Alok Dashora bi analytics journey information excellence

Challenge in

deploying BI Solutions

Alok Dashora, IT Strategy and Consulting

Thank You

for hosting us today

Information Excellence2012 July/August Session

BI & ANALYTICS JOURNEYFROM GROUNDS UP

IN A MEDIA BUSINESS

Alok DashoraInformation Excellence , 4th July,2012

Disclaimer

• This presentation is based on my personal observations being a part of business and will have no relevance to business actuals.

• The contents are intended for technological knowledge exchange and not for business strategy development.

• Presentation does not contain any copyrighted information.

• Any similarities will be merely consequential.

•• Initial StageInitial Stage•• Initial StageInitial Stage1

•• Expansion StageExpansion Stage•• Expansion StageExpansion Stage2

•• Efficiency StageEfficiency Stage•• Efficiency StageEfficiency Stage3

Business Phases

•• DataData•• DataData1•• ReportsReports•• ReportsReports2•• InformationInformation3•• KnowledgeKnowledge4 4

•• InsightInsight•• InsightInsight55

•• ActionAction•• ActionAction66

Data to Insights Cycle

Initial Stage

Business Dynamics

• Establishment • Brand• Technology• Timelines /Events• Finances• Refinements

Data Life Cycle

• Data Quality• System Scalability• Technology Superiority• Information Security• Reliability • Timelines of Business Events

Expansion Stage

Busin

ess V

olum

e Pr

ojec

tion

Initial Stage

Efficiency Stage

Business Dynamics

Growth Stage

Business Scenario

• Business Volumes Increasing• Scalability Testing• Business Models Evolution• Performance Enhancement

Technology Journey• Report Requirements Emerge• Sales Performance Reports• Installation Performance Reports• Call Center Performance Reports• Isolated Time Delayed Data • Business Models Evolution• Multi System syncronization

Dawn of quench for information• Need of integrated picture

across the systems, business functions

• Life Cycle Views of Customer, Inventory, Fund Flows

• Trend Analysis from Multiple touch points

• Integration of heterogeneous reports

• Lead time reduction

Incr

emen

tal

ETL

Customer Decision Mart and Analytical Data Foundation

Solution enabling Data to Action Marketing Lifecycle with integrated Solution Suite

Propensity Models

Deactivationbehavior

Subscriber spends

Package characteristics

Third PartyData Feeds

Subscriber Demographics

CallCenterRecords

Active Services

Subscriber Single View

Subscriber self-care

Master Tables

Summary & RollUps

Marketing Variables

LTD Values and Scores

Bandings

Subscriber Events

Adhoc, Train of Thought Analysis

DashboardsNRC, MMR, ARPU,

Score CardingChurn, Behaviour

Customer Base

Transactions

Cal l records

VariablesBanding

Virtual fields

Exception Dashboards

ADS

Data

Foun

datio

nDa

ta Fo

unda

tion

Campaigns

Data Quality and Process Audit

Iteratively Enriched Marketing Decision Mart

Subscriber

dealer

packages

Recharges

Add Ons

Promos & Campaigns)

CRM Data

Billing

Cust

omer

Dec

isio

n M

art

Vend

or

ACE

ADS

Data Flow across Modeling Environments

KXEN

SAS

Derived Variables

Modeling Variables

Modeling Variables

Cam

paig

n Ex

ecut

ion

&

Trac

king

Data Evolution• Volume Increment• Quality Expectations Mounting• Confidentiality Needs• Emergence of Custodians• Cost / Budget Pressures• Human Skills Needed

Data Foundation

Journey

Subscriber Demographics

Various Components of

Subscriber Spend

Call Center

Records

Service Request Records

Vintage of the

Subscriber

Activity / non Activity History

Subscriber Usage History

Package Type Characteristics

CRM

Billing

3P Service Providers

Data Sources and Dimensions

Snap-shot Recency Normalized

Recharging / Channel adoption

•How many times has subscriber recharged on web?

•When was the last time subscriber recharged?•Was the subscriber early adapter of web recharge?

•What is the latency of recharging?•What is the subscriber affinity towards one recharge mediums?

Product Purchase behavior

•What is current base pack?•How many Add-on packs?

•What was the product purchase behavior in first 90 days?•Were there package drops before subscriber churned?

•What is the average days subscriber uses a package?•What is the maximum no of add-on packs subscriber has used

Transforming Data to Predictive Variables

Snap-shot Recency Normalised

Demographic / Affinity

• Which class of city does subscriber come from?•Is the subscriber package different from the region (South)?

• How soon did subscriber register on portal?•Has there been relocation before churning?

•What is the average spent on Our Company?

Churn behavior•How many times do subscriber deactivates?

• Is the deactivations a recent phenomena?•Is the subscribers new to deactivation

• What is the national average of subscriber deactivations?•Ratio of Active day to total vintage?

