The 'state of the possible' in telecoms is a long way ahead the 'state of the art'. The new science of network performance enables a large leap in customer experience and cost. However, the perception among operators is that only relatively small, incremental improvements are possible.
This presentation explores the reasons for this 'perception gap' between what is seen to be possible, and what actually is. It draws on our work at senior levels for tier 1 operators, as well as examples from outside the telecoms industry.
Overcoming this gap opens the possibility to disruptive innovation. Who will seize the opportunity? Incumbents, challengers or new entrants?
The perception gap: the barrier to disruptive innovation in telecoms
1. THE PERCEPTION GAP:
THE BARRIER TO DISRUPTIVE
INNOVATION IN TELECOMS
28th July 2015
MARTIN GEDDES
FOUNDER & PRINCIPAL
MARTIN GEDDES CONSULTING LTD
2. 2Summary
Industries and organisations vary in their ability to absorb disruptive change. We use examples from 3M
and the US Navy to illustrate the difference. The introduction of network performance science is also
disruptive to the telecoms industry. It is a fundamental change in the way we think about networks. It
radically advances the products we can construct and offer. The cost and value improvements are not small
incremental ones, but rather are by orders of magnitude.
As a result, we see a ‘perception gap’ between what seems to be possible (from the typical network
operator perspective), and what we believe to be actually possible. The idea that the ‘state of the possible’ is
a long way beyond the ‘state of the art’ is hard for those in senior positions to accept. Resolving this
difference opens up an opportunity for disruption. We describe in some detail the nature of the
‘perception gap’, and highlight the technical, organisational and process issues that underlie the ‘blockage’
to disruption.
For operators with an ambition to disrupt the market, there is a need to construct a matching set of
institutional processes and to create a culture of disruption. This poses some difficult questions for senior
management. What kind of disruption do you want? How and where will it be created? We list some
options. We believe that the end game of a successful disruption is a business transformation to a
‘software telco’ business model. The first step is to learn to exploit ‘quality arbitrage’.
3. 3About the authors
Dr Neil Davies
Co-founder and Chief Scientist, Predictable Network Solutions Ltd
Ex: University of Bristol (23 years).
Former technical head of joint university/research institute (SRF/PACT).
Peter Thompson
CTO, Predictable Network Solutions Ltd
Ex: GoS Networks, U4EA, SGS-Thomson, INMOS & Universities of Bristol, Warwick, Cambridge and
Oxford.
Authority on technical and commercial issues of converged networking.
Martin Geddes
Founder and Principal, Martin Geddes Consulting Ltd
Ex: BT, Telco 2.0, Sprint, Oracle, Oxford University.
Thought leader on the future of the telecommunications industry.
4. 4Contents
Embracing the disruption process
The ‘perception gap’ of disruption
Exploring the ‘perception gap’
The journey to a disruptive model
Appendices
1
2
3
4
7. 73M built a management system for innovation
When 3M created the Post-It note, they achieved something that had previously been
unthinkable. Critical to this advance was a management system that was configured to
enable and encourage such innovation. (Whether it was ‘disruption’ for 3M is, admittedly,
debateable – however it did alter the stationery market!)
The culture and processes at 3M legitimised such change. For instance, they allowed
“bootlegging” of ideas between departments, rewarding “innovation without permission”.
Many network operators have indicated a desire to engage in a similar leap ahead of their
competition. We propose network performance science is the new “digital adhesive” to
enable such a radical advance in business capability and results.
Our belief is that the management system of most network operators is currently configured
to reject this disruption. This presentation explores the systemic barriers to progress that we
see operators facing, and proposes a way forward.
8. 8Organisational progress is ‘culture-bound’
USS Whampanoag
(later renamed USS Florida)
Source: Wikipedia
1From “Men, Machines and
Modern Times” by Elting E. Morison
Contrast 3M’s experience with that of the US Navy, which in the
1860s created one of the most innovative vessels ever afloat, the
USS Whampanoag. This purpose-built steamship easily
outperformed all sail boats and first-generation steam vessels.
