You are on page 1of 136

Advanced Analytics

in network
operations
USE CASES LIBRARY
Executive summary

• This document has been prepared in the context of an engagement executed with Telefónica Global
Resources and Telefónica España operations teams aimed at identifying, assessing and prioritizing
advanced analytics use cases in telecommunications network operations

• The document consists of two main chapters


o Chapter number 1 provides a detailed description of 22 uses cases which have been assessed as high
potential in the context of Telefónica (Global and España). It also provides high level business cases and
an standard implementation roadmap for 11 prioritized use cases which provide examples of potential
use cases to industrialize leveraging Telefónica España’s capabilities

o Chapter number 2 provides an overview of all 142 identified use cases with a detail on situation,
potential approach and expected impact levers for each of them

• As an annex to this document, an excel file is provided to further support Telefónica in assessing all use
cases and ideas identified during this effort

2
Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

3. Use cases assessment guide

3
Use case have been assess identifying 20 high potential cases

Use cases proposed for deep-dive phase


High impact / High deasibility High impact / Low feasibility
Use case
Low impact / High feasibility Low impact / Low feasibility
Strategy 1 Smart mobile CapEx allocation based on client and traffic data
Smart CapEx
& network development
2 Smart short term reconfig. of mobile network based on granular profiling

142 3 I+M capacity optimization based on short & medium term events forecasting
Field force
optimization 4 Smart management of field force including optimization of dispatching and technicians journeys
18 Execute
Field 5 CTO's saturation prevention model based on forecasted network deployment and commercial campaigns
operations Works analysis
prioritization 6 Deployments and FARO activities prioritization based on customer traffic pattern analysis vs current
premises analysis
Logistics 7 Dynamic stocking for network equipment and spares (analytics SCM) based on forecasts on deployment /
51 Test & try optimization decommissioning
8 CC calls reduction via automatic diagnosis/problem solutions based on home equipment status
information
Incoming calls &
9 TMO reduction by identifying breakdown cause and providing remote solutions
TMO reduction
Aftersales & 10 CC calls reduction by information calls id to route them appropriately and/or provide information
10
Mature customer
care
11 Ticket generation reduction by crossing network and EDC data to avoid dispatching to client facilities in a
network incidence
Diagnosis &
ticketing reduction 12 I+M costs reduction by improving detection and mgmt. of massive failures
13 Decision making optimization with global technical centers dashboard
63 Asset mgmt. 14 IT HW/SW opex optimization via an analytics based asset management
Further understand
O+M works 15 O+M resources prioritization based on economic impact of incidents
prioritization 16 O+M resources optimization based on an outage duration predictive model
NOC / SOC /TI
17 Network opex optimization using a predictive maintenance and root failure cause analysis
O+M resources
prioritization 18 Network opex optimization using a predictive maintenance model based on service degradation
1 Business feasibility is referred to Telefónica España

SOURCE: McKinsey team analysis 4


1 High potential use cases (1/2) Use cases proposed for
deep-dive phase

Net run-rate impact


Use case Sources of impact Baseline (T.E.), k€ Impact, % (T.E.), k€

Strategy 1 Smart mobile CapEx allocation based on client and traffic data • Optimization of CapEx 145,000 2,0% 2,900
& network development
Smart CapEx
2 Smart short term reconfig. of mobile network based on granular • Optimization of reconfiguration & N/A N/A N/A
profiling optimization costs

Field 3 I+M capacity optimization based on short & medium term events • I+M productivity increase and thus 91,000 3,8% 3,440
operations forecasting reduction of price per work
Field force
optimization 4 Smart management of field force including optimization of • I+M productivity increase by reducing 91,000 1,3-2,6% 1,180-2,380
dispatching and technicians journeys journey times

5 CTO's saturation prevention model based on forecasted network • Reduction of delayed revenues from FTTH 820 45,1% 370
deployment and commercial campaigns analysis provision

Works prioritization
6 Deployments and FARO activities prioritization based on • Increase of revenues from better targeted N/A N/A N/A
customer traffic pattern analysis vs current premises analysis commercial campaigns and deployments

7 Dynamic stocking for network equipment and spares (analytics • Reduction of current assets (stock) 45,000 7% 2,990
Logistics SCM) based on forecasts on deployment / decommissioning • Reduction of urgent transportation costs 1,200 11,7% 140
optimization

Aftersales & 8 CC calls reduction via automatic diagnosis/problem solutions • Reduction of incoming customer care call 17,600
customer based on home equipment status information (WiFi connections) center calls and thus related I+M works and
care equipment 3,4-9,4% 600-1,660
Incoming calls
9 TMO reduction by identifying breakdown cause and providing • Reduction of TMO from incoming customer 7,600
reduction
remote solutions care call center calls
10 CC calls reduction by information calls id to route them • Reduction of incoming customer care call 4,500 18,6% 840
appropriately and/or provide information center calls

5
1 High potential use cases (2/2) Use cases proposed for
deep-dive phase

Net run-rate impact


Use case Sources of impact Baseline (T.E.), k€ Impact, % (T.E.), k€

Aftersales & 11 Ticket generation reduction by crossing network and EDC data • Reduction of I+M dispatching to client sites 660 24,2% 160
customer (B2B) to avoid dispatching to client facilities in a network on network incidents
care (cont.) incidence
Diagnosis &
12 I+M costs reduction by improving detection and mgmt. of • Reduction of ticketing and I+M dispatching N/A N/A N/A
ticketing reduction
massive failures during massive failures

13 Decision making optimization with global technical centers • Optimization of decision making & data N/A N/A N/A
dashboard processing for other use cases

NOC / SOC /TI 14 IT HW/SW opex optimization via an analytics based asset • Reduction in SW licenses renting & HW 35,300 9,0-16,5% 3,180-5,850
Asset mgmt.
management procurement

15 O+M resources prioritization based on economic impact of • Reduction in penalties from network 70,3001 3,5% 2440
incidents availability & SLAs (BRA)
O+M resources
prioritization
16 O+M resources optimization based on an outage duration • Network availability prioritization of N/A N/A N/A
predictive model resources

17 Network opex optimization using a predictive maintenance and • Reduction of O+M operating costs by 17,500 N/A N/A
root failure cause analysis preventing network failures
O+M resources
prioritization
18 Network opex optimization using a predictive maintenance • TBD N/A N/A N/A
model based on service degradation

Total impact ~17 Mn.€

1 Business case based on Telefonica Brazil

6
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC091

1 Smart mobile CapEx allocation based on client and traffic data IMPLEMENTED

DESCRIPTIVE

Situation Approach Impact

• Rationalize current assets and optimize • Analysis of traffic usage, customer • Impact generated
the new deployment in order to reduce value, seasonality, displacements, o CapEx optimization to stretch budget
Capex OMV’s traffic, AFR traffic, roaming efficiency: ~2%
traffic and location of competition’
• Value drivers o Equivalent to 4 Mn. € /year of
antennas
o Infrastructure sharing agreements additional CapEx
• Sources of data:
o Assets valuation
o Internal: voice and data traffic,
o Deployment Prioritization
client value, geolocation, antenna
o CAPEX /year: 200 Mn. € connection…
o External: infoantennas, signal
intensity, network layout

SOURCE: Telefonica España 7


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC040

2 Smart short term reconfiguration of mobile network based on granular profiling of IMPLEMENTED

network nodes DESCRIPTIVE

Situation Approach Impact

• Decide which networks need to be • Network information is gathered at • Decrement of network capacity overload
expanded is a critical task of the network node / layer level and monitored (node situations, and therefore, an increment
strategy department profiling) in service quality and availability
• A deep knowledge of traffic patterns and • Node data monitored: Network Nodes • Better usage of network elements by
trends is needed to be able to expand profiling: top sites, sensitivity to FWA reconfiguration of underused elements
the network capacity to handle an services, to MVO, customers, to
increase in demand without impacting Roamers…
service quality
• Layer data monitored: sensitivity to
• Detect network areas that are being wholesale customers, evolution
underused and therefore can be status…
refarmed is also a source of value
• The information is presented with
• The more granular the information it is, interactive visualization to facilitate the
the better the decisions will be process of understanding the data and
getting insights

SOURCE: Telefonica España 8


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC136

3 I+M capacity optimization based on short & medium term events forecasting

Description Impact

• Helping I+M subcontractors size their workforce Description k€


provides a negotiation lever to reduce costs Development
• Prescriptive model that determines for each Set-up • Model set up (240)
subcontractor: Cost
o Optimal total number of technicians Set-up cost (240)

o Optimal number of technicians / profile


Recurrent
(installation, maintenance)
• Increase technicians productivity by 5%, 2% and 1% the
o Optimal technicians geographical first 3 years respectively
distribution Run rate • Subcontractors negotiation: 50% of efficiency gains
savings • Run-rate for next 3 years respectively over a baseline
• Inputs required (historic subcontractors field 91M
3470
force by typology / region, commercial activities)
are available on Exadata
Run rate • Model maintenance (30)
maintenance

Net recurrent benefit 3440


Payback Year 1

SOURCE: Telefonica Global, McKinsey 9


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC136

4 Smart management of field force including optimization of dispatching and technicians


journeys

Description Impact

• Helping I+M subcontractors optimize the Description k€


dispatching of their workforce provides a Development
negotiation lever to reduce costs
Set-up • Model set up (240)
• Prescriptive model that optimizes the assignment Cost • Additional cost could be incurred to develop a
of technicians to work orders based on dispatching tool
Set-up cost (240)
o Technician profile
Recurrent
o Service type
• 5-10% journey time reduction
o Distance to customer property • 20% automation of dispatchers assigning technicians
time
• Inputs required are available on different data • Subcontractors negotiation: 50% of efficiency gains
Run rate
sources • Run-rate for next 3 years respectively over a baseline
savings 1210- 2410
91M
• Additional impact can be identified if further study

Run rate • Model maintenance (30)


maintenance
Net recurrent benefit 1180 - 2380
Payback Year 1

SOURCE: Telefonica Global, McKinsey 10


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC102

5 CTO's saturation prevention model based on forecasted network


deployment and commercial campaigns analysis

Description Impact

Complete anticipation
• CTOs allow for multiples fiber connections (own
Design anticipation
or wholesale) in one specific area, with a limit
Description k€ k€
• When a CTO is saturated, no more connections Development
can be created and therefore potential new
Set-up • Model set-up (320) (320)
customers can’t connect to Telefonica’s network
Cost • Robot set-up (130) (130)
until a new CTO is built

• A predictive model should be created to detect Set-up cost (450) Set-up cost (450)
which CTO’s are more likely to be saturated in a
close period of time Recurrent
• Advancement revenues 890 440
• Input to the model includes data from BI, EDICRE Run rate
and CELSIX savings
• Once prediction has been made, design and work
orders will be automatically generated • Model maintenance (30) (30)
Run rate
maintenance • Robot maintenance (40) (40)

Net recurrent benefit 820 370


Payback Year 1 Year 1

SOURCE: Telefonica Global, McKinsey 11


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC137

7 Dynamic stocking for network equipment and spares (analytics SCM)


based on forecasts on deployment / decommissioning

Description Impact

• Urgent journeys of stock suppose ~1 – 1.5 Mn.€ a Description k€


year Development
• Optimization of stock distribution and stock Set-up • Model set up (280)
holding to reduce logistic costs Cost • Additional cost could be incurred to develop a stock
management system
• Prescriptive model that determines real time, Set-up cost (280)
geographical stock planning
Impact on • 7% stock cost reduction in the next 3 years 2990
• Inputs required to this model are available on balance sheet
different sources (TeleSap, Edicre, CELSIC..)
Recurrent
• 5% reduction in urgent journeys cost 170
Run rate • Run-rate of 3 years:
savings

