Success Story Mobile Operator Big Data Analytics & Actions Experiences and Benefits Achieved
Agenda 1 2 Company Background Focus Customer Profile 3 The Solution Decision Process 4 5 6 The Solution The Results Looking to the Future
mcentric Background Solutions integrated with Network Elements from best-in-class vendors:
Big Data Activities 1 HADOOP in Roll-out Requirements phase Requirements phase 2 BDAs LIVE BDA in Roll-out 1
1 2 Company Background Focus Customer Profile 3 The Solution Decision Process 4 5 6 The Solution The Results Looking to the Future Focus Customer Profile
Focus Customer s Profile 2 nd Mobile Telecoms Operator in Nigeria with more than 25 Million subscribers with sister companies located in Benin and Ghana. The vast majority of the Company s subscriber base is prepaid. A nationwide network that leverages infrastructure from various best in class providers. There is NO viable fixed data network service provider in the country. I Increase CSR effectiveness & reduce operational costs II Increase Mobile Data Penetration and Revenues III VIP Services & Customer Empowered actionable CEM Facilitate adherence to regulatory and LEA requests Improve the ROI on existing Mobile Data infrastructure Increase Revenue and Fidelity across the entire Subscriber base GOAL provide the business with a 360º up-to-date view of their Customers
1 2 Company Background Focus Customer Profile 3 The Solution Decision Process 4 5 6 The Solution The Results Looking to the Future The Decision Process
Client mcentric s Drivers for Big Data Adoption T Bytes Tier 3 Requirements Tier 1 P Bytes The Decision Process ZEROimpact Capacity scaling Transparent to Applications and Deployment collect it now / analyze it later 1,000,000,000 Events per day 1 Gbps Peak transfer rate +500 GBytes per day(*) ALL unstructured & structured information available 1,000s of CSRs 10s of Analysts First experience with big data for client. 1,000s of small nodes not an option. Fewer high capacity carrier grade servers, guaranteed scalability. Reliable quality, guaranteed, short timeframe delivery Big Data Appliance
1 2 Company Background Focus Customer Profiled 3 The Solution Decision Process 4 5 6 The Solution The Results Looking to the Future The Solution
Architecture Overview Customer Care / Technical Support / LEA / Revenue Assurance Corporate Analyst - Data Scientist Browser Based xdr Navigator Consumer PrePaid / PostPaid Automated & Continuous Learning Clusters / Patterns Big Data Appliance User traffic Prediction MAHOUT GIRAPH Micro Profiling Segments Actions Delivery PROFIT GURU CORE & ACTIONS ENGINE Cloudera Hadoop Net Events & Control Data Continuous ETL CEM & POS Data NoSQL HDFS Firewall/ Border Events MYSQL IMPALA Continuous Actions Processing Real-time Analysis of Data Statistically Representative Sample Unstructured & Structured Data
SUBSCRIBER Data Repository Analytics & Prediction Discovery Operational Telco Event Normal ization Network Elements Stream Event Processing Data Repositories Analytics & Prediction Exploration of Data Dashboards Mahout Supervised learning: classification: method: use training data to create a model; classify new observations output is categorical: eg. yes/no; blue/green/red; loyal/churner recommendation (collaborative filtering): method: use training data to create a recommender; output is score per item: eg. book1 (90%), book2 (56%),.. Giraph system for graph processing: method: construct a network of nodes and connections; calculate properties of each node within the network output: centralities (betweenness, PageRank, EigenVector, degree,..), degree of separation between two nodes, received (event) influence from other nodes; eg Facebook uses Giraph to analyze the social graph formed by users and their connections. Operational Discovery SUBSCRIBER SNA CENTRALITIES SubscriberID CentralityID CentralityValue PartitionDate SUBSCRIBER PREDICTIONS SubscriberID PredictionID PredictionValue PartitionDate Algorithm Adjustments Analysis SNA CENTRALITIES CreateDate CentralityID CentralityName CentralityDescription PREDICTIONS CreateDate PredictionID PredictionName PredictionDescription Mahout Unsupervised learning: clustering: method: no training data; group data points together in clusters output: clusters of data points to explore Dashboard Tools SNAVisual NodeXL Visual Insight D3 Exploration Tools STATXPLORE R Visual Insight REFERENCE TABLES Data Repository Impala Map/Reduce HDFS
1 2 Company Background Focus Customer Profile 3 The Solution Decision Process 4 5 6 The Solution The Results Looking to the Future The Results Success Story
Days Execution Process Oracle s Big Data Appliance Implementation Methodology mcentric s Events Normalization Software & Methodology Oracle ACS team well orchestrated implementation and certification approach Success Story 1 physically deploy and power-up the HW 1 certification validation process Additional Take-Away: 2 resilience testing assistance to mcentric 1 mcentric SW Install & Configuration Historical data challenge - 2x to 3x pipeline capacity!
100 Increased efficiency & Customer Satisfaction Success Story xdr Navigator Real Time browsing, search, and filtering tool Departments benefited; Customer Care, Technical Support, LEA and Revenue Assurance 25% calls Near real-time data access Im sorry Sir can you please call back in some hours - 45,000 CALL PROCESSING MINS/DAY =
CSR and Tech. Support Empowerment Success Story Search By MSISDN, IP Filter by any columns, AND and OR combinations Export content masks Find top 10 Peers, Destinations Trace back inquiry for target Site activity Activity inquiry for a known subscriber Trace Self-care navigation events Navigation trouble-shooting
The Situation: Subscribers that have a device capable of consuming large amounts of data, but that do not have a contracted QoS that permits them to enjoy the full benefit of the device s capability Device - Data Plan Mismatch The Opportunity: Increased revenue via pattern discovery upsell to Subscriptions with higher Bandwidth. The Solution: Continuous synthesis of network information to identify underexpected data throughput in the most frequently visited locations of each Customer with a 3G+/4G capable device where the radio access technology connected matches handset capability and on a restricted QoS data plan, apply pattern segmentation for actions engine treatment. The Results: 500,000 Pattern Matched 40% Conversion Rate Event Synthesis & Pattern Based Segmentation ARPU Increase $3.00 Success Story
1 2 Company Background Focus Customer Profile 3 The Solution Decision Process 4 5 6 The Solution The Results Looking to the Future Looking to the Future CEM,VIP & Turbo Passes Nokia quote on the CEM tool for customer care 9-Sep-2014 - Telenor Denmark, which has trialled the service, claimed it reduced the length of customer care calls by up to 30 percent, from an average of 11 minutes down to eight minutes. mcentric Sep-2014 - mcentric s CEM suite delivers even better business results by adding the experience from the customer s perspective and empowering the customer to take action.
xdr Navigator CEM 360º Millions Actions Pattern Discovery Pattern Matching Push Notification Outbound Dialer SMS
Stakeholders Empowerment SmartAPP that is available on all fundamental mobile device platforms active in today s mobile networks. Enabling Direct feedback Simple 2 click submission. Optional text can be introduced Configurable frequency and events to trigger presentation Detailed Analysis of Coverage Optionally available to Customer Can optionally include other operator coverage Take Action On Demand Bandwidth Upgrade Guaranteed Voice connections Options not provided if for example, insufficient bandwidth available in the network, or congestion issues are not on the mobile network and hence Customer will not perceive a benefit. Control for the Customer Default values set by Operator Customer has complete control over what and how much is collected. Complete opt-out available. Customer becomes a stakeholder improving their own service experience
Other Use Cases as Work-in-Progress THANK YOU!