Yet is was vigorously rejected by the naval establishment as a
cultural misfit: “Lounging through the watches of a steamer, or acting
as firemen and coal heavers, will not produce in a seaman that
combination of boldness, strength and skill which characterized the
American sailor of an elder day…[this ship is] a sad and signal failure,
and utterly unfit to be retained in the service.”
As the author1 notes: “What these officers were saying was that the
Wampanoag was a destructive energy in their society.”
9. 9Two valuable new technologies
Accepted Rejected
Why the difference?
Low-tack adhesives Steam-powered ship
10. 10Why the difference?
Our experience is that the telecoms industry is much more like the US Navy example than 3M.
People see little or no room for radical improvement. Any and all external attempts to disrupt
the status quo will be seen as a “destructive energy” that threatens both the individual’s
role identity as well as the collective social structure.
Yes, there are many small and medium sized technical advances. But these are typically are
limited to a single department’s sphere of influence. Real disruption has a larger sphere of
influence. It cannot be created by “pushing harder” on the existing processes. Instead, it
initially requires new (parallel) structures to incubate and nurture it.
Hence we repeatedly observe a “perception gap” as to what is possible between network
operators and ourselves as network performance scientists. It is a difference in the way we see
and make sense of network cost and QoE optimisation. We believe that understanding this
gap is key to enabling radically better technical and commercial outcomes.
11. 11A language for thinking about progress
Fundamentally impossible
(e.g. faster-than-light communication)
TECHNICALPROGRESS
State of the art
What is successfully
deployed in at least one
operator in the world
State of the possible
State of play
Your current
deployed reality
Deployed as proof of
technical principle, but not
yet in commercial practise
See appendices for examples.
12. 12These are subject to different constraints
Non-negotiable limits of physics and mathematics
TECHNICALPROGRESS
Constrained by
technology
development
& maturity
Constrained by
strategy, cash
and execution
Constrained by
theory and
understanding
State of
play
State of the
art
State of the
possible
13. 13Overcoming constraints has different impactsTECHNICALPROGRESS
State of
play
State of the
art
State of the
possible
INNOVATION
IMPROVEMENT
DISRUPTION
14. 14
The problem of measuring
and modelling network ‘quality’
The issue of quality in networks has been long being troublesome, resulting in endless
deferral. For example, issue after issue of CCITT coloured books in the 1980s and 1990s had
sections about quality marked ‘for further study’. Those issues remain unresolved to this day!
The packet networking pioneers knew they needed compositional approach, where ‘quality’
where could be ‘added’ and ‘subtracted’. This lets you reason about demand and supply:
how much ‘quality’ is needed for a particular purpose, and how it can be shared out. It was a
hard issue as the underlying mathematics was insufficient to support their ambitions. This
meant that to make progress, the goal had to change: it became about mechanisms not
outcomes. The spread of IP meant the history and original goals were lost.
We have returned to those original goals, identified the gap in the conceptual formulation of
‘quality’, and then worked on filling that gap with suitable mathematical foundations. The
culmination of that work is the ∆Q framework that underpins network performance science.
15. 15Network performance science is a disruptionTECHNICALPROGRESS
State of
play
State of the
art
State of the
possible
NEW
MECHANISMSBETTER
PROCESSES
BREAKTHROUGH
THEORY (ΔQ)
16. 16
Advanced operators face a paradox:
past success can inhibit future progress
We have worked with some
of the world’s biggest, most
successful and most
sophisticated network
operators. Any institution
running critical national
infrastructure it is by its
nature quite conservative.
As we shall soon see, this
situation (paradoxically)
creates a psychological
barrier to progress.
Small perceived room for improvement
State of
play
State of the
art
17. 17
There is generally no process to support
real ‘disruption’ at network operators
People will engage with
“disruption” if the
anticipated benefits exceed
the costs. Yet the
“disruption” activities we
have observed at large
operators do not seem to be
structured to initiate the
process of disruption itself.
Instead, they act as defences
against being disrupted by
external agents.