Run rate • Model maintenance (30)


maintenance

Net recurrent benefit 140


Payback Year 2

SOURCE: Telefonica Global, McKinsey 12


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC080

8 9 CC calls and TMO reduction via automatic diagnosis/problem


solutions based on home equipment status information

Description Impact

• New deployed home equipment has very Description k€


advanced variable monitoring and remote control Development
TMO reduction
tools that can be used to get information from 10%
home route performance (e.g. ASSIA in HGUs). Set-up • Model set-up 15%
(2.420)
20%
Cost
• Using that information to develop an advanced
analytics predictive model enables to predict
upcoming access faults and better identify
diagnostics Recurrent Scenario 1: 1% reduction1 Payback Year
340 8
• Predictive modelling allows to take proactive Run rate savings 600 4
actions like reboot, proactive dispatching or even 4
• Calls (~1.9M agent calls) 860
proactive dropship to prevent the fault
• I+M activity (~310k field visits)
• In case clients still call, predictive modeling will • Substitutions (~36k replacements) Scenario 2: 3% reduction
guide agents with the most probable diagnostic Net benefit run • TMO (~1.9M agent calls) 880 4
to give a better and faster answer to the client rate 1.130 3
Run rate costs 1.390 3
• ASSIA maintenance Scenario 3: 5% reduction
• Model maintenance
1.410 3
1.660 2
1.910 2
1 Reduction impacts calls, field visits and substitutions

SOURCE: Telefonica Global, McKinsey 13


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC165

10 Call center calls reduction by information calls id to route them


appropriately and/or provide information proactively

Description Impact

• Call centers receive a massive number of calls


from clients nowadays: ~11,3 M agent calls/year.
Description k€
~30% of the calls received are information
questions Development
• Using historical client data and products owned Set-up • Model set-up (410)
information (e.g.: ASSIA) and developing a Cost
predictive model enables to detect clients more
likely to call 1002 for information after specific
past events
Recurrent
• Proactively sending the client information before
he even calls (SMS, email, etc.) or using an • Calls: 20% reduction over ~1.1M structured 910
automatic process in case the client proceeds Run rate information agent calls (10% of total agent calls
with the call, will reduce significantly information savings ~11.3M calls)
calls getting to an agent

Run rate • Model maintenance (70)


maintenance

Net recurrent benefit 840


Payback Year 1

SOURCE: Telefonica Global, McKinsey 14


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC142

11 Ticket generation reduction by crossing network and EDC data (B2B) to avoid
dispatching to client facilities in a network incidence

Description Impact

• Client equipment platforms (EDC) are


traditionally not interconnected with network
Description k€
monitoring systems
Development
• As a result, long technical calls lead to sending
Set-up • Data set up (65)
technicians to client home when a core network
Cost • Model set up (20)
incidence happens (and there is nothing they can
• Integration set up (200)
do)
Set-up cost (285)
• Combine network and EDC databases to create a
complete picture of client and network at the Recurrent
moment of the incidence. • 100% reduction of massive network breakdowns: 910
o Calls (IVR)
Run rate
• Use a descriptive model to find if the incidence o I+M visits (45% of calls)
savings
cause is in EDC systems or if it is a problem • Run-rate:
derived from the network

Run rate • Model maintenance (30)


maintenance • Integration maintenance (40)

Net recurrent benefit 160


Payback Year 2

SOURCE: Telefonica Global, McKinsey 15


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC171

12 I+M costs reduction by improving detection and management of massive failures

Situation Approach Impact

• B2B customer care services sometimes • Descriptive correlation of client’s data • Avoiding double ticketing during a massive
lacks network failure information and takes (home address, location, nature of failure by ensuring information reaches
some time before realizing a number of failure, etc.) with core network data customer-care services and is correctly
client calls informing about an incidence are (current failures and affected nodes, etc.) correlated with client information may lead
actually related to a massive failure • Sources of data: to a significantly reduction of dispatched:
• Crossing data analysis of network and EDC TBD
o Client equipment
platforms to avoid sending technicians to • Value drivers:
client facilities when there is a core network o Location
o Cost drivers:
incidence and resolve incidences fasters o Failure description from client
and remotely without double ticketing or o I+M visits
o Network data & alarms
even automatically informing the client to • Revenue drivers:
avoid calls o Improve CSI

SOURCE: Telefónica & McKinsey team analysis 16


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC175

13 Decision making optimization with global technical centers dashboard

Situation Approach Impact

• B2B customer care services lacks a global • Descriptive analytical model to present all • Impact is generated by optimizing the day-
dashboard to aggregate and present real- active and recent failures information into to-day decision making with real-time
time data a single dashboard to help decision information
• A global dashboard could help in decision making process, e.g.: • Results of this analysis can be further
making by including failure information on: o Prioritizing dispatching injected into other models
o Affected client equipment information o Understanding problems with specific
mapped by OEM, geography, types of equipment
technology, diagnosis and solution o Events affecting the clients network
o Status of each failure o etc
o Pending intervention
o Quality survey results

SOURCE: Telefónica & McKinsey team analysis 17


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC115

14 IT HW/SW opex optimization via an analytics based asset management

Description Impact

• IT hardware & software capacity utilization has Description k€


been measured and in some cases is extremely Development
low (~5%)
Set-up • Model set-up (240)
• Gathering data based on capacity scanning tools Cost
to assess utilization of HW and SW at a per-unit
level, and running an optimization algorithm to
reallocate capacity, enables to increase asset
utilization Recurrent
• SW licenses: 5-10p.p utilization improv. leads to 9-17% 1430 - 2600
• Maximizing HW & SW asset usage reduces HW & reduction in SW licenses costs
SW supply costs and identifies HW candidates to Run rate (~ 15.7M €)
decommissioning which may lead to cash inflows savings
and decrease in energy consumptions • HW purchases: 5-10p.p utilization improv. leads to 9- 1780 - 3270
17% reduction in HW purchases (~ 19.6M €)

Run rate • Model maintenance (30)


maintenance

Net recurrent benefit 3180 - 5850


Payback Year 1

SOURCE: Telefonica Spain 18


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC112

15 O+M resources prioritization based on economic impact of incidents

Description Impact

• Incidents and faults affect different types of clients Description k€


and have different economic impact for the Development
company
Set-up • Model set-up (180)
• Gathering historical data about previous faults Cost
(e.g. client typology, incident typology, broken
SLA) and building a predictive algorithm to
estimate affected clients and level of impact
enables to identify critical incidences Recurrent
• Network unavailability time: 3.5% reduction over total 2460
• Changing incidence resolution prioritization unavailable time of ~108M min (cost of min is 0.64€,
process to minimize economical impact has Run rate includes lost of revenues and penalties)
significant reductions on penalties and revenue savings
losses

Run rate • Model maintenance (20)


maintenance

Net recurrent benefit 2440


Payback Year 1

SOURCE: Telefonica Spain 19


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC036

16 O+M resources optimization based on an outage duration predictive model


PREDICTIVE

Situation Approach Impact

• Currently tickets are prioritized based in the • Gather historical data about past outages
actual outage duration, with no information and the time needed to fix them.
of how long will it be • Only basic historical outage information
• By following this method critical incident can be used to properly model the status
are not prioritized until the outage has been of uncertainty that happens when a
going for a considerable amount of time outage happens (e.g.: what has • Reduction of revenue loss due to customer
• By knowing beforehand the duration, long happened?). churn by TBD%
expected running tickets can be prioritized, • Some examples of data to be used: • Faster and more precise answer to clients
increasing customer satisfaction o Location of the outage and area affected by the same outage
affected (or disconnected)
o Network variables at the moment of
the outage
o Force field status
• Use a predictive model to get a estimation
of the outage duration / time needed to
recover from the fault

SOURCE: Telefonica Global Resources 20


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC114

17 Network OPEX optimization using a predictive maintenance and root failure cause
analysis

Situation Approach Impact

• Massive network faults or malfunctions • Use historical information to develop an • Decrease of massive network failures by
have a critical impact in customer advanced analytics predictive model to preventing them will mainly decrease O+M
experience, service and repair costs predict an upcoming massive fault operating costs
• Some of those faults would have been • All network elements will be evaluated o O+M OPEX: 17.5 M € workforce cost
easily prevented if they were known regularly, and for each one the  “EEBB”: 12.6 M €
beforehand, with a much lower cost (in probability of an upcoming massive fault
terms of economics and in terms of client in a short period of time (TBD) will be  “Planta Interna”: 5.9 M €
experience) evaluated • By preventing critical faults, service quality
• By digging at network data gathered by • Those elements with a high probability of and client satisfaction will also increase
NOC/SOC, is easy to find strange patterns fault will be reviewed in detail by • Life cycle of network elements will also
long before the massive fault happens maintenance teams, trying to locate root increase
cause of the fault and eventually
preventing it from happening

SOURCE: Telefonica Spain 21


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC114

18 Network opex optimization using a predictive maintenance model based on service


degradation

Situation Approach Impact

• Network faults are usually the consequence • Define different performances KPI for • Reduction in number of fault cause by small
of a prolonged element malfunction over different network elements malfunctions by TBD %
time
• Monitor KPI evolution over time to • Overall performance increase of network
• Those malfunctions do not have a direct detect decreasing patterns trends assets
impact over the network quality, but they
are slowly damaging other network • Perform a root-cause analysis over
elements decreasing trends to understand causes
of those decrements
• Fixing those malfunctions do not have an
immediate impact over network quality,
rather than a very high impact over time

SOURCE: Telefonica Spain 22


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

23
1 The Telefónica & McKinsey team involved in the effort has identified a total of 142 use
cases
Organización Telefónica España
Estrategia y desarrollo de red
Operaciones
Postventa
Soporte y Expl. Red, Serv. y Sist.
Advanced analytics use cases per main areas and type of use case, number of use cases

Network strategy & Mobile Call center Technical centers


Fixed deployment I+M Ops. NOC / SOC / TI O+M Ops. Sites
planning deployment operations ops/mgmt.

Descriptive 6 2 2 5 11 24 5

Predictive 2 4 1 4 6 2 10

1
Prescriptive 8 4 4 9 11 8 1

Total 16 10 1 10 20 24 42 1 6

Non-operations 12 Total use cases 142

SOURCE: Telefónica Global Resources; Telefónica España; McKinsey & Company 24


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

25
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC007

Smart geolocalized business alarms and visualization

Situation Approach Impact

• In order to detect and prioritize alarms Sources of data: • Value drivers:


and business opportunities there should • Internal: adds, churns, arpu, BAF and o Reduction of local churn: 0,2%
be a geographical visualization at the mobile quotes, CSI, breakdowns, o Determinations of areas to develop
local level of business and competition claims, copper network quality… o Improve customer satisfaction
parameters • External: economy trends • Impact:
• Value drivers: (unemployment index), competitors’ o 2.9 Mn. € / year
o Local churn /month: 92000 FTTH deployment
o Determinations of areas to develop
o Improve customer satisfaction

SOURCE: Telefónica & McKinsey team analysis 26


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC022

Capex allocation- micro market optimization

Situation Approach Impact

Telcos need to start looking at their networks • Decisions (both network and • Value drivers:
at ‘micro-market’ level…. commercial) to be made at a o Increase in revenues with same capex
Accelerate Mobile Deploy alternative Maintain “geographical micro-market” level o Improving Network quality
NGN deployment Fixed technology current network
• Individual business cases at each
micro-market need to be based on • Impact:
multiple variables e.g., commercial o ROI uplift of 10%
drivers, competitive dynamics, o NQI improvement of 5%
network characteristics and
operational KPIs
• Combination of network and customer
experience KPIs could reveal powerful
insights for Operators currently
untapped

SOURCE: McKinsey Telcos 27


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC026

Radio Access: Planning


CDO

NEPTUNO: Optimizing CAPEX for RAN

• Network and Marketing simulations • Geographic coverage • Forecasted demand vs planned • Current and potential sites
• Network, traffic and clients data • Space discretization network • “per client” simulations
(ARPU, device) • Coverage maps • Regulatory commitments and • Available tasks to optimize experience
• Interaction among coverage areas obligations
• Available tasks
• Congestion simulations

28
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC027

Capex optimization: investment model for batteries in critical sites

Situation Approach Impact

• Prioritization of sites according to • Analytical model based on: • Value drivers:


batteries investment o Technical criteria o 40% reduction in unavailability time
• CAPEX optimization in batteries o Business criteria
per Network Element
acquisition o Network availability not reduced
o External Factors
• Reduction of unavailability time per despite network elements plant
Network Element increased
• Value drivers:
o 40% reduction in unavailability time
per Network Element
o Network availability not reduced
despite network elements plant
increased

