© 2019

Published by

WHITEPAPER

Advanced Analytics forIntelligent Connectivity:Machine Learning as a Growth Engine forTelcos in the Age of 5G

The world is entering into a new era of ultra-connectivity, as morepeople and devices than ever before tap into operator networksand come online.

The number of global users connected to the mobile internettopped 3 billion in 2017. By 2025, GSMA Intelligence forecasts thatfigure will grow by more than 1.75 billion to reach a total of 5billion1.

In the same timeframe, GSMA Intelligence predicts the market willexperience massive growth in the number of Internet of Things(IoT) connections across cellular and non-cellular technologies.There were already 6.3 billion connected IoT devices in 2017, andthat figure is expected to nearly quadruple to reach 25.2 billion in20252, offering operators a $1.1 trillion revenue opportunity3.

Operator networks provide the magic to fuel this march. Uptakeof 5G technology, with its lightning-fast speeds and ultra-lowlatency, has the potential to enable entirely new user experiencesand a more intelligent kind of connectivity.

Just two years from now, GSMA Intelligence predicts as many as50 million people worldwide will have 5G connections. By 2025,that figure is expected to rise to 1.2 billion4, with Asia, NorthAmerica and Europe leading the charge.

Of course, next generation networks raise the spectre of newcomplexity challenges for operators, and with the influx of newconnections will come a sea of data to navigate. But carriers canrise to the occasion with new data management platforms, usingpredictive analytics and machine learning to correlate network,user, device and application activity for greater customer gainsand operational efficiencies.

Introduction

1 GSMA Intelligence, “Unique subscribers and mobile internet users”. February 2018.2 GSMA Intelligence, “IoT: the next wave of connectivity and services”. April, 2018.3 GSMA Intelligence, “IoT: the $1 trillion revenue opportunity”. May 2018.4 GSMA Intelligence, “5G connections to surpass 1 billion by 2025”. August 2017

Advanced Analytics for Intelligent Connectivity | 3

Intelligent connectivityGSMA defines5 intelligent connectivity as afusion of 5G, IoT and artificial intelligence (AI)technologies which will enable new,transformational capabilities across a wide rangeof verticals, including everything fromentertainment and healthcare to transport andindustry.

5G will play a key role in this mix, providing acritical foundation of ubiquitous connectivitywith increased network capacity, throughputand unprecedented responsiveness. Networkslicing, which allows operators to split a singlephysical network into many layers usingvirtualised architecture, will also become muchmore granular in a 5G world, enabling serviceproviders to tailor connectivity to individualapplications and users.

This will open the door for the advancement ofnew products and services which depend onreliable, low-latency connectivity, such asautonomous cars, augmented reality, real-timegaming, remote surgery and more.

Additionally, operators will gain access to a floodof essential real-world data as 5G drives amassive expansion of the IoT. This informationwill in turn fuel new machine learningapplications, AI development and productpersonalisation.

GSMA’s vision of intelligent connectivitypreviews a world in which insights from big datawill improve decision-making across the board,making roads safer, people healthier, factoriesmore productive, and delivery of products,services and education more on-demand andengaging.

But in order to transform this vision into a reality,organisations need the right tools to managethese next-generation networks and the datawhich flows through them.

5 GSMA, “Intelligent Connectivity: How the combination of 5G, AI, and IoT is set to change the Americas”. September 2018.6 Ericsson blog, “How many network slices are needed”. July 2018.

Managing complexityThough 5G will bring increased opportunities foroperators and other communications serviceproviders, it will also bring increased complexityto the network.

Operators will need to effectively manage eachnetwork slice that is created, as well as the newproducts and services which run on them.

While it’s impossible to know now exactly howmany network slices operators will be juggling,Ericsson tips6 the figure to be “quite high”. That’sbecause even though operators can developslice templates based on the needs of certain

verticals, the specific operational, service andbusiness requirements of each user mean eachtemplate might yield “any number” of individualslices.

In order to keep costs down and preserve thebusiness case for network slicing, operators willneed to turn to automated systems to managethese assets.