State of
play
State of the
art
Believed state
of the
possible
Perceived room for innovation, but not disruption
19. 19
Different cultural assumptions
about how much progress is possible
State of
play
State of the
art
State of the
possible
State of
play
State of the
art
State of the
possible
US Navy view 3M view
20. 20
Different ways of engaging with
higher-order technical change
US Navy
in 19th century
3M
in 20th century
No process
for disruption
Institutionalised
process for disruption
21. 21
The ‘perception gap’ that we see
as network performance scientists
State of
play
State of the
art
State of the
possible
State of
play
State of the
art
Typical network
operator’s
perception
Our perception
Huge untapped
scope for
disruption
State of the
possible
GAP!
22. 22Example: QoE visibility from network metrics
State of
play
State of the
art
State of the
possible
Very weak QoE proxies, slow feedback.
Typically single-point probe data with 15 minute average
utilisation collected every 30 minutes (at best).
Weak QoE proxies, faster feedback.
Single-point probe data with 10 second average utilisation
collected every minute.
Strong universal QoE proxy, immediate feedback.
Multi-point probe data with instantaneous quality
probability distribution with 10 second collection delay.
So what? The difference means months/years of capex-free network growth!
23. 23Example: Packet scheduling
State of
play
State of the
possible
Little or no QoE assurance Strong QoE assurance
So what? The difference can mean major new markets and revenues!
25. 25
Nobody likes to be told they
could be doing a lot better
Over the years, we have worked with several global operators. The
aim of every operator is to extract the highest possible usage whilst
still delivering the required customer quality of experience (QoE).
We have reviewed the optimisation approaches of operators with the
relevant technical and commercial staff. We typically start with
“trouble to resolution” type business processes, where there are
customer experience issues whose source cannot be located.
What we often find is a struggle to engage fully with us: there is a
natural defensive posture from having external critique. Note that the
individuals concerned are not to blame: this is a systemic problem.
So, why is this?
26. 26We have found three blocking issues
Thinking about the potential application of disruptive new
techniques, we commonly find three issues that inhibit progress:
1. The perception gap
“Disruption? What disruption? I don’t see any disruption!”
2. Paradox of perfection
“We’re already the best, so why would we need outside help?”
3. Organisational alignment
“What’s in this ‘disruption’ thing for me, anyway?”
We will now expand on these to give you a better sense of how and
why network operators are (from our perspective) ‘stuck’.
See appendices for details of the underlying technical, organisation and process issues.
27. 27Issue #1: The perception gap
Recent developments in network performance science allow a major leap in performance. This
is a fundamental advance, like spread-spectrum, DWDM, or datagrams. Where these created
new resources, network performance science creates new ways of sharing such resources. It
enables a “state of the possible” where the following are true for any operator:
1. You have visibility of your contribution to delivered QoE and the success (or otherwise)
of the customer experience.
2. You have complete control over any trade-offs over network cost and QoE.
3. You are able to define the appropriate trade-offs for different customers and market
segments.
4. You are able to concurrently deliver all of those service levels, at the lowest collective
cost, by exploiting all of those trade-offs.
5. You are able to isolate any problems that do arise.
How does the typical operator’s “state of play” compare to this ideal?
28. 28The ‘perception gap’ of the ‘state of play’
Operator Perception Our Perception
Visibility of QoE Highly visible Partially visible, at best
Control over
trade-offs
Strong Good at short timescales, weak at longer ones,
and not connected directly to QoE
Definition of
demand
Adequate Have not captured the QoE limits of different
segments, so delivering maximum QoE to
everyone (at highest cost)
Optimal supply Lowest possible
contention
Highest possible cost; many ‘good’ resource
trades not exploited; transmission assets stranded
Isolate QoE issues Not seen as a
problem
Customers are complaining and churning, yet root
cause of QoE issues is not visible
See appendices for details of the cost and QoE perception differences
29. 29Issue #2: Paradox of perfection
Most large operators have highly competent staff, and honourable
intentions of good service delivery. This means they aim high: the service
quality goal they aspire to is “best of the best effort”. In doing so, they have
extracted all (or nearly all) of the “standard” optimisations.
The staff rightly believe they are at or near the state of the art. The problem
is that, in keeping with the rest of the industry, such staff are really highly
skilled and numerate craftspeople, rather than engineers applying an
underlying science. We often experience a low curiosity on how to
improve, and a fear of loss of face when such an improvement is proposed.