SOURCE: Telefónica Global 29


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC038

Strategy, planning, and capacity management

Operations, engi- High-load analysis and configuration data


neering, network
optimization

SOURCE: Telefónica Global 30


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC039

Frequency bandwith monitoring

Situation Approach Impact

• Monitoring the bandwith usage by area • Monitor the bandwith usage by area • Value drivers:
is important because areas with very and/or technology to indentify areas o Increase ARPU: upselling heavy users
high bandwith traffic are liable to be with low, medium or high bandwidth customers
expanded in a future. This information traffic.
can be used for a better price estimation • Analytical model based on:
in the bandwidth auctions o Bandwith usage/ area
• Value drivers: o Traffic Evolution per Technology
o Increase ARPU: upselling heavy users and Freq Band
customers o Network Statu
o Service Coverage

SOURCE: Telefónica Spain 31


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC040

Smart short term reconfiguration of mobile network based on granular profiling of network IMPLEMENTED

nodes DESCRIPTIVE

Situation Approach Impact

• Decide which networks need to be • Network information is gathered at • Decrement of network capacity overload
expanded is a critical task of the network node / layer level and monitored (node situations, and therefore, an increment
strategy department profiling) in service quality and availability
• A deep knowledge of traffic patterns and • Node data monitored: Network Nodes • Better usage of network elements by
trends is needed to be able to expand profiling: top sites, sensitivity to FWA reconfiguration of underused elements
the network capacity to handle an services, to MVO, customers, to
increase in demand without impacting Roamers…
service quality
• Layer data monitored: sensitivity to
• Detect network areas that are being wholesale customers, evolution
underused and therefore can be status…
refarmed is also a source of value
• The information is presented with
• The more granular the information it is, interactive visualization to facilitate the
the better the decisions will be process of understanding the data and
getting insights

SOURCE: Telefonica España 32


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC041

Network usage monitoring

Situation Approach Impact

• A monitor demand tool to visualize • Analytical model based on: • Value drivers:
multiple KPI geographically. This can be o Demand patterns for different o Help make business decisions
used to get multiple business services (reconfiguration, refarming,
conclusions. Very basic forecasting o Network usage evolution
expansion)
models allow for short term
o Spectrum usage
reconfigurations and/or planning
o Basic Forecasts
• Value drivers:
o Help make business decisions
(reconfiguration, refarming,
expansion)

SOURCE: Telefonica España 33


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC078

Infrastructure growth anticipation

Situation Approach Impact

• When special events happen • Predicted model to estimate future • Value drivers:
somewhere, sometimes there are demand increments o Develop proper infrastructures that
infrastructure / service capacity • Sources of data: avoid those problems before they
problems to satisfy the whole demand. o Historic demand/events
happen
This could be avoided by using a
o Future events
predictive model to predict future
demand increments. o Demand forecast/ event

• Value drivers:
o Develop proper infrastructures that
avoid those problems before they
happen

SOURCE: Telefónica Spain 34


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC091

Smart mobile CapEx allocation based on client and traffic data IMPLEMENTED

DESCRIPTIVE

Situation Approach Impact

• Rationalize current assets and optimize • Analysis of traffic usage, customer • Impact generated
the new deployment in order to reduce value, seasonality, displacements, o CapEx optimization to stretch budget
Capex OMV’s traffic, AFR traffic, roaming efficiency: ~2%
traffic and location of competition’ o Equivalent to 4 Mn. € /year of
• Value drivers antenas additional CapEx
o Infrastructure sharing agreements • Sources of data:
o Assets valuation o Internal: voice and data traffic,
o Deployment Prioritization client value, geolocation, antenna
o CAPEX /year: 200 Mn. € connection…
o External: infoantennas, signal
intensity, network layout

SOURCE: Telefónica Spain 35


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC091

Smart mobile CapEx allocation based on client and traffic data IMPLEMENTED

DESCRIPTIVE

Situation Approach Impact

• Rationalize current assets and optimize • Analysis of traffic usage, customer • Impact generated
the new deployment in order to reduce value, seasonality, displacements, o CapEx optimization to stretch budget
Capex OMV’s traffic, AFR traffic, roaming efficiency: ~2%
traffic and location of competition’ o Equivalent to 4 Mn. € /year of
• Value drivers antenas additional CapEx
o Infrastructure sharing agreements • Sources of data:
o Assets valuation o Internal: voice and data traffic,
o Deployment Prioritization client value, geolocation, antenna
o CAPEX /year: 200 Mn. € connection…
o External: infoantennas, signal
intensity, network layout

SOURCE: Telefónica Spain 36


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC164

Profitability assessment of mobile sites

Situation Approach Impact

• Assessing sites by profitability analyzing • Dashboard with all mobile sites • Value drivers:
different variables mapped and grouped by multiple o Improve profitability on sites with
• Value drivers: variables (technology, frequency, etc.) lower profitability rates
o Improve profitability on sites with
and its profitability information.
lower profitability rates Profitability is defined as the
estimation of revenues for each site
minus cost of operating site
• Sources of data:
o Sites location
o Sites technology
o Sites frequency
o Sites operating costs
o Sites revenues

SOURCE: Telefónica & McKinsey team analysis 37


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

38
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC102

CTO's saturation prevention model based on forecasted network deployment and


commercial campaigns analysis

Description Impact

• CTOs allow for multiples fiber connections (own or Complete anticipation Design anticipation
wholesale) in one specific area, with a limit Description k€ k€
• When a CTO is saturated, no more connections can Development
be created and therefore potential new customers
can’t connect to Telefonica’s network until a new CTO Set-up • Model set-up (320) (320)
is built Cost • Robot set-up (130) (130)

• A predictive model should be created to detect which


Set-up cost (450) Set-up cost (450)
CTO’s are more likely to be saturated in a close period
of time Recurrent

• Input to the model includes data from BI, EDICRE and • Advancement revenues 890 440
Run rate
CELSIX savings

• Once prediction has been made, design and work


orders will be automatically generated Run rate • Model maintenance (30) (30)
maintenance • Robot maintenance (40) (40)

Net recurrent benefit 820 370

Payback Year 1 Year 1

SOURCE: Telefonica Global, McKinsey 39


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC107

Design optimization of network deployment based on environment,


demographic & geographical data data

Situation Approach Impact

• Doing a previous study of the • Leverage and analyzing data of the • Value drivers:
construction environment, we could network construction environment o Cost drivers:
adapt our network design to the specific (soil type, step sheet, existence of  # Breakdowns
conditions of each area pipelines, access permit , historical  I + M visits reduction
• Quality will improve and there will be access permit , citizens' income and o Revenue drivers:
less breakdowns willingness to allow passage, etc.)  Improve CSI
• Value drivers:
o Cost drivers:
 # Breakdowns
 I + M visits reduction
o Revenue drivers:
 Improve CSI

SOURCE: Telefónica & McKinsey team analysis 40


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC136

I+M capacity optimization based on short & medium term events forecasting

Description Impact

• Helping I+M subcontractors size their workforce Description k€


provides a negotiation lever to reduce costs Development

• Model set up (240)


• Prescriptive model that determines for each Set-up
Cost
subcontractor:
Set-up cost (240)
o Optimal total number of technicians
Recurrent
o Optimal number of technicians / profile • Increase technicians productivity by 5%, 2%
(installation, maintenance) and 1% the first 3 years respectively
• Subcontractors negotiation: 50% of
o Optimal technicians geographical distribution Run rate
efficiency gains
savings
• Run-rate for next 3 years respectively over a
• Inputs required (historic subcontractors field force by baseline 91M
typology / region, commercial activities) are available 3470
on Exadata
Run rate • Model maintenance (30)
maintenance

Net recurrent benefit 3440

Payback Year 1

SOURCE: Telefonica Global, McKinsey 41


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

42
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC047

Predictive call centers staffing

Situation Approach Impact

• Predicting call center demand, we could • Predictive staffing model based on: • Value drivers:
estimate how many agents we need and o Demand forecasts o Reduce staffing error by 40%
the time we need them o Network status o Increase CSI
• This way agents productivity will increase o Client historical trends
o Increase agents productivity
and waiting times will reduce, driving to o Seasonal / daily trends
a better CSI • Sources of data:
• Value drivers: o Logstrust

o Reduce staffing error by 40%


o Increase CSI
o Increase agents productivity

SOURCE: McKinsey telcos 43


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC066

Service Operation Center (SOC)

Devices performance Data analysis for devices performance using CEI, 4G time and devices info

WORK IN PROGRESS ANALYSIS

Performance Aceptable Tolerable Inaceptable <60 60-80 80-100 30-59 60-79 80-100

Gamas Tiempo LTE CEI Llamadas sobre parque

Alta Media Baja 14


13
16% 18% 12%
25% 22%
64.60% 330K 57.67% 622K 49.13% 1.26MM 28%
15% 15% 10
16%
18.39% 94K 22.50% 243K 24.57% 629K 71%
60%
56%
17.01% 87K 19.83% 214K 26.30% 674K 975K 68% 67%
59%

511K 1.08MM 2.6MM 18% 17% 14%


High Medium Low High Medium Low
For example: Samsung J1  30% with unacceptable performance Mayor concentración de clientes Terminales de gama baja tienen una Menor demanda en clientes con
(+30K customers) con rangos de LTE menor a 60 en mayor concentración de clientes en terminales de gama alta
terminales de gama baja CEI menor a 60

• Devices homologation teams are working with providers to solve working issues
• Version updates in stock devices and recommendation for updates in sold devices

SOURCE: Telefónica Global 44


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC080

CC calls and TMO reduction via automatic diagnosis/problem solutions based on home
equipment status information

Description Impact

• New deployed home equipment has very advanced Description k€


variable monitoring and remote control tools that can be TMO reduction
Development
used to get information from home route performance 10%
(e.g. ASSIA in HGUs). Set-up Cost • Model set-up 15% (2.420)
20%
• Using that information to develop an advanced analytics
predictive model enables to predict upcoming access faults Recurrent
and better identify diagnostics Scenario 1: 1% reduction (1) Payback Year
340 8
• Predictive modelling allows to take proactive actions like
Run rate savings 600 4
reboot, proactive dispatching or even proactive dropship
to prevent the fault • Calls (~1.9M agent calls) 860 4

• In case clients still call, predictive modeling will guide • I+M activity (~310k field visits)
Scenario 2: 3% reduction
agents with the most probable diagnostic to give a better • Substitutions (~36k replacements)
and faster answer to the client • TMO (~1.9M agent calls) 880 4
Net benefit run
rate 1.130 3
Run rate costs 1.390 3
• ASSIA maintenance
Scenario 3: 5% reduction
• Model maintenance
1.410 3
1.660 2
1.910 2

1 Reduction applies to calls, I+M activity (field work) and substitutions

SOURCE: Telefonica Global, McKinsey 45


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC148

Call center customer journey dynamic optimization to ensure unsatisfied customers are
routed to increase CSI

Situation Approach Impact

• Identifying unsatisfied customers and • Dynamic customer journey routing to • Value drivers:
the causes for their dissatisfaction helps increase CSI of unsatisfied customers, o Churn reduction
us a change our behavior towards them, based on: o Improve CSI
routing them to satisfied cutomer o Historical customer journeys
• Value drivers: followed by client
o Churn reduction o Agent rating & match with client
o Improve CSI o CSI
o Failure history
o Benefits

SOURCE: McKinsey others 46


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC165

Call center calls reduction by information calls id to route them appropriately and/or
provide information proactively

Description Impact

Description k€
• Call centers receive a massive number of calls
Development
from clients nowadays: ~11,3 M agent calls/year.
~30% of the calls received are information
questions Set-up
• Model set-up (410)
Cost
• Using historical client data and products owned
information (e.g.: ASSIA) and developing a
predictive model enables to detect clients more
Recurrent
likely to call 1002 for information after specific
past events • Calls: 20% reduction over ~1.1M structured 910
• Proactively sending the client information before information agent calls (10% of total agent
Run rate calls ~11.3M calls)
he even calls (SMS, email, etc.) or using an savings
automatic process in case the client proceeds with
the call, will reduce significantly information calls
getting to an agent
Run rate
maintenance • Model maintenance (70)

