predicting customer experience through hadoop and customer behavior graphs
TRANSCRIPT
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Predicting Customer Experience through Hadoop and Customer Behavior Graphs with Hortonworks and Apigee
We do Hadoop.
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Your speakers…
Sanjay Kumar General Manager, Telecom Hortonworks
Sanjeev Srivastav VP Data Strategy Apigee
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Enhancing Customer Experience with Hadoop We Do Hadoop
Sanjay Kumar General Manager, Telecom Hortonworks
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
The New Landscape of the Telecom Industry
Service Providers
Social Media & Mobile:
Explosion of rich customer data through
Social Media and Mobile Apps for customer
sentiment & Interests
Customer Expectation:
With the cultural impact of web and mobile,
customers are expecting greater levels of service
and responsiveness
Competitive Differentiation:
As other service providers deliver similar levels of telecom service
and coverage, other areas of service levels
are needed
New Digital Ecosystem:
Greater value of Data on digital ecosystem for
Customers and Partners driving Data
Monetization
Internet Of Things: Explosion of data from IOT with benefits aligned with insight not correlated to
data volumes
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Service Provider Focus
Service
Providers
Customer Experience Management - Enhance End-to-end Experience of Customer - Become Trusted Partner to Customer - Awareness of customer’s needs when and where needed
New Business & Consumer Services - New Digital & Infrastructure Services - Data Monetization - M2M, IoT, Analytics-as-a service
Network Optimization - Move to Software Driven Networks - Leverage Network Data Assets - Self optimizing and provisioning
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Traditional systems under pressure Challenges • Constrains data to app • Can’t manage new data • Costly to Scale
Business Value
Clickstream
Geolocation
Web Data
Internet of Things
Docs, emails
Server logs
2012 2.8 Zettabytes
2020 40 Zettabytes
LAGGARDS
INDUSTRY LEADERS
1
2 New Data
ERP CRM SCM
New
Traditional
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Tomorrow: A Data-Centric Model for Your Business
DATA-CENTRIC
Limitations: • Multiple copies of data • Difficult cross-system integration • Upper-limit on data volumes
before harming performance
Advantages: • One version of the data • No need for cross-app integration • System scales linearly
APP-CENTRIC
App1 App 2 App 3 App 4 App 5 App 6
App Centric will break down with x10, x100,x1000… Need to shift to Data Centric
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Modern Data Architecture emerges to unify data & processing
Modern Data Architecture • Enable applications to have access to
all your enterprise data through an efficient centralized platform
• Supported with a centralized approach governance, security and operations
• Versatile to handle any applications and datasets no matter the size or type
Clickstream Web & Social
Geoloca3on Sensor & Machine
Server Logs
Unstructured
SOU
RC
ES
Existing Systems
ERP CRM SCM
AN
ALY
TIC
S
Data Marts
Business Analytics
Visualization & Dashboards
AN
ALY
TIC
S
Applications Business Analytics
Visualization & Dashboards
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
°
HDFS (Hadoop Distributed File System)
YARN: Data Operating System
Interactive Real-Time Batch Partner ISV Batch Batch MPP EDW
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
HDP delivers a completely open data platform
Hortonworks Data Platform 2.3
Hortonworks Data Platform provides Hadoop for the Enterprise: a centralized architecture of core enterprise services, for any application and any data.
Completely Open
• HDP incorporates every element required of an enterprise data platform: data storage, data access, governance, security, operations
• All components are developed in open source and then rigorously tested, certified, and delivered as an integrated open source platform that’s easy to consume and use by the enterprise and ecosystem.