As a result we often find a strong resistance to having the service’s true
QoE being measured, along with a rejection of the applicability of the
science to the organisation (despite its repeatedly proven benefits
elsewhere).
30. 30
Why this intense resistance
to radical improvement?
There is a tension the operator’s Board and senior management are holding: the price of
believing “we are the best” is a lack of inquisitiveness of how to disrupt the status quo. This
means they have no way of understanding the cost optimisation that is truly possible. Why so?
Our experience of working with many large operators is that any shortfall from the state of the
art is seen as a failure. Performance hazards of packet networks are perceived as being
faults, rather than a normal emergent behaviour of a stochastic system. This sets up an
unfortunate dynamic, whereby staff are highly defensive. Incentives are biased towards hiding
issues to avoid punishment for failure; not rewards for exploration and new learning.
What is required is for people to become aware of the (cost and QoE) gap between the “state
of the art” and the “state of the possible”. This requires the development of a culture of safe
critical self analysis. Once the awareness of the gap exists and is widely accepted, it would be
natural to want to quantify it.
31. 31Issue #3: Organisational alignment
There is an apparent issue of alignment between the
development needs of large operators at the group level, and the
managers who own the P&L for each product line or geography.
Our sense is that the internal business owners are unwilling to
engage with a market disruption project because it is not their
business objective. They are unsure how to handle that change
to their purpose. How will it impact their career if they take on
this (seemingly) risky work for the benefit of the Board or Group?
Since it doesn’t fit well with the existing paradigm, operators
need to create space for a new paradigm to grow.
33. 33Key questions for operator senior management
1.How to create disruptive new products without
having to overcome all the internal resistance first?
2.What is the best way to technically validate the
scientific approach?
3.How to generate the hard data necessary to
quantify the commercial benefits of a disruptive
technical strategy?
34. 34What kind of disruption do you want?
Develop
incremental
products based
on current
product set?
34
Develop
radically new
& highly
disruptive
products?
35. 35Where and how will disruption happen?
Force disruption
onto current
structure using
positional power?
35
Incubate disruption
outside current structure?
Invest in
disruption away
from current
structure?
36. 36Where might this journey take operators?
We see network performance science as a transformational business opportunity, not just a
profound shift in technical approach. The immediate opportunity is two-fold:
• A range of potential new product offerings that are segmented by performance,
i.e. versioned by quality and/or resilience.
• A refactoring of business processes from ‘craft’ to ‘science’.
These new products may be created by the operator itself, or by other people using the
operator’s platforms. Furthermore, the underlying infrastructure may be the operator’s, or
that of third parties. In either case the operator is extracting a ‘quality arbitrage’ that
typically exists in all IP networks.
The end game we see is the construction of a ‘post-telecoms’ business: a new generation of
‘distributed computing service provider’, where you dynamically control the matching of
supply to demand along a whole supply chain.
37. 37The end game: the ‘trading platform’
As HBR has observed, the real money is in being a “network
orchestrator” (in the wider sense of “value network”, not
just telecoms network.) SDN/NFV are example technologies
of the shift of “dynamic trading”. But who controls these
mechanisms and prices the trades?
In the new model, you can do end-to-end supply chain
management of QoE, not just the logistics of your own
network. Furthermore, you can differentiate by
quantifying the QoE benefit & show the experience
difference between your own products and those of
competitors.
Executing a ‘quality arbitrage’ play is the first step to a
highly disruptive new business model.
41. 41Example stages of technology development
Description Examples
State of play Your “legacy”: any technology or process
you have deployed, however new.
2G, 3G, 4G, DSL, FTTx, MPLS
State of the art Underlying technical concepts will be
documented in published papers and
textbooks. Aspects of tech are in standards.
SDN/NFV
State of the possible Work that is in R&D or pre-deployment and
there is a working proof of concept. May be
proprietary, and seen as “controversial”. Pre-
standard.
Aspects of 5G, Recursive
Internet Architecture (RINA),
Contention Management
(CM), Quantum computing
Fundamentally
impossible
Falls outside the limits the universe imposes
on us: physics and mathematics
N/A
43. 43Underlying technical issues
Operators generally cannot “see” the performance in the
customer’s terms, since the metrics being used lack the necessary
fidelity. Indeed, all operators lack the tools to give visibility of the
performance hazard space (and hence customer experience).