Net recurrent benefit 840

Payback Year 1

SOURCE: Telefonica Global, McKinsey 47


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

48
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC103

WiFi intrusions detection system

Situation Approach Impact

• Detection and notification of intruders • The intruders can be detected by • Value drivers:
on client's WiFi to have a WIFI overall looking for recurrent connection o Revenues:
performance (less saturation and faults) patterns with a lower power than the  CSI
and a better customer confidence which average connection. After that, the
o Costs:
reflects on CSI client will be informed of this
abnormal activity by SMS/Smart WiFi  N1 cost reduction
• Value drivers: so he can try to the "catch" the
o Revenues: intruder o just change the password.
 CSI
• Data Sources:
o Costs:
 N1cost reduction o Hours / connection records
o Power received in WiFi HGU

SOURCE: Telefónica & McKinsey team analysis 49


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC142

Ticket generation reduction by crossing network and EDC data (B2B) to avoid dispatching to
client facilities in a network incidence

Description Impact

• Client equipment platforms (EDC) are traditionally Description k€


not interconnected with network monitoring
Development
systems

• As a result, long technical calls lead to sending Set-up • Data set up (65)
Cost • Model set up (20)
technicians to client home when a core network
incidence happens (and there is nothing they can • Integration set up (200)
do) Set-up cost (285)

• Combine network and EDC databases to create a Recurrent


complete picture of client and network at the
• 100% reduction of massive network
moment of the incidence.
breakdowns:
• Use a descriptive model to find if the incidence Run rate o Calls (IVR)
cause is in EDC systems or if it is a problem derived
savings oI+M visits (45% of calls)
• Run-rate: 230
from the network

Run rate
• Model maintenance (30)
maintenance • Integration maintenance (40)

Net recurrent benefit 160

Payback Year 2

SOURCE: Telefonica Global, McKinsey 50


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC166

Identification of diagnosis mistakes

Situation Approach Impact

• To identify mistakes in the diagnosis • Generate a scorecard that identifies • Value drivers:
phase, a scorecard must be errors in the diagnosis and its root o Costs:
implemented cause identifying at the same time  Diagnosis improvement
inneficiencies in the processes of
• Value drivers:  Deficiencies detection
incident management, low
o Costs: performance centers, contractors…
 Diagnosis improvement Based on incidents history
 Deficiencies detection
• Data Sources:
o Incidents history by type, georgical
area, contractors company…

SOURCE: Telefónica & McKinsey team analysis 51


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC167

Predictive maintenance of companies’ EDCs, identifying possible breakdowns by


(Manufacturer, Firmware, Model of equipment, etc.)

Situation Approach Impact

• Identification and classification of • Based on inventory information, • Value drivers:


possible breakdowns by firmware, alarms, performance data and o Costs:
manufacturer, equipment model… breakdowns, predict failures due to  Breakdowns reduction
firmware BUGs, failures in specific
• Anticipation to breakdowns based on  CEX services reduction
equipment models, predict
historical data degradation and incommunication,
• Value drivers: etc., and take preventive actions
before breakdown takes place
o Costs:
 Breakdown reductions • Data Sources:
 CEX services reduction
o Incidents history by type, georgical
area, equipment, manufacturer…

SOURCE: Telefónica & McKinsey team analysis 52


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC168

WOKU infrastructure growth prediction

Situation Approach Impact

• Prediction of capacity required in WOKU • Analytical model that allows predicting • Value drivers:
platform to assure data can be safely the increase in capacity required in the o Costs:
stored in the platform, the speed does WOCU infrastructure  Platform breakdowns avoided
not slow down and there are no outages
• Data Sources:  Increase productivity
• Value drivers:
o WOKU historic storing trends
o Costs:
 Platform breakdowns avoided
 Increase productivity

SOURCE: Telefónica & McKinsey team analysis 53


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC169

Identification of customers that turn off equipment to avoid I+M dispatching

Situation Approach Impact

• Identification of customers that turn off • Creation of an analytical model that • Value drivers:
their equipment to predict which detects patterns of behavior that
customers are more likely to have a identifies if a client turns off the o Reduction of I+M dispatching
breakdown equipment in order to avoid
unnecessary breakdowns
• Value drivers:
• Data Sources:
o Customer equipment on/off time

SOURCE: Telefónica & McKinsey team analysis 54


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC170

Improve Access network capacity management

Situation Approach Impact

• Prediction of capacity required in • Analytical model to identify capacity • Value drivers:


different platforms to assure data can be management trends both in access o Costs:
safely stored in the platform, the speed network (CdM, Pes, JDS), and in the  Platform breakdowns avoided
does not slow down and there are no EDCs). Data Sources:
 Increase productivity
outages
o Capacity in EDC, capacity in CdM,  TMO reduction
• Value drivers: capacity in PES, capacity in JDS o Revenues:
o Costs:
 Quality improvement
 Platform breakdowns avoided
 Increase productivity
 TMO reduction
o Revenues:
 Quality improvement

SOURCE: Telefónica & McKinsey team analysis 55


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC171

I+M costs reduction by improving detection and management of massive failures

Situation Approach Impact

• B2B customer care services sometimes • Descriptive correlation of client’s data • Avoiding double ticketing during a
lacks network failure information and (home address, location, nature of massive failure by ensuring information
takes some time before realizing a failure, etc.) with core network data reaches customer-care services and is
number of client calls informing about an (current failures and affected nodes, correctly correlated with client
incidence are actually related to a etc.) information may lead to a significantly
massive failure • Sources of data: reduction of dispatched: TBD
• Crossing data analysis of network and o Client equipment • Value drivers:
EDC platforms to avoid sending Cost drivers:
technicians to client facilities when there o Location o

is a core network incidence and resolve o Failure description from client o I+M visits
incidences fasters and remotely without o Network data & alarms • Revenue drivers:
double ticketing or even automatically o Improve CSI
informing the client to avoid calls

SOURCE: Telefónica & McKinsey team analysis 56


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC172

Identification of access nodes failures

Situation Approach Impact

• Prediction of incidents in access nodes • Analytical model that based on alarms • Value drivers:
will help us to actively act and repair it (equipment losses, jitter, etc.) and o Costs:
before customer calls, I+M visits are inventory, can identify incidents in an  Reduced technical centers calls
scheduled… access nodes that affect several clients
 TMO reduction
• Value drivers: • Data Sources:  Reduced I+M visits
o Costs: o Revenues:
o Access nodes historic breakdown
 Reduced technical centers calls types, locations…  CSI imporvement
 TMO reduction
 Reduced I+M visits
o Revenues:
 CSI imporvement

SOURCE: Telefónica & McKinsey team analysis 57


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC175

Decision making optimization with global technical centers dashboard

Situation Approach Impact

• B2B customer care services lacks a global • Descriptive analytical model to present • Impact is generated by optimizing the
dashboard to aggregate and present all active and recent failures day-to-day decision making with real-
real-time data information into a single dashboard to time information
• A global dashboard could help in help decision making process, e.g.: • Results of this analysis can be further
decision making by including failure o Prioritizing dispatching injected into other models
information on: o Understanding problems with
o Affected client equipment specific types of equipment
information mapped by OEM, o Events affecting the clients network
geography, technology, diagnosis and
solution o etc

o Status of each failure


o Pending intervention
o Quality survey results

SOURCE: Telefónica & McKinsey team analysis 58


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC181

Automatic diagnosis of individual Centrex faults for not registered terminals

Situation Approach Impact

• Prediction of incidents in voice services • Automatic resolution or enrichment of • Value drivers:


due to faults in data elements faults with the information of terminal o Costs:
registratioin in the NGN, being able to  Reduced technical centers calls
• Value drivers: perform remote tests for
o Costs:  TMO reduction
autoresolution (login or revision of
 Reduced number of calls wiring)
 Reduced TMO
• Data Sources:
o NGN

SOURCE: Telefónica & McKinsey team analysis 59


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC183

Advanced diagnosis and routing of breakdowns due to massive damage to Ibercom IP and
hosted services

Situation Approach Impact

• Identification of massive incidents to • Unique traceability for service in • Value drivers:


Ibercom IP and hosted services to route Ibercom IP and Ibercom IP Hosted o Costs:
individual customers breakdowns caused (MPLS (PEs) and CdS network (CEs))  Automate calls
by massive incidents
• Data Sources:
• Value drivers:
o Ibercom IP
o Costs:
 Automate calls o Ibercom IP Hosted
o CdS network

SOURCE: Telefónica & McKinsey team analysis 60


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC184

Enrichment of information in Bussiness trunking service faults

Situation Approach Impact

• Enrichment of information in Business • Automated consultations in cases • Value drivers:


trunking service faults of uncommunicated customers o Costs:
with ASBT service, incorporating  TMO reduction
• Value drivers: the status of trunksip. Fault routing
o Costs: decision is based on the result.
 TMO reduction
• Data Sources:
o ASBT
o Trunksip

SOURCE: Telefónica & McKinsey team analysis 61


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC186

Analysis of terminal plants by headquarters

Situation Approach Impact

• Due to the mobility that customers’ IPs • Analysis by MAC and VLAN in the • Value drivers:
may have, an analysis of terminal plants EDC for accounting of terminals in
by headquarters should be done to each headquarters and plant o TBD
better control their position update.
• Value drivers: • Data Sources:
o EDC: terminals by MAC, terminasl
by VLAN
o Headquarters
o Plant update

SOURCE: Telefónica & McKinsey team analysis 62


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC187

Samsung KME integration for massive enlistment of Samsung terminals

Situation Approach Impact

• Samsung KME enables automatic • Advanced analytics model to • Value drivers:


enrollment (similar to Apple's DEP integrate Samsung terminals o Revenues:
program) allowing fast enrollment of  Billing per enrolled device
devices on the MDM platform, ensuring • Data Sources:
revenue (billing per enrolled device) o MDM platform: devices enrolled,
• Value drivers: billing/ device
o Revenues:
 Billing per enrolled device

SOURCE: Telefónica & McKinsey team analysis 63


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

64
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC020

Network maintenance Dez/17 Jan/18

Contractors companies’ performance management


FLM
82% 84%
74%
70% 67%
• Analytical model based on TEF data … 63%
o Preventive maintenance
o Corrective maintenance
o Quality items
Implementation • … and historic data from contractor previous
activities
• Ranking of contractors, from best to worst
• Contractors get bonuses according to
improvements in the ranking NOC CO RAN CAMPO

Main results
• Better response to critical failures • 4% increase FLM tickets under SLA
Main goals • Increase network availability (response in less than 4 hours)
• Better management of contractors • Network availability increased
• Contractors ranking defined

SOURCE: Telefónica Global 65


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC021
Data science
Network maintenance

Analytical model to select the best contractor for each


Capex optimization network work: based in previous works’ performance
and current agreements Performance
and quality
Example measurements
in previous
ICU ICU por Saldo works
Proveedor Score acumulado nueva obra contrato Precio obra
Contratista 6 70 0.50 0.1 260,000 50,000 Best contractor
recom-
Contratista 7 74 0.84 0.1 173,000 50,000 mendation

Contratista 2 87 0.78 0.1 200,000 50,000


Agreements’
Contratista 8 90 0.39 0.1 90,000 50,000 restrictions and
Not selected due capacity of
Contratista 1 92 3 0.1 293,000 50,000 to lack of “saldo” contractors
or lack of
Contratista 3 94 0.5 0.1 39,000 50,000 capacity to make
the works
Contratista 5 99 0.64 0.1 172,000 50,000

Contratista 4 102 1.00 0.1 150,000 50,000

SOURCE: Telefónica Global 66


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC046

Improve contactability by subcontractors

Situation Approach Impact

• Contactability should improve to • Determine the best time to contact • Value drivers:
increase the number of adds and reduce each client in order to set an o Cost drivers:
the number of churns appointment for an installation or  Increase Baf adds/ month: 20%
• Value drivers: breakdown service day
o Revenues: • Sources of data:  Increase Cu-Fi Migration /month:
o Internal: number of calls, TMO, 20% day
 Baf adds/ month: 92500
data connections, Tv audiences  Improve CSI
 Cu-Fi Migration /month: 50000
/housevoice o Costs
 CSI  Reduction of Baf churn / month:
o Costs: 0,002%
 Baf churn / month: 26000  Reduction of Cu-Fi Migration
 Cu-Fi Migration Churn/ month: Churn/ month: 0,002%
5500 • Impact:
o 0.6 Mn.€ / year

