Marketing Analytics

How to make the most effective use of advanced Analytics Techniques

A Briefing for Professional Marketers

Chair
Dougy Watt, Head of Retail Analysis, RBS

Speakers
Dr Judy Bayer, Director of Advanced Business Analytics, Teradata EMEA
Tom Breur, Principal, XLNT Consulting
Dr Stuart Clarke, Senior Consultant, KXEN
Gordon Farquharson, Director, Streetwise Analytics Ltd
David Harris, Director of Statistics, CACI Marketing Solutions
Corrine Moy, Director of Marketing Sciences, GfK NOP
John Tansley, Senior Consultant, CACI Customer Analysis Group
Francesco Vivarelli, Head of Data Analysis and Consulting, Acxiom
Programme

Panelist
Martin Squires, Head of Customer Intelligence & Analysis, M & S Money

Advanced Analytics and Data Mining
• A brief introduction to the techniques that will be covered during the day, explaining how they fit together, descriptive vs. predictive techniques, classical methods vs. artificial intelligence, use of multiple techniques and some dos and don’ts.

Multivariate Analysis for Segmentation
Key techniques: Factor Analysis, Cluster Analysis
• Why segment?
- Business benefits of segmentation
• Key issues in segmentation
- Define objective of segmentation
- Identification of base for segmentation (people vs businesses vs occasions, global vs local)
- Identification and development of input variables (needs, behaviours, demographics)
- Selection of appropriate segmentation approach
- Analytic issues (stability testing, outliers etc)
- Creating allocation models for databases
• Segmentation techniques – pros and cons
- K-Means
- Latent Class
- Hierarchical Cluster Analysis
- Two-Step
- CHAID
- Canonical segmentation
- New approaches – Genetic Algorithms, Neural Network approaches (self-organising maps etc)
• Implementation of segmentation

Decision Trees for Response Modelling
Key technique: Decision Tree Analysis
• Characteristics of decision trees:
- Applicable business problems
- Main decision tree concepts
- Use for segmentation, data exploration, visualisation
• Learning decision trees from data:
- Learning algorithms: CHAID, Gini, C5.0, CART
- Bayesian decision tree learning: Use for variable selection
• Comparison with Logistic Regression – advantages and drawbacks of both approaches

Regression Analysis for Lifetime Value Prediction
Key technique: Multiple Regression
• Importance of Regression Analysis
- Business problems that can be solved
• Key principles and features of Regression Analysis
- Measuring associations using correlations
- Building simple linear regression models
- Extension to many variables – Multiple Regression
- Selecting and transforming appropriate variables
- Assessing the model – Will it support its intended application?
• Case Study: Predicting Future Customer Value
• Alternative method – Survival Analysis

Logistic Regression for Churn Modelling
Key technique: Logistic Regression
• Addressing business problems using Logistic Regression
• Concepts of Logistic Regression
• How to apply Logistic Regression for best effect
• Case study example
- Data preparation
- Preliminary analysis
- Model development
- Interpretation of results
• New approach to classification modelling – Structured Risk Minimisation

Affinity Analysis for selecting ‘Next Best Activity’
Key techniques: Association and Sequence Analysis
• Importance of next best offer for a customer
• Principles of affinity analysis techniques
• Descriptive analysis to explain how customers move through the product assortment
• Predictive application to planning customer purchase paths
• Alternative methods

Data Integration Techniques for Database Enrichment
Key techniques: Imputation, Data Fusion
• Enhancing database quality - how imputation methods can be used to improve confidence in the data
- Review of imputation techniques, from simple to advanced
• Using market research and other data to derive new variables
- Model based methods
- Data fusion
• Example applications of the enriched database
• Measuring the cost-benefit trade off from enriching your database

Techniques for Price Optimisation
Key technique: Price Elasticity Modelling
• The challenges of optimal pricing
• Principles and main features of price elasticity models
• The Analytics of optimal end-of-life pricing decisions
• The Analytics of optimal new product pricing decisions
• Discovery-based cross-sell and cannibalisation analysis and modelling
• Availability-based Affinity Analysis (ABAA) to support pricing
• Discovering how much customers will pay for new service features
• Market research based approaches to optimal pricing
• Case study of using price optimisation – an Italian retailer

Neural Networks for Time Series Forecasting
Key technique: Supervised Neural Network
• Introduction to Neural Networks
- First principles
- Main network types – supervised vs. unsupervised
- Business problems that neural networks can solve
- Strengths vs. weaknesses, when to use, which to use
• Neural Network applications – contrasts with more traditional approaches
- Time series forecasting
• Case study on predicting customer attrition with neural networks