As a result, operator control over QoE and cost is significantly
limited by the fidelity of the current measurements and mechanisms
being used. The assumed optimality cannot be depended upon.
Furthermore, the mechanisms in use only expose a portion of the
“good” trades (over the many timescales), so the “frontier” of
possible QoE and cost is restricted compared to the ideal.
Resolution: Higher fidelity measures, better mechanisms and fuller
exploitation of the available resource trades.
44. 44Underlying people issues
Few operators have institutional experience of using high-fidelity
instantaneous measures and doing network performance engineering
using them. What we are suggesting is beyond the received and
accepted wisdom of what is possible; aspects of it can be found in the
literature but not yet in books on the subject.
There are two resulting sets of human anxieties to deal with:
• The fear of the unknown, and the risk of trying something new.
What’s in it for any individual staff member or team that tests
something new and untried to the organisation?
• A possibility of loss of face if that process calls into question prior
expertise. How do I explain the shortfall in what I was doing before,
when I am the company’s domain expert?
Resolution: Skills transfer in a “safe” context (e.g. new business area).
45. 45Underlying process issues
The processes within network operators are not configured to
think in terms of performance hazards and their relationship to
the network trading space.
Thus there is a costly disconnect between the business and
the network operation. The potential to capture different
customer QoE and cost requirements, and deliver that as a
portfolio of services, is being missed.
All operators are aware of the risk of cannibalisation between
products, and have business processes to mitigate this. Yet there
are no processes actively looking for arbitrages to exploit.
Resolution: New metrics and methods that simultaneously
represent the customer experience and the network performance.
47. 47Exploring the perception gap
There is a physical reality to both the network operation and
the QoE delivered. Since we cannot track every packet and
application use, we use proxies to manage the network.
Common proxies might be average link utilisation or packet
loss rates.
We then use these proxies to make trade-offs of cost and
quality. These trade-offs require making resource
allocation decisions at all timescales, as per the chart shown
to the right. (See separate presentation for more detail.)
The pervasive industry use of low-fidelity proxies to QoE
results in a difference in perception (between networks
operators and us) as to how well optimised the network is in
terms of QoE and cost.
From “Get more out of the network”
48. 48Typical operator perception of QoE and cost
QoE
“The best available metrics and methods have been used
to deliver a system with very high QoE. Over-provisioning
ensures low levels of contention and negligible packet
loss in the core network. Customers do complain, but
there is no evidence that their performance problems
are due to how we manage our network.”
Cost
“The system has been fully optimised: it is being run as
‘hot’ as possible whilst still delivering the intended QoE.”
49. 49QoE: our perception of the typical operator
In general, operators are managing well to their current metrics. However, there is no
measurement or modelling of the very short-term contention effects in the network.
Therefore the presumed visibility of customer experience is lacking.
There are often disputes with customers over QoE problems, but there is no scientific system
for attribution of blame. Customers are seeing direct effects of these QoE issues, and are
(rightly or wrongly) blaming the carrier. Often these issues may be due to on-premises WiFi
access, or transient application or device ‘freezes’, and nothing to do with the operator
network.
The method being used to manage QoE is to keep the link busy-period down, which lowers
contention. However, this leaves operators’ services at the mercy of the packet arrival
patterns generated by their customers. Performance hazards can be armed by customers,
without any visibility available to the operator.
50. 50Cost optimality: our perception
We see much room for cost improvement at the typical operator:
• The short-term statistical properties are not visible, so there is no visibility of short-term
QoE breaches.
• There is only one QoE control lever being used, which is utilisation and lowering of the busy
period.
• There is insufficient visibility of the relationship between utilisation and delivered QoE.
As a consequence, most networks are likely to be flipping between running too idle (wasting
capex cost), and too hot (a precursor to churn). By over-provisioning to eliminate contention,
the network operator has unwittingly taken a cost-maximisation route to network design.
Operators can have more “levers” over QoE by taking control of more of the resource trades at
all timescales. They can also gain better visibility of QoE. This combination allows for safe
lowering of costs.