SOURCE: Telefonica Spain 67


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC048
Cliente Planta externa Técnica Ventas
Customer service delivery TI Masiva Instalación Otros
Field
force

Field force – use case #3


• Analysis of why a high number of scheduled field visits end up without solving customer’s problem
o Main cause of failed field visit: customer is not found at his house ~60%
o Second cause: network or external-plants-related issues ~15%
2017 2018
Motivo Grupo T1, percent T2, percent T3, percent T4, percent T1, percent
Casa cerrada s/cont/cli Cliente 23,76 18,51 19,04 17,88 17,79
Cliente posterga instalación … Cliente 14,26 16,44 17,07 16,31 16,26
Cliente Cliente 8,62 8,73 12,27 16,84 15,69
Cliente no cumplió cita Cliente 7,72 9,55 9,39 9,42 9,30
Pendiente cita futura Ventas 10,60 10,60 7,05 1,54 1,41
Problema planta externa Planta externa 17,19 12,28 0,93 0.01 0.03
Sistémica TI 1,17 3,45 3,13 1,65 1,65
Pendiente cable externo Planta externa 0,76 3,89 4,30 3,83
Pendiente despacho de lín … Planta externa 1,75 1,61 2,63 3,14 4,07
Pendiente cable local Planta externa 0,90 4,00 3,60 3,73
Pendiente ba local Planta externa 0,64 3,15 3,39 3,69
Pendiente oficina central Planta externa 2,02 1,65 2,04 1,15 1,24
Planta externa asignada con … Técnica 0,53 0,29 1,04 4,68 4,32
Cable reparado Planta externa 0,39 1,69 1,01 2,06 2,48
Pendiente ba externo Planta externa 0,31 1,62 2,08 2,03
Cliente no cumple cita. Sin … Cliente 1,87 0,86 0,80 0,78 0,52
Datos erróneos planta exter … Planta externa 0,64 0,48 0,86 0,54 0,34
Pendiente falla masiva plant … Masiva 0,37 0,49 0,59 0,64 0,53
Servicio no sincroniza por p … Técnica 0,58 0,44 0,65 0,40 0,32
Pendiente recurso adsl Planta externa 0,57 0,41 0,51 0,55 0,68
Se llama al número de conta … Cliente 0,41 0,55 0,51 0,52 0,38
Planta externa ok Planta externa 0,37 0,41 0,24 0,59 0,70
Falta equipo fttx (ont, hiwi, l … Técnica 0 0,16 0,42 0,82 1,34
Nuevo servicio de banda an … Técnica 0,37 0,32 0,33 0,38 0,38
Pta ext ok defecto hacia clie … Planta externa 0,24 0,29 0,38 0,41 0,47

• Actions taken
o Several initiatives related to appointment schedule have been put in place (SMSs, appointment reminders, tracking of the technician by the customer (in Argentina, this project is called Glympse), etc.

SOURCE: Telefónica Global 68


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC049

Smart management of field force including optimization of dispatching and technicians


journeys

Description Impact

• Helping I+M subcontractors optimize the Description k€


dispatching of their workforce provides a
Development
negotiation lever to reduce costs
Set-up • Model set up (240)
• Prescriptive model that optimizes the assignment • Additional cost could be incurred to develop a
Cost
of technicians to work orders based on dispatching tool
Set-up cost (240)
o Technician profile
Recurrent
o Service type • 5-10% journey time reduction 1210- 2410
• 20% automation of dispatchers assigning technicians
o Distance to customer property
time
• Inputs required are available on different data • Subcontractors negotiation: 50% of efficiency gains
Run rate
sources • Run-rate for next 3 years respectively over a baseline
savings
91M
• Additional impact can be identified if further study

Run rate • Model maintenance (30)


maintenance
Net recurrent benefit 1180 - 2380
Payback Year 1

SOURCE: Telefonica Global, McKinsey 69


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC051

Customer service delivery Field force


Field force– Use case #5

Review of the point scores used to pay to the local-loop companies

• In Brazil, we are using these information in the negotiation with the local-loop companies for establishing the new scales of their local-
loop activities to obtain the related cost-efficiency gains

70
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC052

Predicting customers with high probability of orders cancellation PRELIMINARY

PREDICTIVE

Situation Approach Impact

• Customers with high probability of • Machine Learning model to predict • Value drivers:
cancelling orders can be predicted with when and why installations orders will o Cost drivers:
machine learning be cancelled (POC). Main causes are:
o Smart management of the field force
• Knowing which customers have high o Main causes for cancellations
• Impact:
cancellation probability can help us o Installation type
managing the field force work o Location
o Technical data
o Time lapse between the request
from customers and execution
• Sources of data:
o Internal: customer segmentation,
customer historic behavior,
installation types
o External: location

SOURCE: Telefonica 71
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC053

Training of technicians
PRELIMINARY

Situation Approach Impact

• Training technicians on their weak points • Analytical model to assess capability • Value drivers:
will make them multi-task and ease as building requirements based on:
dispatching o Tracking of productivity and
o Current and future network capabilities by technician
• Telefonica is already training their N1 deployment/maintenance
technicians in Chile capability needs o Optimization of capability building
programs
• Value drivers: o Current technicians capabilities
o Reduction of I+M OpEx as better
o Tracking of productivity and • Sources of data: qualified technicians will be available
capabilities by technician o Internal: technicians skills, • Impact:
productivity, CSI
o Optimization of capability building o Improve CSI
programs
o Reduction of I+M OpEx as better
qualified technicians will be available

SOURCE: Telefonica Global 72


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC062

Fault Finding Maps: identification of blackspots in the last-mile network PRELIMINARY

PRESCRIPTIVE

Situation Approach Impact

• If we discover network elements • Analyzing historical data of line • Value drivers:


concentrating the highest number of stability (xDSL connection) along with
failures we could repair/ substitute them line faults, we discover the network o Massive breakdowns reduction
proactively before customers notice and elements (cabinets, splices and boxes) o Technical centers calls reductions
complain concentrating the highest number Of
failures. o Reduction of I+M visits
• It would help us prioritize repairs order • Impact:
of network elements • Sources of data:
o Internal: network elements, o Improve CSI
• Value drivers: historical breakdowns, #1002 calls
o Reduction of call center opex
o Massive breakdowns reduction due to network elements, # I+M
visits due to network elements, o Reduction of I+M cost
o Technical centers calls reductions frequency of breakdowns
o Capex reduction
o Reduction of I+M visits

SOURCE: Telefonica Global 73


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC063

Customer service delivery CPEs

CPEs – use case #9 WORK IN PROGRESS

• Fraud detection in the material handled to the local-loop companies (Chile)


o Chile has developed a project to correlate the material handled to the technicians for their field visits and the material really used by them either when they perform a installation
or a repair with equipment's replacement

Descripción
• El presente gráfico mide a nivel de reclamo
• Los equipos pendientes corresponden a la diferencia de las
cantidades de material registradas por el técnico (TOA) y a las
devueltas por el mismo en aquellas claves de cierre que les
corresponde devolución de equipo (las devoluciones se miden a
nivel de reclamo ya que solo hay registro de la cantidad devuelta
pero no del material devuelto por reclamo)
• El porcentaje de cumplimiento de devolución muestra qué tanto
cumplen las empresas contratistas respecto a los equipos que
han devuelto versus a lo que deberían o lo que les
correspondería devolver. Entre mayor es su porcentaje de
cumplimiento de devolución están en una mejor posición. El
porcentaje de cumplimiento de devolución es una forma de
comparar a todas las empresas contratistas en una misma escala
Interacción
• El gráfico de barras total tiene la opción de ver detalle
• El gráfico de barras por empresa contratista tiene la opción de
ver detalle
La selección del gráfico de porcentaje de cumplimiento de
devolución resalta al de cantidades por empresa contratista

SOURCE: Telefónica Global 74


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC094

CSI improvement by substituting technical services by stock replacement PRELIMINARY

when technicians are overloaded PRESCRIPTIVE

Situation Approach Impact

• In order to decrease average installations • Algorithm to offer the best available • Value drivers:
time, number of cancelled services and spaces and the most appropriate o Revenues:
improve customer satisfaction, we technician according to their skills,  Increase Baf adds/ month: 1 day
should prioritize services by product profile, performance history and  Increase Cu-Fi Migration /month: 1
value, technician’ expertise and house type day
instalaltion complexity • Sources of data:  Improve CSI
• Value drivers: o Internal: product value, client value, o Costs:
o Revenues: client history…  Reduction of Baf churn / month:
 Baf adds/ month: 92500 o External: technician profile, service 0,01%
 Cu-Fi Migration /month: 50000 orders history, calendarization tool,  Reduction of Cu-Fi Migration
 CSI cadastre, VISORD… Churn/ month: 0,01%
o Costs: • Impact:
 Baf churn / month: 26000
 Cu-Fi Migration Churn/ month:
o 2,9 Mn. € / year
5500

SOURCE: Telefonica Spain 75


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC134

CSI improvement by substituting technical services by stock replacement


when technicians are overloaded

Situation Approach Impact

• Network deployment works, mainly in • Dynamic descriptive advanced analytics • Increase in revenue advancement in B2B
the access network is typically prioritized model to prioritize works based on clients (GGCC): TBD
based on customer value, or first-in first- actual business impact adding data • Reduction of penalties applied: TBD
out basis from: • Optimization of supply and works: TBD
• Prioritizing work based on actual o Supply & SCM costs
business value of each of the works o Provision works
could: o Customer value
o Increase revenue advancement o Probability of claims based on
o Reduce penalties actual contract & history of claims
o Other work orders for each client

SOURCE: Telefonica España 76


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC134

CSI improvement by substituting technical services by stock replacement


when technicians are overloaded

Situation Approach Impact

• Identifying customers preferences • Sources of data: • Value drivers:


(waiting for maintenance vs going to a o Internal: Shop stock analysis, o Cost drivers:
shop to pick up new shock) we could breakdown classification by type/  I+M visits
improve customer satisfaction complexity, localization, client  Stock Cost
• When I+M are over loaded, maintenance profile, client behaviour (willingness  Logistic
services could be prioritize according to to wait) o Revenue drivers:
customer profile and behaviour  Improve CSI
• Value drivers:
o Cost drivers:
 I+M visits
 Stock Cost
 Logistic
o Revenue drivers:
 Improve CSI

SOURCE: Telefónica & McKinsey team analysis 77


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC137

Dynamic stocking for network equipment and spares (analytics SCM)


based on forecasts on deployment / decommissioning

Description Impact

• Urgent journeys of stock suppose ~1 – 1.5 Mn.€ a Description k€


year
Development
• Optimization of stock distribution and stock • Model set up (280)
Set-up
holding to reduce logistic costs • Additional cost could be incurred to develop a stock
Cost
management system
• Prescriptive model that determines real time, Net recurrent benefit (280)
geographical stock planning
Impact on • 7% stock cost reduction in the next 3 years 2990
• Inputs required to this model are available on balance sheet
different sources (TeleSap, Edicre, CELSIC..)

