Customer
Retention in the
Airline IndustryJamesTaylor,
Decision Management Solutions
Matthew Kitching,
Apption
The Value of Predictive
Analytics and How
Using Decision
Modeling Helps You
Succeed
Your Presenters
James Taylor
We work with clients to
improve their business by
identifying and modeling
decisions, and applying
business rules and analytic
technology to automate &
improve these decisions.
Spent 12 years championing
Decision Management.
DMN Submitter
Matthew Kitching
We help our clients leverage
their data for better
decision-making – analyzing
large data sets allows us to
build robust predictive
models which you can embed
in your operations.
Developers of analytics
solutions since 2007.
Senior Data Scientist
Agenda
Introductions
Case Study:Airline Customer Retention
An Approach for Big Data Analytics
Framing Requirements
Data Preparation
Data Exploration
Predictive Modeling
Reporting and Evaluation
Deliverables
Wrap
© Decision Management Solutions, 2014 3
Customer Retention in the Airline Industry
Goal
Retain valued customers to
maximize profit
Candidate Predictions
Likelihood of churn
Customer lifetime value
Customer response
Based on an Apption engagement
with a major airline
Apption Big Data Analytics Workflow
Data
Exploration
Predictive
Models
Reporting
and
Evaluation
Data
Preparation
Design Implementation Delivery
Deliverables
Requirements
Gathering and
Design
Decision Models Add Value Throughout
©2015 Decision Management Solutions 6@jamet123 #decisionmgt
Data
Exploration
Predictive
Models
Reporting
and
Evaluation
Data
Preparation
Requirements
Gathering and
Design
Deliverables
Design Implementation Delivery
©2015 Decision Management Solutions 7
Framing Analytics Methodology
@jamet123 #decisionmgt
Design specifications
Data governance strategy
Validation and success criteria
Business goals
Business process model
Data
Exploration
Predictive
Models
Reporting
and
Evaluation
Data
Preparation
Requirements
Gathering and
Design
Deliverables
Design Implementation Delivery
©2015 Decision Management Solutions 8
"This is the critical path to monetizing
advanced models."
Head of Analytics
"Decision modeling enables us to model our
business by dividing it into concrete parts
that are understandable to business people
without being too detailed.“
Process Director
@jamet123 #decisionmgt
"What used to be one week
of requirements work was
done in a few hours.“
Lead Business Analyst
Airline Customer Retention – High Level
©2015 Decision Management Solutions 9@jamet123 #decisionmgt
Airline Customer Retention – Churn
©2015 Decision Management Solutions 10@jamet123 #decisionmgt
Information
Knowledge
Decision
©2015 Decision Management Solutions
Power of Decision Requirements Models
Precise
Definition
Identify decision to be
improved
Uses a non-technical
notation
@jamet123 #decisionmgt 11
Why Frame Predictive Analytic Projects?
Provides structure (who, what, how, when)
Provides transparency of decision process
Promotes buy in
Fosters innovation
Standardizes approach to decision making
Provides an audit trail for decisions
Improves/changes the business model
Steve Knode, University of Maryland University College
©2015 Decision Management Solutions 12@jamet123 #decisionmgt
Apption Big Data Analytics
Data
Exploration
Predictive
Models
Reporting
and
Evaluation
Data
Preparation
Requirements
Gathering and
Design
Deliverables
Design Implementation Delivery
Step 2 - Data Preparation
•Hadoop Infrastructure Setup
•Data assessment and consolidation
•Cleanse data
de-duplication
de-identification
unstructured text processing
At this point, the data is ready to be analyzed
Airline Case Study – Data Preparation
What we learned from the data:
• Shorter than expected timeline
for survey data
• Impact of omitting customer
survey results can be visualized
• Decision Model can be updated
with information about the data
sources
We can adjust our assumptions after analyzing the data
Apption Big Data Analytics
Data
Exploration
Predictive
Models
Reporting
and
Evaluation
Data
Preparation
Requirements
Gathering and
Design
Deliverables
Design Implementation Delivery
Step 3 - Data Exploration
•Identify actionable insights from data:
• statistics about data features
• correlations between features
• aggregation of data
• creation of new features
•Convert into a visual or tabular format
•Data Requirements Models focus data
scientists on most relevant data