Operators will also need to find an efficient wayto sift through and monetise a tidal wave of dataas the number of connected consumer and IoTdevices rises.

4 | WHITEPAPER

Power of predictionPredictive analytics and machine learning canhelp operators tackle a number of theaforementioned challenges, giving them toolswhich can be used across business processesincluding everything from customer experiencemanagement, network performance analytics,capacity planning and revenue assurance tosecurity analytics, fraud detection and myriadother data monetisation use cases.

Simply put, machine learning10 is a tool whichcan help identify patterns and relationships in adata set and predict outcomes based on thoseinputs using statistical models. The insightsgleaned through the use of advanced analyticsand machine learning allow companies to beproactive rather than reactive, and offer bothbetter customer service and better products.

Recognising and tapping information from theappropriate information sources11 is the first stepin turning data into dollars. Sources run thegamut, including everything from network data,call detail records (CDR), sensors and otherconnected devices, apps, services and CRMsystems.

Once the data is gathered, it can be used togenerate models that predict future actions andoutcomes, such as the probability of customerchurn or how an impending event will impactnetwork performance.

There are applications for predictive analyticsacross business segments. For example,predictive analytics can help operators provideservice assurance and improve user experienceon their networks. Additionally, operators andvendors can use it as a tool for IoT datamanagement and predictive maintenance fordevices. Media companies and over-the-top(OTT) video providers can apply predictiveanalytics to gather audience metrics andpersonalise content offers. It can also helpadvertising technology firms crunch data fasterto measure ad performance and reallocateresources accordingly.

7 GSMA, “Scaling the IoT Becomes the Next Frontier for the Mobile Industry in 2018”. January 2018.8 Ericsson, “Exploring IoT Strategies”. April 2018.9 Mobile World Live, “Nokia CTO puts value on edge cloud”. May 2018.10 Microfocus, “Unlock Machine Learning for the New Speed and Scale of Business”. September 201711 Microfocus, “Unlock Machine Learning for the New Speed and Scale of Business”. September 2017

Svetlana Grant, GSMA’s IoT Program Director,wrote7 the quantity of data generated by IoTdevices will be “unthinkable,” adding “in its raw,unstructured form, this data is of limited value toanyone, and current analytics tools will be simplyunable to cope.” But she also noted that datacan become a “bankable asset” if managed andused properly.

An Ericsson survey8 of 20 leading operatorsreleased in April 2018 found they were pursuing

multiple paths to generate revenue from the IoT,but concluded 70 % lacked a well-definedstrategy for tackling the market opportunity.

In order to reap the most benefit from both 5Gand the IoT and be successful in the long-term,Nokia CTO Marcus Weldon said9 operators mustexpand beyond their traditional role asconnectivity providers and use data to offervalue-added services.

Advanced Analytics for Intelligent Connectivity | 5

Telco and Mobile Operator Use Cases Data Sources

Capacity Planning and Management Network Usage

Network Performance Monitoring Call Detail Records (CDR)

Customer Churn Analysis Internet Protocol Detail Records (IPDR)

Targeted Marketing and Promotions Network Traffic

Content Personalization Equipment Probes

Fraud Detection Sensors

Revenue Assurance Log Files

Customer 360 Web and Mobile Applications

Network Optimization User Events

Security Analytics Content Metadata

Clickstream

Latency or Bitrate

Business Applications

Customer Relationship Management (CRM

Enterprise Resource Planning (ERP)

Marketing Automation

Digital Advertising

Impressions

Clicks

Conversions

6 | WHITEPAPER

Hitting a wallTools like predictive analytics and machinelearning go hand in hand with the idea ofintelligent connectivity. Absent the ability toprocess the massive amounts of data generatedby mobile networks – including 5G and the IoT –operators and other organisations will be unableto effectively capitalise on the opportunity tooffer not only network innovation but also new,personalised products and services.

However, legacy systems are ill-equipped for thechallenge.