Recurrent
• 5% reduction in urgent journeys cost 170
• Run-rate of 3 years:
Run rate
savings

Run rate • Model maintenance (30)


maintenance
Net recurrent benefit 140
Payback Year 2

SOURCE: Telefonica Global, McKinsey 78


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC138

Improvement in client equipment traceability

Situation Approach Impact

• Improvement of customer devices • Sources of data: • Value drivers:


monitoring to control breakdowns  Customer theoretical and real o Cost drivers:
• Value drivers: plant (the real one can be  Equipment losses reduction
o Cost drivers: obtained from devices ASSIA)  Fraud reduction
 Equipment losses  I+M performance sheets and  Logistic optimization
 Fraud records  Repair control optimization
 Logistic optimization  Information of customer park o Revenue drivers:
 Repair control optimization renewal  Improve CSI
o Revenue drivers:  Provisions and actions carried out
 Improve CSI  Warehouse inventory

SOURCE: Telefónica & McKinsey team analysis 79


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC139

Optimization of I+M workforce by assigning on-the-go tasks based on geolocalization and


capabilities

Situation Approach Impact

• Typically a significant amount of low • Optimization of the use of I+M • Value drivers:
value added tasks are left un-assigned resources by asigning on-the-go tasks
due to high costs related to dispatching a to technitians by leveraging following o Cost drivers:
technician to execute just that task. This info:  Reduction of I+M OpEx by assigning
creates a backlog of unattended works tasks on a geolocation basis
o Geolocalization and/or access to
• Assigning on-the-go tasks to qualified specific sites o Increase in network avail
technicians based on their capabilities a
geolocalization can lead to executing a o Technical qualifications
larger amount of maintenance tasks and o List of unattended, low priority,
thus increase avail on network elements tasks

SOURCE: Telefónica & McKinsey team analysis 80


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC147

Correlation of I+M work sheets vs work orders to ensure work in done


according to orders

Situation Approach Impact

• Comparing and correlating work sheets • Analytical model based on external • Value drivers:
vs work orders we could identify what factors, KPI parameters, business and o Cost drivers:
has been done differently from agent technical criteria to identify:  Fraud prevention
recommendations and why o Contractors' fraud  Work control
• This way we could identify if agents o Field force mistakes... corruption or  Equipment control
made a diagnosis mistake, if technicians simple mistakes o Revenue drivers:
make a mistake due to vague skills or o Agents diagnostic mistakes  CSI improvement
corruption • Sources of data:
• Value drivers: o Work orders
o Cost drivers: o Work order assignment
 Fraud prevention o Work bulletins
 Work control
 Equipment control
o Revenue drivers:
 CSI improvement

SOURCE: Telefónica & McKinsey team analysis 81


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC150

Advanced service provisions

Situation Approach Impact

• Check if there are recurring incidents in • Analyzing historical data of circuit • Value drivers:
any circuit part/ factor, either in typology stability (xDSL connection) along with o Revenue drivers:
or in any responsible area. If there are, cirquit faults, we discover the parts/  Advance provision revenues
the milestone of the order should be factors (concentrating the highest
changed to incorporate this activity, in number of failures.
order to try to start earlier and reduce
provision time • Sources of data:
• Value drivers: o Frequency of breakdowns/ type
o Revenue drivers: o Frequency of breakdowns/ area
 Advance provision revenues o Order milestone

SOURCE: Telefónica & McKinsey team analysis 82


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC151

Advance network fall-back

Situation Approach Impact

• With the statistics of success of • Historical analysis of trends and • Value drivers:
migrations to fiber of the different quantity of migrations copper- fiber o Cost drivers:
services, we should prioritize the for different services  Equipment shutdown
migrations of those services that • Sources of data: o Revenue drivers:
statistically end up migrating before and o Copper-fibber migration trend/  Advance migration revenues
in greater number. This is intended to service type
accelerate the shutdown of equipment, o Number of copper-fiber service
so that the average release path migrations / service type
accelerates the first few weeks.
• Value drivers:
o Cost drivers:
 Equipment shutdown
o Revenue drivers:
 Advance migration revenues

SOURCE: Telefónica & McKinsey team analysis 83


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC154

Using Big Data in the technical (EDICRE) and accounting (TELESAP) inventories databases to
apply it to the audit process.

Situation Approach Impact

• Using Big Data in the technical (EDICRE) • Real-time monitoring of the current • Value drivers:
and accounting (TELESAP) inventories state of inventory o Cost drivers:
databases would minimize
• Creating alarms for works with errors  Process automation
inconsistencies between inventories and
automate processes, reducing the • Sources of data:
number of manual interventions. o EDICRE
• Value drivers: o TELESAP
o Cost drivers:
 Process automation

SOURCE: Telefónica & McKinsey team analysis


84
Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

85
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC115

IT HW/SW opex optimization via an analytics based asset management

Description Impact

• Incidents and faults affect different types of clients Description k€


and have different economic impact for the
Development
company
Set-up • Model set-up (180)
• Gathering historical data about previous faults
Cost
(e.g. client typology, incident typology, broken
SLA) and building a predictive algorithm to
estimate affected clients and level of impact
enables to identify critical incidences Recurrent
• Changing incidence resolution prioritization • Network unavailability time: 3.5% reduction over total 2460
process to minimize economical impact has unavailable time of ~108M min (cost of min is 0.64€,
significant reductions on penalties and revenue Run rate includes lost of revenues and penalties)
losses savings

Run rate • Model maintenance (20)


maintenance

Net recurrent benefit 2440


Payback Year 1

SOURCE: Telefonica Spain 86


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC115

IT HW/SW opex optimization via an analytics based asset management

Description Impact

• IT hardware & software capacity utilization has Description k€


been measured and in some cases is extremely
Development
low (~5%)
Set-up • Model set-up (240)
• Gathering data based on capacity scanning tools
Cost
to assess utilization of HW and SW at a per-unit
level, and running an optimization algorithm to
reallocate capacity, enables to increase asset
utilization Recurrent
• Maximizing HW & SW asset usage reduces HW & • SW licenses: 5-10p.p utilization improv. leads to 9-17% 1430 - 2600
SW supply costs and identifies HW candidates to reduction in SW licenses costs (~ 15.7M €)
decommissioning which may lead to cash inflows Run rate • HW purchases: 5-10p.p utilization improv. leads to 9-
and decrease in energy consumptions savings 17% reduction in HW purchases (~ 19.6M €)
1780 - 3270

Run rate • Model maintenance (30)


maintenance

Net recurrent benefit 3180- 5850


Payback Year 1

SOURCE: Telefonica Spain 87


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC036

O+M resources optimization based on an outage duration predictive model


PREDICTIVE

Situation Approach Impact

• Currently tickets are prioritized based in the • Gather historical data about past outages
actual outage duration, with no information and the time needed to fix them.
of how long will it be • Only basic historical outage information
• By following this method critical incident can be used to properly model the status
are not prioritized until the outage has been of uncertainty that happens when a
going for a considerable amount of time outage happens (e.g.: what has
• By knowing beforehand the duration, long happened?).
expected running tickets can be prioritized, • Some examples of data to be used: • Reduction of revenue loss due to customer
increasing customer satisfaction churn by TBD%
o Location of the outage and area
affected (or disconnected) • Faster and more precise answer to clients
affected by the same outage
o Network variables at the moment of
the outage
o Force field status
• Use a predictive model to get a estimation
of the outage duration / time needed to
recover from the fault

SOURCE: Telefonica Global Resources 88


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC114

Network OPEX optimization using a predictive maintenance and root failure cause analysis

Situation Approach Impact

• Massive network faults or malfunctions • Use historical information to develop • Decrease of massive network failures by
have a critical impact in customer an advanced analytics predictive model preventing them will mainly decrease
experience, service and repair costs to predict an upcoming massive fault O+M operating costs
• Some of those faults would have been • All network elements will be evaluated o O+M OPEX: 17.5 M € workforce cost
easily prevented if they were known regularly, and for each one the  “EEBB”: 12.6 M €
beforehand, with a much lower cost (in probability of an upcoming massive
terms of economics and in terms of fault in a short period of time (TBD)  “Planta Interna”: 5.9 M €
client experience) will be evaluated • By preventing critical faults, service
• By digging at network data gathered by • Those elements with a high probability quality and client satisfaction will also
NOC/SOC, is easy to find strange patterns of fault will be reviewed in detail by increase
long before the massive fault happens maintenance teams, trying to locate • Life cycle of network elements will also
root cause of the fault and eventually increase
preventing it from happening

SOURCE: Telefonica Spain 89


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC156

Network opex optimization using a predictive maintenance model based on service


degradation

Situation Approach Impact

• Network faults are usually the • Define different performances KPI for • Reduction in number of fault cause by
consequence of a prolonged element different network elements small malfunctions by TBD %
malfunction over time
• Monitor KPI evolution over time to • Overall performance increase of network
• Those malfunctions do not have a direct detect decreasing patterns trends assets
impact over the network quality, but
they are slowly damaging other network • Perform a root-cause analysis over
elements decreasing trends to understand
causes of those decrements
• Fixing those malfunctions do not have an
immediate impact over network quality,
rather than a very high impact over time

SOURCE: Telefonica Spain 90


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC001

Proactive maintenance to improve network availability

Situation Approach Impact

• Incidents and faults affect significantly to • Optimization based on: • Increase revenues due to improvement
sites holding critical service o Operative data: in Network Availability in sites that hold
 Data traffic
critical services
• Failure on sites that hold critical service
can draw a huge negative impact over  Concentrations • Improvement in Customer Satisfaction
clients if they are not resolved fast  Recharge sites locations
enough o Business data: Example Colombia:
• Optimization of the maintenance tasks  Benefits (ARPU – Costs) • Total ARPU increased
impact on client satisfaction levels  Companies categorization o Jan 2017 33%
• Combine both aspects to develop an o Feb 201719%
analytical model o Mar 2017 19%
• Prioritize maintenance tasks for critical
sites to improve network availability

ARPU = Average Revenue per User

91
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC002

Proactive maintenance: Network availability supervision dashboard

Situation Approach Impact

• Unavailability of network data results in: • Develop a centralized dashboard • Improve management time
Slow response to incidents including global information of
o • Resolution time reduction (operations
network availability
o Increased management time area has available online information)
o Increased non-value added tasks • Monitor centrally global KPIs

• A centralized dashboard provides easy • As a further step, develop an analytical


and quick access to network availability model “on-top” of the dashboard for
information availability management
o Prioritize network availability on VIP
sites

SOURCE: Telefonica Spain 92


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC009

Increase voice quality service by increasing Minutes Per Drop

Situation Approach Impact

• Drops in mobile voice services decrease • MPD KPI tracking through real time • Increase quality of mobile voice service
quality and client satisfaction mobile voice service monitoring • Increase customer satisfaction
• Real-time control of drops in mobile • Detect lowest MPD services
voice services allows to keep track of
Minutes per Drops • Prioritize actions to increase service
quality of lowest performance.
• By detecting lowest MPD services, Improve ISC of mobile voice service.
relevant actions can be taken to increase
customer satisfaction

93
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC030

Prioritize incident resolution based on clients’ revenue

Situation Approach Impact

• Incidents and faults affect relevant clients • Gather historical data to asses client • Increase revenue by prioritizing highest
revenue generation revenue clients
• Increased resolution time for most
valuable clients can draw a huge negative • Identify clients with highest revenue • Increase on relevant clients engagement
impact over client satisfaction generation
• Prioritizing significant clients increases • Optimization algorithm to prioritize
client satisfaction and reduces revenue incident resolution based on client
losses revenue generation

94
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC031

Real time detection of anomalies in radio links

Situation Approach Impact

• Poor performance of radio links results • Monitor system variables to analyze in • Improved service quality
in low service quality real time throughput behavior • Improved customer satisfaction
• Real time monitoring of system variables • A throughput drop trigger detects
allows early identification of issues when bandwidth on a radio link drops
• Detection and proactive review of the significantly below a dynamic
issues improves service quality threshold
• An alert system will enable a proactive
review of the issue

Link with throughput drop

95
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC032

Network OPEX optimization through outages modelling

Situation Approach Impact

• Outages lead to a significant decrease on • Develop a descriptive model of • Reduction on network OPEX:
customer satisfaction outages (FO and Cu) to understand o Reduction on agent calls
• Lack of information on historical outages trends, patterns and behaviors o Reduction on sustitutions
increases the complexity to detect future • The model is based on the following • Increased customer satisfaction
possible outages criteria:
• Understanding outages patterns and o Pext: Equipment
behaviors will enable to better predict o Pint: Equipment
and prevent future possible outages o CPE’s
o Geographical
o Client profile
o Weather
o Service
• Outages models will enable to
anticipate future outages and take
preventive actions

96
ADVANCED ANALYTICS IN NETWORK OPERATIONS. UC035

Identification of assets and network elements with most incidents

Situation Approach Impact

• Incidents and faults affect assets • Gather data to identify network • Increase network quality
according to asset typology and elements with most incidents from: • Reduce number of incidents
utilization o UDO (CEDEX IP) • Increase asset lifecycle
• Assets with higher failure rates tend to o Network where the assets operates
have lower lifecycles o Incidents
• Understanding “normal” failure rates • Detect assets or elements with a
and a prioritization of maintenance higher failure rate than the average.
procedures could results in higher asset Get conclusions from assets
lifecycle performance, operation and other
relevant insights.
• Based on the results, execute
proactive maintenance procedures to
increase assets life cycle.

