor it is richer than that.
Technologies like DSL (used for copper-based broadband access) have ‘line speed
retrains’ and re-authentication that cause ‘micro outages’. When we bond together
different access technologies, or share radio networks, we may get brief service
interruptions due to transient loading effects. A mobile phone may experience gaps in
continuity due to hand-offs between cells, and transitions between 2G, 3G and 4G.
Capacity is the ‘quantity’ aspect of supply. If you want to download a gigabyte of movie
to a user in a minute, there is an absolute minimum capacity level required in terms of
megabits per second.
Finally, consistency is the most misunderstood factor, being the ‘quality’ part of supply.
The timeliness with which information is exchanged at the 'socket' needs to be
constrained. It also cannot vary too much or too fast. Network scientists call this
essential stability property ‘stationarity’. Without it, even the cleverest adaptive
application cannot work.

10
More speed is not the answer
Not all bandwidth is the same
Different kinds of access technology naturally vary in their properties across these
factors, even when they have the same ‘bandwidth’.
A key reason for deploying fibre-optic cable to the home is that is has strong continuity
of service, and often excellent consistency due to idle resources. That it has enormous
(over)capacity too is a bonus! The good operational maintenance costs compared to
copper help to lower prices in the long run.
For media like cable and cellular the capacity of the access medium is dynamically
shared between users, so by default they have poorer consistency than unshared
access like DSL and fibre to the home. Satellite has enormous coverage and good
capacity, but also lacks the consistency we would like for real-time use.
The UK has invested a lot of resources in higher capacity through fibre to the cabinet
and VDSL. This technology has different coverage, continuity and consistency
properties than simple DSL. The result is not always an improvement on its
predecessor technology.
Success in delivering ‘superfit’ broadband is simple, at least in theory. The demand for
timely information exchange from our “information socket” needs to be met through
these four supply factors. Each factor can affect the availability, performance,
dependability and affordability of that information exchange.
Dynamically matching supply to demand
The whole computing and communications industry is involved in a long, slow process
of moving from fixed and inflexible supply to a responsive on-demand model. For
computing we have cloud-based service delivery, and for telecoms we have
technologies like software-defined wide area networks. An on-demand model targets
supply to match demand on ever-tighter timescales.
At the longest timescales, we make architecture choices over coverage, like where to
locate street cabinets, and how long the local copper loop will be. We select and install
equipment that defines the supply service capability.
Different technologies have different “switching” times for service coverage, affecting
whether changing demands can be met with a fresh supply. With dial-up you could
connect to a new ISP service within seconds. In contrast, today’s commercial model for
fixed access broadband coverage is a “single transferrable monopoly”. You subscribe
to a one service provider at a time per access line, and the switching time is typically in
the range of months to years.

11
At shorter timescales, we can reconfigure our broadband lines to have different
coverage range, capacity, and continuity. At the shortest timescales still, we can
schedule packets differently to offer multiple levels of quality with different levels of
consistency.
Success in future is not about predicting demand by gazing into a crystal ball, and
creating a static supply with a Soviet-style five-year plan. What is required is a system
that can dynamically flex to support changing demands. This requires us to have
“option value” baked-in, so we can support a diversity of needs for availability,
performance, dependability and affordability.
This still requires long-term infrastructure planning, a job sometimes better suited to
governments and public institutions than markets. It is all too easy to bake-in
constraints that are expensive or impossible to remedy later. The adoption of
standards -- as happened for domestic power supplies -- may not happen in a
fragmented industry without intervention.
‘Speed’ is not the same as fitness-for-purpose
We have examined the nature of demand and supply, and the need to match them
dynamically. How well are we doing at this with today's policy approach?
There is evidently a richness to both application demand and the properties that
broadband supply offers to meet it. Yet the retail broadband market has adopted a
naïve approach to characterising supply. The primary focus is on ‘speed’, by which we
mean a user runs a ‘speed test’ application. This spits out a number for peak upstream
and downstream data transfer rates.
These numbers are the result of what happens when many factors interact with each
other: transport protocols, on and off-net link capacity, traffic management, transient
loads from other users, as well as the speed test application design and hosting. Many
of these factors are outside of a broadband service provider’s control. These ‘speed
tests’ (when run at scale) also act as denial-of-service attacks on the customer
experience since they stress the network by design. This makes ‘speed’ an odd (if
common) choice of policy goal.
There is one important technical fact here that every policymaker needs to
understand: the essence of network demand is not for this ‘speed’ as measured by
‘speed tests’. What is required is sufficient coverage, continuity, capacity and
consistency. The measured “speed” is definitely not the same thing, not even for
capacity! Indeed, what we colloquially call ‘speed’ for things like Web pages is really
‘user application responsiveness’. This is a result of both capacity and consistency
interacting.
The figure for peak transfer rate under maximum load is not a good reflection of
performance for applications other than a ‘speed test’. A network optimised for speed
tests may be very poor at real-time applications, for example. A single metric simply