YARN: Data Operating System (Cluster Resource Management)
1 ° ° ° ° ° ° °
° ° ° ° ° ° ° °
Apa
che
Pig
° °
° °
° ° °
° ° °
HDFS (Hadoop Distributed File System)
GOVERNANCE BATCH, INTERACTIVE & REAL-TIME DATA ACCESS
Apache Falcon
Apa
che
Hiv
e C
asca
ding
A
pach
e H
Bas
e A
pach
e A
ccum
ulo
Apa
che
Sol
r A
pach
e S
park
Apa
che
Sto
rm
Apache Sqoop
Apache Flume
Apache Kafka
SECURITY
Apache Ranger
Apache Knox
Apache Falcon
OPERATIONS
Apache Ambari
Apache Zookeeper
Apache Oozie
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Only HDP delivers a Centralized Architecture HDP is uniquely built around YARN serving as a data operating system that provides multi-tenant Resource Management, consistent Governance & Security and efficient Operations services across Hadoop applications.
Hortonworks Data Platform
YARN Data Operating System • A centralized architecture of
consistent enterprise services for resource management, security, operations, and governance.
• The versatility to support multiple applications and diverse workloads from batch to interactive to real-time, open source and commercial.
Key Benefits
• Multiple applications on a shared data set with consistent levels of service: a multitenant data platform.
• Provides a shared platform to enable new analytic applications.
• Delivers maximum cost efficiency for cluster resource management. Fewer servers fewer nodes.
Storage
YARN: Data Operating System
Governance Security
Operations
Resource Management
Existing Applications
New Analytics
Partner Applications
Data Access: Batch, Interactive & Real-time
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
HDP: Any Data, Any Application, Anywhere
Any Application • Deep integration with ecosystem
partners to extend existing investments and skills
• Broadest set of applications through the stable of YARN-Ready applications
Any Data Deploy applications fueled by clickstream, sensor, social, mobile, geo-location, server log, and other new paradigm datasets with existing legacy datasets.
Anywhere Implement HDP naturally across the complete range of deployment options
Clickstream Web & Social
Geoloca3on Internet of Things
Server Logs
Files, emails ERP CRM SCM
hybrid
commodity appliance cloud
Over 70 Hortonworks Certified YARN Apps
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Social Media
Sentiment
The View from the Customer
Call Center Interaction
Quality of
Service
Lifestyle & Interests
Clickstream
Geolocation
Web Data
Internet of Things
Docs, emails
Server logs
Streaming: Network Probes, Click Stream, Sensor, Location
Batch: Call Detail Records
On-Line: Customer Sentiment
Unstructured: Txt, Pictures, Video, Voice2Text
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
DELIVERY
The Destination: Data-Centric Operations
Clickstream
Geolocation
Web Data
Internet of Things
Docs, emails
Server logs
Streaming: Network Probes, Click Stream, Sensor, Location
Batch: Call Detail Records
On-Line: Customer Sentiment
Unstructured: Txt, Pictures, Video, Voice2Text
Personal Data Analysis & Customer Insight Services To Customer & Partners
Hadoop Distribution with Yarn: Allows central source of data across all mediums of ingestion and interaction
Existing & Legacy Systems can Contribute and Participate: May extend the life of existing and legacy systems from enriched data
New Applications interact with Data Lake, not each other: Next Generation Apps build around data and can deliver to customers and partners
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Hadoop Driver: Enabling the data lake SC
ALE
SCOPE
Data Lake Definition • Centralized Architecture
Multiple applications on a shared data set with consistent levels of service
• Any App, Any Data Multiple applications accessing all data affording new insights and opportunities.
• Unlocks ‘Systems of Insight’ Advanced algorithms and applications used to derive new value and optimize existing value.