Airline Case Study - Data Preparation
Results found provided interesting and surprising insights:
• Useful positive or negative indicators for predicting churn
• Surprisingly not useful indicators for predicting churn
Apption Big Data Analytics
Data
Exploration
Predictive
Models
Reporting
and
Evaluation
Data
Preparation
Requirements
Gathering and
Design
Deliverables
Design Implementation Delivery
Step 4 - Predictive Models
Data science and technology at work:
• Algorithms: Segmentation, classification, clustering,
regression…
• Technologies: Hadoop, Spark, Python, R, SAS…
A data asset is created that can be reused over time
Airline Case Study - Predictive Models
Update the Decision Model based on the results:
• Original definition of churn did not lead to a stable model
• Many passengers who churned in year 1 did not in year 2
• No correlation between lost baggage claims and churn
The Decision Requirements Model to be updated
Airline Case Study – Churn Model results
98%
2%
Customers identified as low risk of
churn based on year 1 data
Customers did not churn in year 2 Customers did churn in year 2
49%51%
Customers identified as high risk of
churn based on year 1 data
Customers did not churn in year 2 Customers did churn in year 2
Low-churn group shows model accuracy
High-churn group identifies target market
Step 5 - Reporting and Evaluation
Data Exploration
• statistics
• correlations
Predictive Models
• performance versus
success criteria
Further iterations of the Implementation cycle
based on the results obtained
Report on the results obtained in the previous steps:
Data
Exploration
Predictive
Models
Reporting
and
Evaluation
Data
Preparation
Deliverables
Implementation Delivery
©2015 Decision Management Solutions 24
Business Case and Project Comparison
Analytic Decisions
KPIs
Processes
and Systems
Organizations
Monitoring
@jamet123 #decisionmgt
Apption Big Data Analytics
Data
Exploration
Predictive
Models
Reporting
and
Evaluation
Data
Preparation
Requirements
Gathering and
Design
Deliverables
Design Implementation Delivery
step 6 - Finalize•New reusable software asset is deployed
•Knowledge transfer allows the business to integrate
this asset in their enterprise processes
•Potential roadmap for evolution of the asset
•Reports
Visualizations
Actionable insights
Predictive model results
Decision Requirements Model
Step 6 - Deliverables
Questions?
Takeaways
©2015 Decision Management Solutions 29
Lessons Learned
Decision Models frame and communicate
analytic requirements accurately
Decision Models are accessible to all teams
involved, building shared understanding
Decision models show how analytics will
add value and deliver business impact
The requirements for deployment and
usage are clear
@jamet123 #decisionmgt
Big Data Lessons Learned
Strong Case for Big Data Analytics
Big Data Analytics extracts actionable insights
from unstructured and often messy data
Meaningful actionable insights can be
achieved within a reasonable amount of time
Big Data Analytics assets created allow
ongoing insights as data changes
©2015 Decision Management Solutions 31
Decision Management Solutions
Providing Decision Management consulting
and training since 2009
DMN Submitter
BABOK ® Contributor
IIBA Endorsed Education Provider
Recognized experts in decision management
Services
Free Resources
Consulting
Decision Modeling Software and Services
Training and Workshops
http://decisionmanagementsolutions.com
info@decisionmanagementsolutions.com
@jamet123 #decisionmgt
DecisionsFirst Modeler
A collaborative decision modeling software that
conforms to the new Decision Model and Notation
(DMN) standard.
©2015 Decision Management Solutions 32
DecisionsFirst Modeler is available as a free BasicVersion (SaaS) and a paid
Enterprise Edition (SaaS or on-premise). Sign up at www.decisionsfirst.com
@jamet123 #decisionmgt
Apption
Data Science, Data Management and Analytics Software
Development and Consulting Experts
Founded in 2004
Full Stack Big Data Analytics Services
Data Engineering
Data Science and Analytics experience
Data Visualization
Custom Software Development
Focus on Security Analytics and Customer Intelligence
Website: http://www.apption.com
Email: info@apption.com
Thank You
JamesTaylor
james@decisionmanagementsolutions.com
Matthew Kitching
matt@apption.com

The Value of Predictive Analytics and Decision Modeling

  • 1.