12
cannot capture the richness of coverage, continuity, capacity and consistency
requirements.
Why ever more ‘speed’ is not the answer
The initial leap from dial-up to broadband was not only a jump in ‘speed’, but was also
a great improvement in continuity, as the service was always on. By definition more
speed has required more capacity, which has value for many applications. It is true that
historically more speed has been a good proxy for better consistency, as the network
could be ‘over-provisioned’ more readily with idle capacity.
However, there are decreasing benefits to more speed. To the surprise of many, the
unthinking pursuit of speed can result in lower fitness-for-purpose. As it is a relatively
technical discussion, a list of reasons is captures in the appendix.
An important collective effect of these technical factors is a loss of that essential
consistent stationarity, and a growth of inconsistency as ‘non-stationarity’. This results
in the widespread failure of interactive and real-time services. Whilst these terms may
not yet be common in the lexicon of policymakers, they ought to be. You might think
of it as a statistical ‘global warming’ inside of networks due to our profligate resource
use policies!
The current UK policy incentive is to drive people to select 'faster' broadband networks
and packages. However, a ‘faster’ ISP service doesn’t necessarily imply greater fitness
for purpose. It may have worse coverage, continuity and/or consistency. In a 2014
report1
commissioned by Ofcom, it was found that speeds above 10mbit/sec had little
impact on improved user experience.
Furthermore, to deliver the consistency required we have a de facto a policy of
‘superfat’ and ‘superidle’, rather than ‘superfast’. This works against the need for
affordability, using large amounts of capital to run networks idle to avoid resource
contention effects.
1
http://stakeholders.ofcom.org.uk/binaries/research/technology-research/2014/performance-eval.pdf

13
More speed alone risks a policy ‘car crash’
Focus on speed leads to the capability gap
The capability gap is a shortfall in the fitness-for-purpose of the UK's digital
infrastructure for the applications we can reasonably anticipate. It crosses all four
supply factors, especially coverage, continuity and consistency.
The technical issues could occupy a book, not just a white paper. To briefly summarise,
the key problems are:
• Coverage: Users are only able to select a single service provider, resulting lock-
in and lack of diversity of network services. Switching times are very long, and
the process is prone to errors. There is a lack of security isolation of services
needed for 'industrial internet' uses.
• Continuity: There are persistent and widespread ‘micro outages’ on fixed
access from architecture and operational management choices. The systems for
service monitoring do not automatically detect performance faults, and their
isolation in the supply chain is more craft than science.
• Capacity: Users in rural and select urban locations face a significant shortfall in
capacity, especially in the uplink. The amount of (user and taxpayer funded)
network capacity that is effectively unusable (as it must be idle to maintain
consistency) is large and unquantified.
• Consistency: Weak performance isolation between users resulting in growing
non-stationarity and application failure. Users are unable buy assured quality for
their critical applications. The ability of the fixed broadband network to support
mobile backhaul is significantly limited.
The gap will harm the UK if not addressed
There are very specific harms that can be foreseen should we fail to address these
capability gaps.
The immediate problem is a growing unfitness-for-purpose. Users have a diversity of
needs, yet we are attempting to deliver everything with the cost structure of real-time
interactive video. This ‘coals by motorcycle courier’ model has inherently
unsustainable economics. As a result, ISPs drop service quality to avoid spending on
(yet more idle) capacity, and applications that previously worked for users stop
working.
Any reversal of performance improvement creates a consumer loss of trust and
confidence in their suppliers and a crisis of legitimacy for Ofcom. Humans have a
cognitive bias called the ‘endowment effect’ that makes them over-sensitive to losses.
The 'turning point' in service quality that triggers this varies by access technology and
provider.