Drivers: 1. Cost Optimization 2. Advanced Analytic Apps
Goal: • Centralized Architecture • Data-driven Business
DATA LAKE
Journey to the Data Lake with Hadoop
Systems of Insight
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Journey to Enhanced Customer Experience
Real-time Event Action based on Customer Context & Location
Collection of Data for Customer Insights (Data Science)
Actionable Insights through Customer Models (BI)
Current State of the Customer (Dynamic Customer Profile - NPS)
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
HDFS Raw Event Storage
Customer Experience Journey – Data Lake for Collection
1 ° °
° ° °
° °
HBase Processed Event Storage
° ° °
° ° °
° ° N
°
Mul3tenant Processing: YARN
(Hadoop O
pera.ng System)
Metadata M
anagement HCatalog
Hive / Tez
(Interac.ve Query)
ISV
(YARN Apps, i.e. HPA / LASR)
Slider
(Always-‐on Services)
Message Queues
Log Files
Web Services
JMS
Update Data Lake
Network Probe Events
ODBC / JDBC
Rest API
Native API
External Customer Data References
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
HDFS Raw Event Storage
1 ° °
° ° °
° °
HBase Processed Event Storage
° ° °
° ° °
° ° N
°
Mul3tenant Processing: YARN
(Hadoop O
pera.ng System)
Metadata M
anagement HCatalog
Hive / Tez
(Interac.ve Query)
ISV
(YARN Apps, i.e. HPA / LASR)
Slider
(Always-‐on Services)
Message Queues
Log Files
Web Services
JMS
Update Data Lake
Network Probe Events
ODBC / JDBC
Rest API
Native API
External Customer Data References
Customer Experience Journey – Actionable Insights
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
HDFS Raw Event Storage
1 ° °
° ° °
° °
HBase Processed Event Storage
° ° °
° ° °
° ° N
°
Mul3tenant Processing: YARN
(Hadoop O
pera.ng System)
Metadata M
anagement HCatalog
Hive / Tez
(Interac.ve Query)
ISV
(YARN Apps, i.e. HPA / LASR)
Slider
(Always-‐on Services)
Message Queues
Log Files
Web Services
JMS
Update Data Lake
Network Probe Events
ODBC / JDBC
Rest API
Native API
External Customer Data References
State of the Customer - Net Promoter Score - Appetite for Info score - Target Advertising Profile - # Dropped calls today/month - Sentiment / Churn Score
HBase /Accumulo Real-‐3me Serving
° ° °
° ° °
° ° N
In-‐Line Memory
(Spark)
Update Customer Profile and Scores
Customer Experience Journey – Current State of Customer
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
HDFS Raw Event Storage
Customer Experience Journey – Real-time Event Interaction
1 ° °
° ° °
° °
HBase Processed Event Storage
° ° °
° ° °
° ° N
°
Mul3tenant Processing: YARN
(Hadoop O
pera.ng System)
Metadata M
anagement HCatalog
Hive / Tez
(Interac.ve Query)
ISV
(YARN Apps, i.e. HPA / LASR)
Slider
(Always-‐on Services)
HBase /Accumulo Real-‐3me Serving
° ° °
° ° °
° ° N
Streaming Event Processor: Storm
Machine Learning
(Spark)
Indexing
(Lucene)
Rules Processing
(Drools)
In-‐Line Memory
(Spark) Message Queues
Log Files
Web Services
JMS
Enrich Events with Customer info
And Score Matrix
Update Data Lake
Real-time Intelligent Action - Marketing Promotions - Next Best Action - Dynamic Network Provisioning
Network Probe Events
ODBC / JDBC
Rest API
Native API
Messaging Platrom
: KaLa Update Customer
Profile and Scores
External Customer Data References
State of the Customer - Net Promoter Score - Appetite for Info score - Target Advertising Profile - # Dropped calls today/month - Sentiment / Churn Score
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Hortonworks for Customer Experience
Functional Area
Solution Components / Use Case
Description Problem Addressed Business Benefits
Customer Experience Management
Central Data Lake for 360 Customer View
Visibility of customer household view across services and accounts through ingestion of account service and event data into a central Data Lake with views into granular customer’s service experience
- Silo view of customer in different systems
- Social media unstructured data does not fit into existing EDW
- Incorporating customer call center discussions into customer view
- Complete view into customer experience across all services
- Reduction in Customer Churn
Dynamic Customer Profile
Summarized instant view of customer across service identifiers and customer key performance metrics and ‘net promoter scores’. Used for immediate view of customer profile
- How to react to customer contact & events based on their experience
- What is the customer’s experience level and context at this moment in time
- Next Best Action based on customer’s experience with service provider (Retail /Call Center)
- Greater targeted marketing/advertising
Real-time Next Best Action
Real-time streaming of network event data to identify customer location
- How to determine next best action when and where they are most appropriate to customer
- Marketing and CEM analysis is after the fact; need for real-time
- Context sensitive promotions 10x customer acceptance
- Improves customer experience levels and customer retention
© Hortonworks Inc. 2011 – 2015. All Rights Reserved
Customer Experience Management & Marketing Journey
HDP Landing Zone
HDP DataLake
Real-‐.me Streaming HDP DataLake
Dynamic Customer Profile
360 Customer Awareness & Household
View
Loca3on Based CEM & Real-‐3me
customer response Next Best Ac3on
Customer Sen3ment
Customer Aware Loca3on Based Promo3ons
Mul3-‐channel Customer Scoring Models
Telco Customer Care using Hadoop & Customer Behavior Analytics
Sanjeev Srivastav / [email protected]
Agenda
The changing landscape in customer care • Interactions across multiple channels • Key experience management strategies
Data lakes for ongoing analysis and action Visualize and predict • Sequences of activity across call center and operations data • Next-best-actions, churn, etc. based on activity sequences
23
24
The Changing Landscape !in Customer Care!