    Customer Retention in the AirlineIndustryJamesTaylor, Decision Management Solutions Matthew Kitching, Apption The Value of Predictive Analytics and How Using Decision Modeling Helps You Succeed
  • 2.
    Your Presenters James Taylor Wework with clients to improve their business by identifying and modeling decisions, and applying business rules and analytic technology to automate & improve these decisions. Spent 12 years championing Decision Management. DMN Submitter Matthew Kitching We help our clients leverage their data for better decision-making – analyzing large data sets allows us to build robust predictive models which you can embed in your operations. Developers of analytics solutions since 2007. Senior Data Scientist
  • 3.
    Agenda Introductions Case Study:Airline CustomerRetention An Approach for Big Data Analytics Framing Requirements Data Preparation Data Exploration Predictive Modeling Reporting and Evaluation Deliverables Wrap © Decision Management Solutions, 2014 3
  • 4.
    Customer Retention inthe Airline Industry Goal Retain valued customers to maximize profit Candidate Predictions Likelihood of churn Customer lifetime value Customer response Based on an Apption engagement with a major airline
  • 5.
    Apption Big DataAnalytics Workflow Data Exploration Predictive Models Reporting and Evaluation Data Preparation Design Implementation Delivery Deliverables Requirements Gathering and Design
  • 6.
    Decision Models AddValue Throughout ©2015 Decision Management Solutions 6@jamet123 #decisionmgt Data Exploration Predictive Models Reporting and Evaluation Data Preparation Requirements Gathering and Design Deliverables Design Implementation Delivery
  • 7.
    ©2015 Decision ManagementSolutions 7 Framing Analytics Methodology @jamet123 #decisionmgt Design specifications Data governance strategy Validation and success criteria Business goals Business process model Data Exploration Predictive Models Reporting and Evaluation Data Preparation Requirements Gathering and Design Deliverables Design Implementation Delivery
  • 8.
    ©2015 Decision ManagementSolutions 8 "This is the critical path to monetizing advanced models." Head of Analytics "Decision modeling enables us to model our business by dividing it into concrete parts that are understandable to business people without being too detailed.“ Process Director @jamet123 #decisionmgt "What used to be one week of requirements work was done in a few hours.“ Lead Business Analyst
  • 9.
    Airline Customer Retention– High Level ©2015 Decision Management Solutions 9@jamet123 #decisionmgt
  • 10.
    Airline Customer Retention– Churn ©2015 Decision Management Solutions 10@jamet123 #decisionmgt Information Knowledge Decision
  • 11.
    ©2015 Decision ManagementSolutions Power of Decision Requirements Models Precise Definition Identify decision to be improved Uses a non-technical notation @jamet123 #decisionmgt 11
  • 12.
    Why Frame PredictiveAnalytic Projects? Provides structure (who, what, how, when) Provides transparency of decision process Promotes buy in Fosters innovation Standardizes approach to decision making Provides an audit trail for decisions Improves/changes the business model Steve Knode, University of Maryland University College ©2015 Decision Management Solutions 12@jamet123 #decisionmgt
  • 13.
    Apption Big DataAnalytics Data Exploration Predictive Models Reporting and Evaluation Data Preparation Requirements Gathering and Design Deliverables Design Implementation Delivery
  • 14.
    Step 2 -Data Preparation •Hadoop Infrastructure Setup •Data assessment and consolidation •Cleanse data de-duplication de-identification unstructured text processing At this point, the data is ready to be analyzed
  • 15.
    Airline Case Study– Data Preparation What we learned from the data: • Shorter than expected timeline for survey data • Impact of omitting customer survey results can be visualized • Decision Model can be updated with information about the data sources We can adjust our assumptions after analyzing the data
  • 16.
    Apption Big DataAnalytics Data Exploration Predictive Models Reporting and Evaluation Data Preparation Requirements Gathering and Design Deliverables Design Implementation Delivery
  • 17.