Transforming Data to Predictive Variables

Efficiency Stage

Business View

• Oligopoly• Market Aggression • Innovate of Perish• Cost Pressures• Multiple pursuits to

same human and technology resources

Technology Eco System

• Emergence of Cloud• SaaS, PaaS, Iaas evoltion• Market Heading towards

specialist on demand • Replication Technologies• Columnar DB

alternatives emergence

Knowledge Need

• Customer Profiling• Campaign Efficiency • Churn Prediction• Inventory Models• Capacity Optimization

Saas For BI

Key Criteria• Speed to market, agility• Lack of internal expertise• Fluctuations in requirements• Disparate Set of Metadata within enterprise• Predictive Modelling

Experience

Upsides• Infrastructure and technology issues

streamlined within weeks – connectivity, instance, extraction

• Started with customer analytics and headed to predictive modelling

• Integration from and to multiple sources, Call centers, CRM System etc

• End Result – ARPU above industry avg.

Challenges

• Information Security and Data Access• Integration with heterogeneous systems• Scalability to enterprise levels• Risk Mitigations• Arbitration between multiple solution providers• Fault tolerance and Reliability• Technology Evolution

Direction

Yes

Yes Yes

No

Small Large

Application Size

Real Time

Take your time

Summary

• Define your challenges– Technological as well as business

• Take Ecosystem and Technology Paradigm in Mind

• Mastery is not achieved overnight• Journey and a pursuit for excellence is more

important than goal attainment.

QUESTIONS?

Moving to Predictive Analytics

32

Problem Definition

Validated?

Hypothesis creation

Dimension & Data Model

Data exploration

Modeling process – step by step

No

Approach

Model Deployment

Model evolution

Regeneration of Model definition

Yes Trend Analysis

Model Building & Validation

Business problem / Business Opportunity

This is where a lot of Business inputs come in from Our Company team

33

Problem Definition

Validated?

Data exploration

Modeling process – step by step

No

Approach

Model Deployment

Model evolution

Regeneration of Model definition

Yes Trend Analysis

Model Building & Validation

Model usage recommendations are provided and model is ready for roll-out

Hypothesis creation

Dimension & Data Model

Moving to Campaign management

Actionable Analytics

“ Let me find a group of people to talk about it.”

“ I have an offer …”

offer

Current Campaign Management

“ Let me find the best offer to fit this person ’s need. ”

offeroffer

offeroffer

“ Let me find the best offer to fit this person ’s need. ”

offeroffer

offeroffer

offeroffer

offeroffer

“I have a person with a change in behaviour that suggests a

need…”

Subscriber Campaign Management

Campaign Management

Let me find the best offer to fit this person ’s need.

offeroffer

offeroffer

“ offer ’ ”

offeroffer

offeroffer

offeroffer

offeroffer

“I have a person with a change in behaviour that suggests a need…”

Subscriber Campaign Management

Campaign Approach

• How do I find the “Right Offer” for the “Right Subscriber”?• How do I differentiate the subscribers based on their current status

with Our Company?• What is the order of campaign events for each of the opportunity

with subscriber?

Campaign Framework

Target Identification

Develop Rule Engine

Track & Measure Refine

Approach 1: Rule based

Approach 2:Behavioral Classification

Approach 2: Clustering

Develop the Rule engine which will define the campaign structure for each of the target segment

Track & measure the campaign effectiveness and conversion on Test & Control approach; Identify the factors effecting response uplift

Develop a process to refine the target selection & rule engine based on campaign history

Our Company

Model Data Base Creation

Cluster Profiles

Outlier Treatment

Missing Value Treatment

Multicollinearity Treatment using Factor Analysis

Variables Standardization

Cluster Solution Development & Validation

Variables across Demographic, Transaction, & Call and Service specific parameters taken into consideration

Data & Methodology

Advocate

Nomads

Overlay Segments ActionBuild Model ~

Targeting

Selection Universe

Marketing Objective

Acquire New Customers

Develop Existing Customer Relationship

Retain Customer Relationship

Revenue Growth

Customer Strategy

Gather Data

Usage/Payment Behavior

Calling Behavior

Newbies

Platinums

Offer/ Treatment

Attrition Model

RevenueRevenueComponent Component

ModelsModels

Build Model

Dynamic Pricing

Differentiated Service at call centre

Differentiated offers for each segment

Up-sell

Different creatives by segment

Reduce targeting of non profi table segments

Test different channels for communication

Reactive and Proactive Retention

Bargainers

Switch oners

Campaign Approach

Community Focused

Volunteer Driven

Knowledge Share

Accelerated Learning

Collective Excellence

Distilled Knowledge

Shared, Non Conflicting Goals

Validation / Brainstorm platform

Mentor, Guide, Coach

Satisfied, Empowered Professional

Richer Industry and Academia

About Information Excellence Group

Progress Information Excellence

Towards an Enriched Profession, Business and Society

About Information Excellence Group

Reach us at:

blog: http://informationexcellence.wordpress.com/

linked in: http://www.linkedin.com/groups/Information-Excellence-3893869

facebook: http://www.facebook.com/pages/Information-excellence-group/171892096247159

presentations: http://www.slideshare.net/informationexcellence

twitter: #infoexcel

email: [email protected]