14
The service quality ‘performance cliffs’ are not being measured or modelled by the
industry or regulator. The scaling limits of our architectures are not well understood. In
short, the safety case for our national infrastructure is absent, and the risks to
continuity and consistency are not quantified. Should other industries (media, travel,
retail, etc.) experience failure of their digital distribution system, it will create intense
political pressure for action.
The standard approach to scaling (by using idle capacity to solve consistency issues)
results in huge wastage. This is typically 50-90% of network resources in the
experience of my industry colleagues. This drives up costs, reduces investment, and
impairs our ability to service the needs of the weakest in society. By better scheduling
and sharing of resources it is possible to have a massive increase in efficiency and
effectiveness.
If this is not addressed, the unsustainable economics of broadband will create a
pressure for vertical integration, monopoly, and taxpayer subsidy or bailout.
The lack of fungibility of services means that a fully functioning competitive market is
not possible. End users cannot judge whether a 300mbit cable service or a 10mbit DSL
service will be more suitable for their online gaming application. This problem extends
from the retail level all the way down the supply chain. Buyers of wholesale services
cannot be certain that their inputs will result in a working application, now or in future.
The consequence of this continuing is twofold. Firstly, there will be a failure of the UK
policy, as the only way of ensuring services work with clear attribution of responsibility
will be vertical integration. Secondly, other countries who manage their supply chain
better will become the 'digital hubs'. This mirrors, say, the transfer of business from
London to Rotterdam with the arrival of container shipping.
Looking forward, a key transition is to the ‘industrial Internet’, with 5G being a vital
enabling technology. The readiness of UK infrastructure to support 5G backhaul is
somewhere between ‘unknown’ and ‘unlikely’. This will affect the UK's future position
in the ‘Internet of Everything’, with consequences for jobs and growth. The acquisition
of ARM by Softbank is an indicator the importance of this sector.
Is the problem real? Show me!
The sceptical reader will no doubt want some evidence that these issues are real. Here
is a selection of data points. It is not meant to be a survey of the fitness-for-purpose of
UK digital infrastructure.
The Institute of Directors, not an organisation given to hyperbole, says the UK has a
“poverty of ambition”2
when it comes to broadband. Their demand for high-speed
services should be seen as a proxy for fitness-for-purpose.
2
http://mkhan6131.com/uk-poverty-ambition-comes-broadband-says-iod/

15
The Engineering Employers Federation, a manufacturers association, says3
that
download speed not biggest broadband concern. Their issues are reliability and
resilience.
A survey of tourism businesses in Suffolk stated4
that “37% of businesses feel that the
quality of broadband in the area is holding them back”.
According5
to Fix Britain's Internet, a movement set up by Sky, TalkTalk, Vodafone and
the Federation of Communication Services, “two-thirds [of people] say they feel 'let
down" by their internet connection at least once a month. 56 per cent of users said
they were unable to perform even the simplest of online tasks”.
Bell Labs forecasts6
that “By 2020, 19 per cent of the worldwide consumption demand
forecast for connectivity will not be able to be met by today’s network operators”.
3
http://bit.ly/1X6PJ4j
4
https://twitter.com/Suzanne_UCS/status/723190866026860545?s=03
5
http://www.craveonline.co.uk/design/1017407-bad-internet-connections-affecting-mental-health
6
http://www.telecomtv.com/articles/telcos-and-service-providers/bell-labs-warns-that-telcos-will-be-
unable-to-profitably-meet-traffic-demands-by-2020-13484/

16
The time for ‘superfit’ has arrived
A framework for superfit regulation
To address these systemic capability issues, we need a systemic response. This
demands a framework for ‘superfit’ regulation that has three key properties.
Firstly, it must be technically sound. Having the science and engineering in good order
cannot be taken for granted. For instance, the FCC and BEREC have both embarked on
‘net neutrality’ rules that contradict what the science tells us. In contrast, Ofcom
laudably took a lead in getting high quality science advice7
before rule-making.
Secondly, it must be market-based. Progress is typically made via market entry and
exit, no by incumbents acquiring a revolutionary zeal for change. The job of the
regulator is to help the market to function, and the hard work of trial and error that
involves. This requires a reinvigoration of retail competition through better
transparency and lowered switching costs. It also requires the UK to preserve and grow
its successful wholesale access strategy, since this is what enables new technology and
commercial experiments to happen without incumbent permission.
Finally, it must be forward-looking, in two ways. We need to anticipate and enable the
technical needs of the future, and prepare for the industrialisation of the Internet. We
also need to manage failure to protect the weakest in society. This means we need to
have a long-term safety case that ensures that we limit the harms from absent,
inadequate or unaffordable infrastructure.
Within this framework, there are two broad areas that have to be addressed: new
metrics for demand, and moving the money for supply.
New metrics for demand
A demand-led superfit model is centred on user outcomes, not network inputs. This
requires a change is what is measured and managed.
The prerequisite is for Ofcom to establish a common scientific base and framework
from which demand-led regulation can be performed. Unlike power distribution,
where we know it is volts and amps that matter, we haven't even yet agreed industry
standard metrics and measurement tools for basic things like network quality. Getting
the science right should be an easy political ‘win’, especially given the UK is a leader in
the field of network performance measurement.
7
http://stakeholders.ofcom.org.uk/market-data-research/other/technology-research/2015-
reports/traffic-management