97
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC064

Increase customer experience by monitoring mobile network performance

Situation Approach Impact

• Poor mobile network quality leads to low • Monitoring TCP retransmissions • Improved customer experience
customer satisfaction enables to supervise the mobile
• The lack of regular TCP information network quality given to clients
• Detect quality issues on specific Example of TCP retransmissions monitoring:
impedes the identification of clients
suffering low customer experience mobile networks • Consequence increase 10 p.p in CSI
• Monitoring TCP retransmissions enables • Take faster actions to mitigate the
to detect clients experiencing low mobile problem (e.g.) :
network quality and improve their o Performance analysis
customer experience o Monitoring included in their normal
procedures
o Worst performance detection
o TILT correction

98
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC065

Fraud detection and prevention through analysis of client data (Tuenti fraud)

Situation Approach Impact

• Fraudulent users impact negatively on • Analyze and monitor client data • Increase in revenues
the company results consumption to detect significant
traffic consumption with no associated
• The lack of regular tracking of traffic billing Example Tuenti Fraud:
consumption impedes the identification
of fraudulent users • Change security team resolution
prioritization process to focus on
• Early detection and prevention of potential fraudulent clients
fraudulent clients results in revenue
increases • Prevent potential fraudulent users

• Incident: Aprox, 40 users were


responsible for 250 Gb per day through
VPN
• Consequence: DNS configuration solved

99
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC067

Fraud detection and prevention through analysis of abnormal traffic consumption patterns

Situation Approach Impact

• Fraudulent users impact negatively on • Analyze and identify abnormal or • Increase in revenues
the company results unusual traffic consumption patterns
to detect potentially fraudulent users
• The lack of regular tracking of traffic
consumption impedes the identification • Change security team resolution
of fraudulent users prioritization process to focus on
potential fraudulent clients
• Early detection and prevention of
fraudulent clients results in revenue • Prevent potential fraudulent users
increases

100
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC072

IPTV service quality monitoring to identify faulty network elements

Situation Approach Impact

• Faults in network elements result in poor • Monitor service quality on a Live IPTV • Increase customer satisfaction
quality IPTV service and low quality (e.g. detect frequent “pixel episodes”) • Increase asset lifecycle
satisfaction • Identify network elements that need
to be replaced • Reduce CC OPEX
• Monitoring quality IPTV service and
identifying elements to be replaced • Replacement of faulty network
improves service quality, increases elements will result in:
customer satisfaction and increases o Overall quality improvement
asset lifecycle o Better CSI
o Longer asset lifecycle (asset
replacement based on performance
rather than age)

101
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC073

IPTV proactive server maintenance detecting available content that cannot be played

Situation Approach Impact

• Failures in the servers result in IPTV • Monitor and detect available content • Increase customer satisfaction
content that cannot be played that cannot be played by consumers • Reduce CC OPEX
• Identify failures in the servers emitting
• The lack of regular monitoring of
that content. In case no content is
displayed content increases the
being played, massive outage may be
complexity of identifying server failures
the cause of the issue
• Early detection of server failure reduces • Take proactive maintenance action on
resolution time and increases customer the affected servers
satisfaction

102
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC079

Using a Telegram BOT chat information to get insight about IPTV network status

Situation Approach Impact

• Failures in video network result in • Leverage text analytics over telegram • Increase customer satisfaction
content that cannot be played and BOT conversation to get insights on • Reduce CC OPEX
decreases customer satisfaction global video service delivery
• Detect massive faults that were not
• The lack of additional non-traditional
detected by monitoring more
sources to detect faults increases the
traditional sources of information
complexity of identifying IPTV network
failures • Take specific actions to solve detected
faults
• Early detection of failures reduces
resolution time and increases customer
satisfaction

103
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC089

OPEX optimization through analytics based network asset management

Situation Approach Impact

The leading TSOs and DSOs are applying • Creating an integrated large data • Reduce OPEX
advanced analytics to asset management in platform with models for all the users • Reduce CAPEX
order to: to come before the failures and the
• Maintenance of operation (inspection vs. consequences of the failure
preventive vs. corrective) Example:
• Reduce capex (extend the life cycle of the
asset) 2025 OPEX 2025 CAPEX 2025 TOTEX
• Improve quality (reduce the number of US$M US$M US$M
power interruptions and increase the
availability of assets) 2.92 14.12 17.0

Staff Reputa-
Security Reliability motivation tion

TSO = Transmission System Operators


DSO = Distribution System Operators

104
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC101

Social network monitoring to detect faults unnoticed by the systems or unreported by


customers

Situation Approach Impact

• Failures in systems network result in • Monitor social network activity by • Increase customer satisfaction
increased CC calls and decreases using relevant keywords • Reduce CC OPEX
customer satisfaction • Detect faults unnoticed by the system
or unreported by customers
• The lack of additional non-traditional
sources such as social network scanning • Take proactive action to resolve issues,
increases the complexity of identifying reduce resolution time by using an
network failures additional source of detection

• Scanning social networks reduces system


failures detection and resolution time

105
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC106

Unicast video definition dynamic changes based on demand analytics

Situation Approach Impact

• Poor quality on video service leads to • Develop an advanced analytics model • Increase customer satisfaction
low customer satisfaction with unicast video usage data to
predict peaks in bandwidth
• The lack of adaptation of video definition
consumption
depending on demand volumes
decreases quality of the service • Take actions accordingly (e.g.):
• Adjusting video definition depending on o Lower video definition to avoid
demand consumption will increase pixelated images
quality of the service and therefore o Turn off multicast channel that are
increase customer satisfaction not being watched
• Improving video quality increases
customer satisfaction

106
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC113

Systems coordination for a fast network restoration after a failure

Situation Approach Impact

• Long resolution of network faults results • Develop a machine learning algorithm • Reduce OPEX:
in high OPEX costs and low customer that takes the best action possible o Reduce field work
satisfaction after a outage to restore network
systems as fast as possible o Reduce time of field work
• Non coordinated sources and network
• Coordinate actions and data from
elements lead to longer fault resolution
multiple sources to increase resolution
times
efficiency and reduce resolution time
• The lack of information of previous • In case that no human intervention is
procedures taken to resolve previous need to restore the failure the system
issues increases resolution inefficiency will act automatically without human
supervision.
• Learning from resolution experience and
coordination systems will reduce OPEX
and increase customer satisfaction

107
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC127

Software license usage monitoring

Situation Approach Impact

• Low current utilization of software • Gather real-time data based on • Reduce software maintenance cost
licenses, in some cases extremely low capacity scanning tools to assess
(~5%) license utilization at a per-unit level
• Typically allocation of software licensing • Optimization algorithm to reallocate
is only based on equipment and maximize license usage
requirements, not leveraging the
utilization of current licenses • Undergo non-used licenses

• Higher levels of software utilization can • Detect usage of non-licensed software


be reached by smart reallocation and that could potentially lead to significant
undergoing unused licenses penalties

108
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC157

Post mortem advanced analytics root failure cause analysis

Situation Approach Impact

• Failures in systems network increases CC • Analyze of historical faults to study • Reduce network OPEX
calls and decreases customer satisfaction main root failure causes
• The lack of root cause analysis on • Develop advanced model to help
failures: identify potential root causes
o Prevents the identification of the root • Develop specific strategies to address
cause and therefore the correct root causes will reduce network OPEX
resolution of the issue
• Solving directly root causes also:
o Limits learning for future failures
o Prevents from future failures
• Root cause analysis increases efficiency
of resolution processes and reduces o Avoids re-work of superficial issues
operating costs of failure repair

109
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC162

Real time detection of anomalies in network elements

Situation Approach Impact

• Poor performance of network results in • Monitor system variables to analyze in • Improved service quality
low service quality and a decrease on real time throughput behavior
customer satisfaction • Improved customer satisfaction
• A throughput drop trigger detects
• Real time monitoring of system variables when the value of multiple network
allows early identification of issues KPIs drops significantly below a
dynamic threshold
• Detection and proactive review of the
issues improves service quality • An alert system will enable a proactive
review of the issue

110
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC013/UC019

Develop an analytical model to run root cause analysis

Situation Approach Impact

• Faults increase CC calls and decrease • Analyze of historical faults to study • Reduce network OPEX
customer satisfaction main root failure causes
• The lack of root cause analysis on • Develop advanced model to help
failures: identify potential root causes
o Prevents the identification of the root • Develop specific strategies to address
cause and therefore the correct root causes will reduce network OPEX
resolution of the issue
• Solving directly root causes also:
o Limits learning for future failures
o Prevents from future failures
• Root cause analysis increases efficiency
of resolution processes and reduces o Avoids re-work of superficial issues
operating costs of failure repair

SOURCE: Telefonica Global 111


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC028

Network diagnostic: Amount, time, impact

Network Diagnostics improve Network performance regarding amount, time, and impact of outages, by
applying Machine Learning and reducing manual effort

SOURCE: Telefonica Global 112


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC029

Enabling optimal decisions near real time

Enabling Optimal Decisions Near Real-Time


• Network-OPS has to solve
critical incidents within 4
hours
• Near real-time
infrastructure to deliver
fast decision support
• Machine learning
algorithms for decision
support are key value add
of business & Market
Intelligence

SOURCE: Telefonica Global 113


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC033

Prediction of recurrent outages

• Recurring outages are predicted by a combination of text


Prediction of Recurring Outages mining features from incident work logs and other incident
features

SOURCE: Telefonica Global 114


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC037

Traffic management real-time analytics

Applying analytical techniques to traffic data to detect anomalies, bottle necks, load
forecasts and “What-if” scenarios
Traffic Management

SOURCE: Telefonica Global 115


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC044

Call center: Analysis based on IT service Management resolution code and description

Analysis based on IT service management resolution code & description

• Although data analysis is disaggregated by OB, regional


distribution, technology and service dimension, some
outcomes are common to all TEF OBs:
o Only 1 out 4 call center contact is resolved by
recommending a field visit
o ~25% of the technical contacts related to high speed
internet services are WiFi issues related
 Roughly, 40% of the Wifi issues are WiFi network
Name and password related
o A high percentage of the second customer call are
due to a scheduled field visit that never happened

Actions taken
• SMART WiFi project to facilitate our customers the management of their own WiFi

SOURCE: Telefonica Global 116


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC017/UC018/UC054/UC077/UC088/UC090/UC128/UC131

NOC SOC summarized ideas (McKinsey + Telefonica)

Use case Description


1 RNA (Neutral Network system) Analyse anomalies and netowork elements and develop an alarm system to
detect them

2 Network optimization, Predictive maintenance, Leverage machine learning algorithms to identify signals and make predictions
Network outages and services degradations in advance of outages and network degradations

3 E-E analytics led diagnosis End to end supervision to ensure right decision making and efficiency
maximization

5 Reactive maintenance Automate resolution protocols and actions for main problems

5 Process mining to detect anomalies and Apply specialized data mining algorithms to data in order to detect anomalies,
bottlenecks trends and bottlenecks

7 Network quality optimization by using SON Improve network quality through advanced technologies e.g. self-organizing
network (SON)

8 Identification of a pattern of common breakdowns Analyze historical breakdowns, diagnosis and commands

9 Analysis of breakdowns per geographic area Analyze frequency of breakdowns per area and identify their root cause

SOURCE: Telefonica Global 117


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

118
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC155

Price outlier detection for site contracts

Situation Approach Impact

• Detection of price outliers of sites in • Apply geolocation techniques to site • Value drivers:
specific areas contracts to detect price outliers (sites o Price renegotiation
• Value drivers: with a very high contract when
o Price renegotiation
compared with other close sites).