24
Traditional models of customer behavior were largely based on
customer profiles.
New focus on behavior models.
25
Data Lakes!
25
Investments in storing customer engagement data across various
interaction channels
Enterprises see interaction fragments
27
Email Analytics
Mobile Analytics
Web Analytics
Retail Analytics
Social Analytics
Call Center Analytics Time
Contextual and individualized actions
30
Customer journey insights
Individual and contextual actions
Real time
?
Use case – call center + operations data Predict the
customers likely to contact the
call center
Predict the likely reason of the
next call
Avert the next call
31
Use case – call center + operations data Predict the
customers likely to contact the
call center
Use sequence of events +
customer profile
Predict the likely reason of the
next call
Illustrate with individual customer
experience
Avert the next call
Act during call or after call
32
Use case – digital channel abandonment
Interaction sequences ending in abandonment of
a digital channel
Profile of customers and relevant metrics
Interactions after abandonment
Evaluate customer
preferences
Likelihood of abandonment
Influence engagement
with customers
33
Responding to trends: Key technology strategy
Data Lake for cross-channel experience
Journey analytics
Predictive analytics
Actionability § Batch and real-time engagement § APIs and interfaces with existing systems
§ Models for recommendations and targeting § Next best actions
§ Customer journey visualization § Time and sequence query interface
§ Raw data, not summarized or aggregates § Time sequences preserved
Users Apps Developers Models + APIs Data + APIs Backend
End–to-end System
34
Apigee Insights Demo • Customer interaction across various “channels”
– Web site – IVR – Call center agents – Retail stores
• Visualize – “Channel surfing” – Customer satisfaction
• Act – Segmentation, based on customer experience
35
Visualize: repeated calls to a call center • Calls center data • Operations data • Network data
36
IVR change 25
Escalation 226,359 (66,532 dropoff)
Network 9,340
Education 18,588 (6,119 dropoff)
Internal issue 9,121
Equipment issue 9,121
Internal issue 661
(blank) 576
Network 528
No sound 27
No pic 29
Education 107 (35 dropoff)
Equipment issue 765 (243 dropoff)
Predict: using customer profiles and events
Event sequences
before t (training)
Prediction at time=t
Events after t (test)
38 Machine learning @scale
Ok to have small number of
historical events that correlate with predicted events
Apigee Insights platform and tools
Machine learning (ML)
Segment customers
Individualized actions
ML feedback § APIs for customer engagement and recording customer responses
§ Engine for hyper-personalization; in-context recommendations
§ Tool for business analysts; data interfaces to business systems
§ Models of behavior (Apigee GRASP + 3rd party models) § Flexible modeling time windows
Users Apps Developers Models + APIs Data + APIs Backend
End–to-end System
39
Next Steps…
Download the Hortonworks Sandbox
Learn Hadoop
Build Your Analytic App
Try Hadoop 2
More about Apigee & Hortonworks http://hortonworks.com/partner/apigee
Contact us: [email protected]