17
The first step is a consultation to scope this activity: what are the requirements that
should be met for a good framework of user-centric metrics? Key concerns must
include how well the metrics reflect the actual user experience; and the ability to
(de)compose supply chains metrics (including a retail/wholesale split); and the
isolation and attribution of faults (including in-building vs network supplier).
Ofcom can then tender for the scientific and technical advice needed to fully develop
this framework. It may be necessary to work in collaboration with standards and
engineering bodies.
In the meantime, the use of ‘speed tests’ should be reconsidered and limited to where
truly appropriate and absolutely necessary. These create perverse incentives and high
costs. Ofcom needs to consider how it will move to a more experience-led reporting
system, both for the overall state of the market, as well as individual suppliers and
users.
A functioning market requires transparency, so retail users can compare providers.
Ofcom should establish typical baskets of applications for different segments,
reflecting both present and future demand. The demand these place on the network
can then be quantified and become a benchmark against which different supply
services can be compared. Initially the resulting supplier metrics may only be
consumable by industry technical experts. Over time Ofcom should develop a
consumer-friendly means of presenting the contracted service experience.
In this process, the UK should take advantage of a benefit of Brexit. The European
regulations on broadband transparency focus on irrelevant internal network
mechanisms. An industrial policy for packets is unnecessary. These requirements
should simply be ignored, focusing instead on the overall end-to-end service quality
and fitness-for-purpose.
Ofcom should examine the safety case of the UK's digital infrastructure. There should
be longitudinal measures of technical hazards like the growth of non-stationarity.
Furthermore, there should be a public risks register that Ofcom should (in
collaboration with industry) maintain and manage so that we do not get unpleasant
technical surprises when we discover unforeseen scaling limits, or other design
constraints.
Finally, a demand-led approach should define the network termination point at the
information exchange level. What does the ‘socket’ logically offer? This is a key point
of regulation, and should also abstract away the physical implementation. It should
not matter if the network is a single technology (e.g. 3G, cable) or multiple
technologies bonded together, if the user outcome is the same. It may be necessary to
regulate the physical outlet as well as include measurement functionality in customer
access gateways.

18
Change the market incentives
Move the money to create supply
An industry associate often quips “The answer is ‘money’. Now, what is your
question?”. If the UK wants to have different and better outcomes for its digital
infrastructure, the only way of making this happen is to create different financial
incentives. These can be seen as aligning to the four categories offered earlier:
coverage, continuity, capacity and consistency.
With respect to coverage, BT has shown a remarkable ability to innovate and act when
faced with an overbuilder offering a competing network. This emphasises the value of
competition in creating incentives, but at a price of allowing BT to unfairly leverage its
dominant market position. The government should simply commit to enforcing
existing Competition Act provisions to prevent this in future. If this is insufficient, a
more proactive approach to licensing BT's build-out plans may be required.
The fees that Openreach can charge wholesalers and ISPs should be restructured to
reflect user value, not input cost. These should create a significant upside for enabling
new and valued capabilities, such as multiple concurrent retail services with fast
switching. There should also be a downside to delivering technology that cannot be
unbundled, falls short of its advertised capacity, or lacks the continuity or consistency
required.
Openreach should be rewarded for offering services with more tightly defined
performance characteristics (which are then successfully delivered). A loose or,
ambiguous service definition should result in a lower payment, as should failure to
deliver on the promise. Rather than having a war of attrition with the regulator over
open access to holes and poles, Openreach should be focused on raising the value of its
offering to society.
Ultimately questions of BT's structure as a supplier should be answered by BT
responding to demand-led incentives, rather than being engineered by Ofcom (who
then take all the blame). If BT can sweat its copper assets and still deliver the end user
service outcomes, there's not a problem to be addressed. The evidence to date is that
there are significant limitations in fitness-for-purpose that result from BT's technology
choices.
Ofcom should study the technical feasibility and economic benefits of opening up the
Openreach Equivalence Management Platform (EMP) to multiple input suppliers. This
platform is the nucleus for a commodity trading system in network access. The ability
for users (individually, or as communities) to secede from Openreach and ‘declare
independence’ should be investigated. If Openreach refuses to raise the value of their
offer, others can enjoy the benefits instead.