SOURCE: Telefónica & McKinsey team analysis 119


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC163

Energy efficiency increase in network related building

Situation Approach Impact

• Comparing each energy/unit with all the • Advanced analytics descriptive model • Value drivers:
close neighbors to identify outliers and to identify outliers in network related o Increase building energy efficiency
optimize energy efficeincy buildings based in a energy
consumption/network unit criteria
• Value drivers:
(e.g. Watt, GB, etc.)
o Increase building energy efficiency
• Sources of data:
o Energy consumption/ network unit

SOURCE: Telefónica & McKinsey team analysis 120


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

o Network strategy & planning

o Fixed deployment

o Call center operations

o Technical centers operations & management

o I+M / O+M field operations

o NOC / SOC /TI

o Sites management

o Other areas

3. Use cases assessment guide

121
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC011

Mobile churn prediction

Situation Approach Impact

• Predicting customers with high • Analysis of customer behaviour to • Value drivers:


probability of churn, to launch comercial identify use patterns that increase o Churn reduction
campaigns and customer churn probability
• Value drivers: • Sources of data:
o Churn reduction o SOC: XDRs

SOURCE: Telefónica Global 122


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC061

Customer Service Delivery


Access Network – Use Case #8

Vista 1: Argentina a 14/12/2016 Vista 2: Argentina a 12/02/2018


Velocidad de sincronismo: 10+Mbps - 15Mbps Velocidad de sincronismo: 10+Mbps - 15Mbps
MABR_DS: 10+Mbps - 15Mbps MABR_DS: 10+Mbps - 15Mbps
Número Líneas: 34.328 Número Líneas: 79.547.328
% Línes agregado: 97,96 % Línes agregado: 94,04
Total [%] de líneas Avanzado: 34.328 Total [%] de líneas Avanzado: 79.547

300 With no network investment, we 1,600 300 1,600


hav done upselling with these
customers 1,400 1,400
250 250

1,200 1,200

200 200
1,000 1,000

150 800 150 800

600 600
100 100

400 400

50 50
200 200

0 0 0 0

SOURCE: Telefónica Global 123


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC105

Customer demand based FTTH migrations

Situation Approach Impact

• Customer-based (e.g. OTT video) • Identification of customer patterns • Value drivers:


identification of copper clients to which make them high probable o Extra Revenues
proactively migrate them to fibre customers for migration
o Improve CSI
• Customers could be changed to fiber • Sources of data:
either with no extra costs or with o Traffic analysis (e.g. OTT video)
upselling
o Value of clients
• Value drivers: o Surveys
o CSI o Commercial questions on fiber
o Revenues coverage
o Capacity of mobile sites dedicated
to ground broadband

SOURCE: Telefónica & McKinsey team analysis 124


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC140

Failure predictive models aimed at preparing commercial campaigns to migrate clients from
CU to FTTH

Situation Approach Impact

• Identification of customers that are more • Failure predictive models aimed at • Value drivers:
probable to suffer breakdowns to preparing commercial campaigns to o Available mirror commercial offers /
prepare and launch commercial migrate clients from CU to FTTH upselling / crosselling
campaigns
• Sources of data: o Lower churn
• Value drivers: o Client type / contracted services / o Opportunities for ARPU increase
o Available mirror commercial offers / equipment (upselling, crosseling)
upselling / crosselling o FTTH availability o Better CSI
o Lower churn o Failure predictions in the Cu
o Opportunities for ARPU increase network
(upselling, crosseling)
o Better CSI

SOURCE: Telefónica & McKinsey team analysis 125


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC141

Use client locations to optimize dynamic advertisements

Situation Approach Impact

• Sell aggregated customer information to • Identification of customers common • Value drivers:


customize dynamic advertisements patterns and aggregation of those o Revenues from data sale
campaigns to client profiles customers
• Value drivers: • Sources of data:
o Revenues from data sale o Client type / contracted services /
equipment
o Client location
o Customer house type
o Customer value

SOURCE: Telefónica & McKinsey team analysis


126
ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC144

Understand home allocation of convergent clients to identify mobile users not living in main
client home

Situation Approach Impact

• Understand home allocation of • Advanced analytics model which • Value drivers:


convergent clients to identify mobile identifies which convergent customers o Identification of mobile candidates for
users not living in customer main home do not live in customer main home convergent or mobile only
• Value drivers: • Sources of data: o Low cost commercial campaigns
o Identification of mobile candidates for o Main location of all mobiles within
convergent or mobile only a convergent offer
o Low cost commercial campaigns

SOURCE: Telefónica & McKinsey team analysis 127


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC174

Management of NGN capacity and data flow

Situation Approach Impact

• Prediction of a cusotmer’s growth of use • Advanced analytics model that • Value drivers:
of NGN channels and data flow to predicts customer growth of use of o Identification of candidates with high
automaticly notify commercials as a sale NGN channels and data flow upselling opportunities
opportunity
• Sources of data:
• Value drivers: o Customer NGN use trend
o Identification of candidates with high
upselling opportunities

SOURCE: Telefónica & McKinsey team analysis 128


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC180

Management of flow capacity contracted in APV and AUIP services (NGN)

Situation Approach Impact

• Management of flow capacity contracted • Advanced analytics model that • Value drivers:
in APV and AUIP services by identififying automatically diagnoses and enriches o Identification of candidates with high
and ennriching breakdowns with breakdowns with information on the upselling opportunities
information on the occupation of NGN occupation of NGN channels
channels, and routing it, if required, to
• Sources of data:
commercials
o APV
• Value drivers:
o AUIP
o Calls automation
o TMO reduction

SOURCE: Telefónica & McKinsey team analysis 129


ADVANCED ANALYTICS IN NETWORK OPERATIONS- UC185

Proactive monitoring of traffic profiles

Situation Approach Impact

• Proactive monitoring of traffic profiles • Advanced analytics model that


proactively supervises and
automatically generates faults
• Sources of data:
o NGN: traffic thresholds variation

SOURCE: Telefónica & McKinsey team analysis 130


Contents

1. High potential advanced analytics use cases

2. Advanced analytics use cases library

3. Use cases assessment guide

131
For each use case, we will run an standardized assessment in order to understand both
value at stake and feasibility
Assessment
Dimensions Description Subdimensions Low Medium High
Impact on revenues

Sources of value Impact on cost


1 Impact (conceptual) & value at
stake Impact on quality

Impact on customer experience

Data quantity
Data availability and
soundness
Data quality / engineering

Feasibility of mathematical model


Modelling complexity
Analytics tools & expertise

Architecture requirements
Operative
2 Feasibility requirements
Implementation & tech. support

Business case

Business integration
Business
feasibility
Process ownership

Change management

Use case typology • Advanced Descriptive Analytics • Predictive analytics • Prescriptive analytics • Deep learning and NLP
132
2 Feasibility - Assessment detailed description (1/3)

Dimensions Subdimensions Low Average High


• Data quantity • Main event data is not registered • Data granularity is high enough to isolate • Data is available at the most basic level or at
significant events a level of granularity that is significantly
• Not enough historical data available to create
higher than required (e.g. second-by-second
representative train/test datasets • Some data sources have the desired
when PdM needs to be performed on an
granularity while others are aggregated, and
• Data granularity level can’t isolate significant hourly basis)
therefore, losing part of its descriptive power
events (e.g. tickets not stored individually for
• All data sources share data granularity or is
intelligent care)
even higher than needed
• Each data source has a different granularity
level – some detailed information can’t be
obtained

• Data quality • Low confidence intervals for relevant data (e.g. • Average confidence intervals for relevant • High confidence intervals for relevant data
accuracy < 95% for sensor data) data (e.g. accuracy > 90% for sensor data) (e.g. accuracy>99% for sensor data)

Data Availability • High amount of erroneous / missing data points • Some erroneous / missing data points that • Very limited amount of erroneous/missing
and Soundness still allow for an algorithm to generalize over data points
• Target variable can’t easily be created
available data
• Data checks with proven logic is in place to
• Data is stored in separate sets that’s cannot be
• Target variable can be created from available detect outliers or erroneous data points
easily merged
data
• Target variable has already been defined and
• No data dictionary is available
• Data dictionary can be improved (e.g. add stored
• No consistent format / naming convention has new info such as variable format, number of
• A master table is already created with key
been defined missing values, etc.) and expanded
synthetic variables
• Mostly consistent format / naming
• A clear data dictionary is available explaining
conventions across files
each variable structure
• Consistent format / naming conventions
across files

133
2 Feasibility - Assessment detailed description (2/3)

Dimensions Subdimensions Low Average High


• Feasibility of • The studied phenomenon is unpredictable • Only a subset of data presents a behavior • The problem can be solved using simple
mathematical model • The problem requires the use of very complex trend that can be predicted algorithms
algorithms to be solved • A good performance can be obtained with a • Data trends are easy to found and interpret
• The expected accuracy of a mathematical mathematical model, but the use of • Extremely good results can be found by using
model is extremely low elaborated techniques is needed complex algorithms

Modelling • Analytics tools & • Required tools and algorithms are difficult to • Required algorithm and tools are easy to • Required algorithm and tools are available
Complexity expertise access (need to be purchased or coded from access (e.g. open source software) but not and ready to use in company systems
scratch) currently available • Strong internal experience using relevant
• Non-conventional OS are required • Some internal experience in using relevant tools/algorithms in other use cases in the
• No internal expertise in dealing with relevant tools past
tools / algorithms • Some expertise in tuning and applying • Strong internal expertise in tuning and
• No similar use cases developed in the past relevant algorithms applying relevant algorithms

• Architecture • The process can’t be host inside company IT • The model can be hosted inside actual • The process can be hosted inside actual
requirements infrastructure company IT infrastructure, but some company systems
• Current IT infrastructure processing power concessions need to be done • The company architecture is powerful
doesn’t meet the process requirements • The infrastructure used to host the process is enough to handle the process
• A specific external architecture is needed to temporary and will change in a close future • IT infrastructure is stable and no change in
handle the process close future is planned
Operative • Implementation and • The data sources availability does not meet the • The model can partially be automated • The process can be automated end to end
requirements technical support implementation requirements • Limited knowledge of IT teams in case of • The process can be scheduled
• No support nor supervision of the process technical problems • A dedicated team supervises the model
• Complex coding changes are required to deploy • Medium changes in the code are needed to execution in case of tech. problems
the model in production follow company coding guidelines • The model can be easily implemented inside
• Complex security permission request needed to • Complex coordination of data sources company’s systems
deploy the model in production temporal availability • Agile transition from dev. to production

134
2 Feasibility - Assessment detailed description (3/3)

Dimensions Subdimensions Low Average High


• Business case • The initial investment and/or operational cost • The model performance needed to pay off • The minimum model performance needed to
are high enough to make the use case the initial investment and the operational pay-off is easily achievable with traditional
profitable cost is elevated and complex data AA techniques or easy model
• The minimum model performance to pay off is engineering / modeling will be required
unrealistic

• Business integration • New processes need to be designed from • Current processes need to be partly adapted • Current processes are readily available for
scratch to use the model for the use case the use case to be implemented with
• Several pain points need to be resolved for use • Some paint points need to be resolved for minimum modifications required
case to be implemented use case to be implemented • Little to no pain points need to be resolved
• A business process bottleneck is blocking • Higher levels business processes can be for use case to be implemented
process performance – an AA model will be debottlenecked to increase performance got
useless from AA model
Business
feasibility
• Process ownership • No clear owner of the process • The IT support team is the owner of the • A specific owner is responsible of
• No one in charge of reviewing model operation model maintenance and update of the mode
and performance • Constant review of model performance but • The owner has full control and knowledge of
limited model retrain capabilities model insides

• Change • The area where the model will be implanted is • Moderate resistance to operative change • Department wiliness to change its operative
management not used to work with AA model outputs • Brief learning sessions are needed to train based on the AA processes
• Outputs of the model not understood by the employees about how to interpret machine • Full employers understanding of model /
own users learning results process results
• Complex training sessions are needed to train
the employees

135

You might also like