One issue common to many organisations issiloed data12, which occurs when differentbranches of a single company have control overdifferent data sets. Data from one branch maybe stored in the cloud, while another is kept inan on-premises data warehouse and still anotheris stored in a Hadoop data lake.

These historic silos make it difficult fororganisations to get a unified picture of theirdata and run analytics across all of theseinformation pools. Efforts to bridge the gapbetween these resources can be both expensiveand time consuming since they requirespecialised skillsets, making projects which relyon analytics more complex to manage andsubject to delays. The process of merging datafrom disparate systems into a unified set canalso leave companies working with inconsistentdata points and incomplete information, yieldingresults which may not live up to expectations.

Legacy systems also present obstacles to scaleand speed. Companies are now collecting datafrom more sources than ever before – cellphones, sensors, machine equipment, thenetwork itself. But this influx of informationcreates a number of challenges since the oldtools companies have at hand were not built tohave so many users running complex analyticson such a large volume of data.

One of the key value propositions offered byanalytics is the ability for companies to gleanactionable information from data sets veryquickly.

When corporate SLAs and high customerexpectations are on the line, the speed andaccuracy of analytics are critical, particularly ascompanies move to automate more processes. Inthe case of a network outage, for example, theright system could help an operator reduceresponse times from hours, minutes or secondsto just milliseconds, and pinpoint which servicesand revenue streams are most impacted toprioritise restoration efforts. It might evenidentify problems with a key network asset so itcan be replaced or maintained in advance, thuspreventing the outage altogether.

Of course, the more users – both internal andexternal – who can access those kinds ofinsights, the more value companies can generatefrom a given data set.

However, attempts to run too many queriesconcurrently on a legacy database can slowdown processing significantly, meaningorganisations and their clients cannot be as agilewith their data as they might like.

Companies which have recognised theseproblems often double down on legacy systems,investing more money on the solutions theyhave in place in an effort to head off the issues.But all too often these systems rely onproprietary technology, which limits companies’hardware and software choices.

12 Information Week, “Data Silos: Now and Forever”. November 2018

Advanced Analytics for Intelligent Connectivity | 7

Analytics in action – success storiesThe difference between legacy analytics systemsand a solution like Vertica can be stark. Beloware descriptions of three separateimplementations of Vertica for very different usecases which illustrate just how wide the gap is:

Anritsu: service assuranceAnritsu, a provider of test and measurementsolutions for network operators, faced an acutepair of challenges in 2012: rapidly growing datastorage needs and increased demand among itsclients for shorter response times from dataanalysis. Where formerly customers were happyto receive insights within 15 to 30 minutes, theybegan asking for results from larger data setswithin a few seconds.

After evaluating factors including cost,scalability and time to deployment, Anritsuopted to switch to the Vertica Analytics Platformrather than spending big to upgrade its legacysystem.

In addition to sparing the company around $2million per year in upgrade costs, the moveallowed Anritsu to improve performance andincrease employee productivity.

With Vertica, the company is now aiming to usepredictive analytics and machine learning to helpoperators prevent issues before they even arise13.

Catch Media: personalisationCatch Media similarly turned to Vertica to solveissues related to scale and speed, albeit with adifferent goal in mind.

The company’s database simply wasn’t scalableand enterprise-ready enough to grow with them.Catch Media needed a data analytics platformcapable of tracking hundreds of millions ofusers, with tens of billions of data eventtransactions processed in near-real-time on amonthly basis, with no data aggregation orreduction in query performance or featurelimitations.

Vertica offered a solution to this. BecauseVertica is so scalable, Catch Media can query anunlimited amount of historical data without theneed for pre-aggregation. Vertica gives CatchMedia the performance, scalability, flexibility, andpricing structure to help their clients engageconsumers, drive lifetime value, and increaserevenues. As a result, they have shown a 50percent increase in engagement through theeffective segmentation of consumers intodistinct audiences that can be targeted in-realtime with messaging and recommendationspersonalized to their content consumptionbehavior14.