19
The need for continuity of service is important for real-time applications, as well as
mobile backhaul. Again, the financial rewards that Openreach can claim for its service
should be tied to what value it delivers to end users.
It is also important to keep some level of powered copper infrastructure in case of
disasters and emergencies. The recent flooding in Lancaster offers a study in what can
go wrong. There is also a need for diversity in supply for the same reason: the market is
not set up to address 'tail risk' events. Ofcom needs to audit the safety case for the UK's
digital infrastructure. Just as banks face ‘stress tests’, so should our broadband supply.
Finally, there is the consistency of the offer. Advocates of ‘net neutrality’ are de facto
arguing for a system of rationing of quality. This is both technically inefficient and
commercially unfair. The UK should commit to a market-led approach where network
resources are given on the basis of willingness and ability to pay.
UK policy should be to lead the way in developing more technologically advanced
broadband networks, with richer commercial models. These should include having the
‘optionality’ needed from more levels of performance and resilience. The UK should
encourage development of quality-assured services needed for the ‘Internet of
Everything’. This is the opposite of the EU approach, which encumbers their
development with heavy-handed regulation.
Where next for UK broadband?
The proposals in this paper are intended as the beginning of a debate, not the
conclusion. Moving from a supply-led industry to a demand-led one is going to be a
long process.
The broadband industry faces a transformation from a ‘fat’ to ‘lean’. This revolution in
quality management — from ‘push’ to ‘pull’ — has already happened in manufacturing
and other services industries. By adopting a superfit model, the UK can be at the
vanguard of this inevitable process of change. Is already underway with cloud services
for computing, and networks need to catch up. The question is, where to begin?
Ofcom has in the past solicited public comment on a proposed universal service
obligation (USO) for broadband. This is an opportunity for the UK to up its digital
infrastructure game, and do it in a way that reduces cost and delivers more public
benefit.
The present consultation asks questions about the service definition centred on speed.
This is the wrong measure, and will have perverse effects. For instance, by setting
aiming to deliver 10mbit/sec, the result will be to maximise cost (i.e. subsidised capex
spending on often idle capacity) and minimise coverage (reach of DSL) and continuity
(more line retrains).
The review of the universal service obligation is an opportunity to redefine the market
in terms of outcomes, not inputs. This will encourage innovative technological
solutions to delivering the user benefit to more people at lower cost.

20
All it takes is the political will to indicate that what matters to government is the
experience its citizens can enjoy, not the prestige of league tables on speed. The
'superfit' policy works to prioritise people, not packets.

21
Technical appendix
Why is ‘speed’ such a problem?
More speed can hurt performance. This happens in many ways:
• By driving our copper infrastructure harder for speed, we may find we trade
more capacity for lower continuity. Access lines being driven to their limit are
less stable.
• Statistical effects and protocol interactions in faster networks result in the
performance "insulation" between users breaking down. This impairs
consistency.
• Architectural changes make it easier for users to "interfere" with each other, also
impairing consistency. This can be due to adding additional buffers in the
network (e.g. in street cabinets), or changing design rations (like the relative
capacity of the edge and middle of the network).
• For mobile, there are trade-offs of coverage with capacity and consistency.
Driving the market to compete on peak burst speed impairs coverage and
utility for interactive applications.
• With mobile, there is also a trade-off of capacity and continuity. A denser
network to handle higher speeds has more hand-offs, which result in lower
service continuity.
• "Faster" bearers (4G, cable, VDSL) may have significantly worse consistency than
"slower" ones (DSL) and cannot support the same applications at the same
quality.
• The capital cost of more speed drives more network sharing, and the dynamic
load effects of interacting populations of users can negatively affect continuity
and consistency.

Martin Geddes Consulting Ltd
Incorporated in Scotland, number SC275827
VAT Number: 859 5634 72
17-19 East London Street, Edinburgh, EH7 4BN
To learn more
I recommend reading the presentation Essential science for broadband
regulation.
To understand the criteria for a good regulatory metric for broadband, see
How should regulators measure broadband quality?.
Read more about the science of quality at qualityattenuation.science.
There is a reading list of articles, papers and presentations on my website.
This includes a detailed reading list on this topic of broadband quality
regulation.
Sign up to my free Future of Communications newsletter and follow me
on Twitter.
Get in touch
Thank you for reading this paper.
Feel free to reach out to contact me:
mail@martingeddes.com
I offer educational workshops in this
subject area, and frequently work as
a ‘thinking partner’ to senior
executives.
To learn more about the services on
offer visit www.martingeddes.com.

Superfast or superfit? The case for UK broadband policy reform

More Related Content

What's hot (20)

Similar to Superfast or superfit? The case for UK broadband policy reform (20)

More from Martin Geddes (20)

Recently uploaded (20)

Superfast or superfit? The case for UK broadband policy reform