    Step 3 -Data Exploration •Identify actionable insights from data: • statistics about data features • correlations between features • aggregation of data • creation of new features •Convert into a visual or tabular format •Data Requirements Models focus data scientists on most relevant data
  • 18.
    Airline Case Study- Data Preparation Results found provided interesting and surprising insights: • Useful positive or negative indicators for predicting churn • Surprisingly not useful indicators for predicting churn
  • 19.
    Apption Big DataAnalytics Data Exploration Predictive Models Reporting and Evaluation Data Preparation Requirements Gathering and Design Deliverables Design Implementation Delivery
  • 20.
    Step 4 -Predictive Models Data science and technology at work: • Algorithms: Segmentation, classification, clustering, regression… • Technologies: Hadoop, Spark, Python, R, SAS… A data asset is created that can be reused over time
  • 21.
    Airline Case Study- Predictive Models Update the Decision Model based on the results: • Original definition of churn did not lead to a stable model • Many passengers who churned in year 1 did not in year 2 • No correlation between lost baggage claims and churn The Decision Requirements Model to be updated
  • 22.
    Airline Case Study– Churn Model results 98% 2% Customers identified as low risk of churn based on year 1 data Customers did not churn in year 2 Customers did churn in year 2 49%51% Customers identified as high risk of churn based on year 1 data Customers did not churn in year 2 Customers did churn in year 2 Low-churn group shows model accuracy High-churn group identifies target market
  • 23.
    Step 5 -Reporting and Evaluation Data Exploration • statistics • correlations Predictive Models • performance versus success criteria Further iterations of the Implementation cycle based on the results obtained Report on the results obtained in the previous steps: Data Exploration Predictive Models Reporting and Evaluation Data Preparation Deliverables Implementation Delivery
  • 24.
    ©2015 Decision ManagementSolutions 24 Business Case and Project Comparison Analytic Decisions KPIs Processes and Systems Organizations Monitoring @jamet123 #decisionmgt
  • 25.
    Apption Big DataAnalytics Data Exploration Predictive Models Reporting and Evaluation Data Preparation Requirements Gathering and Design Deliverables Design Implementation Delivery
  • 26.
    step 6 -Finalize•New reusable software asset is deployed •Knowledge transfer allows the business to integrate this asset in their enterprise processes •Potential roadmap for evolution of the asset •Reports Visualizations Actionable insights Predictive model results Decision Requirements Model Step 6 - Deliverables
  • 27.
  • 28.
  • 29.
    ©2015 Decision ManagementSolutions 29 Lessons Learned Decision Models frame and communicate analytic requirements accurately Decision Models are accessible to all teams involved, building shared understanding Decision models show how analytics will add value and deliver business impact The requirements for deployment and usage are clear @jamet123 #decisionmgt
  • 30.
    Big Data LessonsLearned Strong Case for Big Data Analytics Big Data Analytics extracts actionable insights from unstructured and often messy data Meaningful actionable insights can be achieved within a reasonable amount of time Big Data Analytics assets created allow ongoing insights as data changes
  • 31.
    ©2015 Decision ManagementSolutions 31 Decision Management Solutions Providing Decision Management consulting and training since 2009 DMN Submitter BABOK ® Contributor IIBA Endorsed Education Provider Recognized experts in decision management Services Free Resources Consulting Decision Modeling Software and Services Training and Workshops http://decisionmanagementsolutions.com info@decisionmanagementsolutions.com @jamet123 #decisionmgt
  • 32.
    DecisionsFirst Modeler A collaborativedecision modeling software that conforms to the new Decision Model and Notation (DMN) standard. ©2015 Decision Management Solutions 32 DecisionsFirst Modeler is available as a free BasicVersion (SaaS) and a paid Enterprise Edition (SaaS or on-premise). Sign up at www.decisionsfirst.com @jamet123 #decisionmgt
  • 33.
    Apption Data Science, DataManagement and Analytics Software Development and Consulting Experts Founded in 2004 Full Stack Big Data Analytics Services Data Engineering Data Science and Analytics experience Data Visualization Custom Software Development Focus on Security Analytics and Customer Intelligence Website: http://www.apption.com Email: info@apption.com
  • 34.