Tuenti: user experienceTelefonica-owned mobile virtual networkoperator Tuenti Technologies also selectedVertica to process data faster and glean deeperinsights to enhance its mobile app.

As its user base grew, Tuenti found itincreasingly difficult and time-consuming toanalyse data stored across its MySQL andHadoop pools. It wanted deeper granularly andfaster insights from its data sets to allow it toadd and remove features from its app inresponse to customer demand.

Using Vertica, Tuenti was able to compress itsraw Hadoop data set from 1TB to 50GB, andaggregate and measure multiple metrics daily.Vertica also made it possible for moreemployees to tap into insights from the data set,opening it to more than 150 internal developersrather than the three-person analysis team thatpreviously had access to the information.

Now, the company is able to push out three tofour app updates per month using insights fromuser data provided by Vertica15.

13 https://www.vertica.com/wp-content/uploads/2017/01/r24-HPE-Vertica-ROI-case-study-Anritsu.pdf14 https://www.brighttalk.com/webcast/8913/340965/4-steps-to-drive-customer-lifetime-value-gain-real-time-behavior-intelligence15 https://www.vertica.com/wp-content/uploads/2018/02/Tuenti-Case-Study.pdf

8 | WHITEPAPER

ConclusionIntelligent connectivity is on the horizon, but the data flowing in from agrowing number of consumer and IoT devices remains an underutilisedresource.

The advent of 5G will bring with it even more potential, offering a potentcombination of customisable network slices, faster speeds and ultra-lowlatency. Such capabilities will give companies the chance to offer new andpersonalised experiences to users around the globe, but truly responsive andintelligent data management is the key to unlocking this future of possibility.

The companies that will thrive in the age of intelligent connectivity will bethose which scrap outdated data storage and analysis systems in favour ofsolutions which offer unified visibility across data silos, massive scalability,faster insights and an ability to generate accurate predictions using advancedanalytics and machine learning.

With a solution like Vertica, companies will finally be able to fully leverage thepower of advanced analytics and machine learning to effectively serve new usecases, offer highly-contextual user experiences, improve operationalefficiencies, deliver better network quality and manage the increasingcomplexity of our connected world.

Advanced Analytics for Intelligent Connectivity | 9

© 2019

Produced by the mobile industry for the mobileindustry, Mobile World Live is the leading multimediaresource that keeps mobile professionals on top of thenews and issues shaping the market. It offers dailybreaking news from around the globe. Exclusive videointerviews with business leaders and event reportsprovide comprehensive insight into the latestdevelopments and key issues. All enhanced by incisiveanalysis from our team of expert commentators. Ourresponsive website design ensures the best readingexperience on any device so readers can keep up-to-date wherever they are.

We also publish five regular eNewsletters to keep themobile industry up-to-speed: The Mobile World LiveDaily, plus weekly newsletters on Mobile Apps, Asia,Mobile Devices and Mobile Money.

What’s more, Mobile World Live produces webinars,the Show Daily publications for all GSMA events andMobile World Live TV – the award-winning broadcastservice of Mobile World Congress and exclusive hometo all GSMA event keynote presentations.

Find out more www.mobileworldlive.com

Disclaimer: The views and opinions expressed in this whitepaper are those of the authorsand do not necessarily reflect the official policy or position of the GSMA or its subsidiaries.

Vertica is trusted by thousands of leading data-drivenenterprises around the world, including AT&T, Etsy,Cerner, Intuit, Uber and more to deliver speed, scaleand reliability on mission-critical analytics. The VerticaAnalytics Platform is purpose built from the very firstline of code for Big Data analytics. It is designed foruse in data warehouses and other big data workloadswhere speed, scalability, simplicity, and openness arecrucial to the success of analytics. Vertica relies on atested, reliable distributed architecture and columnarcompression to deliver blazing fast speed. All basedon the same powerful, unified architecture, the VerticaAnalytics Platform provides you with the broadestrange of deployment models – including on-premises,in the clouds, and on Hadoop - so that you havecomplete choice as your analytical needs evolve.

Visit www.vertica.com