Big Data Executive Forum Data Discovery, Modern Architecture & Visualization Driving Value From Big Data Bill Franks Chief Analytics Officer, Teradata
It s Not So Much Big Data As it is different data. The differentness of big data can be more challenging than the bigness of the data
Utilize New Analytic Disciplines Augment traditional analytic approaches with new approaches Statistics Forecasting Graph Analysis Geospatial Text Analysis
Leverage Analytics In Diverse Ways Perform discovery analysis alongside confirmatory analysis to maximize benefits Discovery Analysis Full scope not defined Interactively evolving hypotheses Business problem is developing Aim is to identify new theories Confirmatory Analysis Examining predefined problems Assessing specific hypotheses Business problem well defined Aim is to validate a theory
Historical Perspective: The Case For Data Warehousing The Problem OPERATIONAL SYSTEMS DECISION MAKERS OPERATIONAL SYSTEMS The Solution DECISION MAKERS Integrated Data Warehouse (IDW)
Today s Perspective: The Case For A Discovery Platform Data Warehouse/ Business Intelligence The Problem Advanced Analytics The Solution SQL Framework Access Layer Integrated Discovery Platform (IDP)
Examples: Behavioral Time Pattern Analysis Use cases around discovering patterns & sequences of Events of Interest & then moving insights into production. Omni Channel Customer Behavioral and Event Analysis Fraudulent Paths Fraud Pattern Sequence Detection (multi-channel) Call Center & Customer Satisfaction Analysis Sequential Process and Procedure Associated Analysis Omni Channel Event Analysis Customer Service Paths Sequential Procedure Analysis
Service Cancellation Discovery Paths Identify which actions each customer took and in which order Correlate these patterns (or pattern subsets) to churn
Comparison with Existing Churn Model Current Statistical Model Churn Potential Behavioral Path Model Current State Churn Potential Current Model Future State Churn Potential Common Prediction Incremental Churner Improved Model
Aster s Prebuilt SQL-GR Functions Leveraged Closeness Function Measures how long it will take to spread information from a given node to any other node. Local Clustering Function Local clustering coefficient quantifies how connected a node s neighbors are to each other The advantage of social network analysis is that it focuses on interaction (rather than on individual behavior)
Predicting Churn: Old and New Methods Social Network Analysis Behavioral Path Analysis Current Statistical Model Social Network Analysis Current Statistical Model Behavioral Path Analysis Old way New way
Examples: Relational Affinity Analysis Use cases concentrated on understanding relationships between data (products, services, procedures, occurrences,...) Products & Services Affinity Analysis (recommendations) Warranty Part Failure Affinity Analysis Customer Incident & IVR Affinity Analysis Online Gaming Social Network Connectivity Incident & IVR Affinity Online Gaming Social Network Connectivity Products & Services Affinity Analysis Warranty Part Affinity Analysis
Teradata Unified Data Architecture Your environment must enable any analysis against any type or volume of data at any time
Start from the Right Perspective Data Doesn t Lead You to a Business Problem... Your Business Problem Leads You to the Right Data!
To Succeed With Big Data, Start Small!
Move IT From Serving To Enabling Old Way: Server prepares You pay per cup & topping (Traditional IT Model) New Way: You prepare You pay per ounce (Discovery Model)
Join The Analytics Revolution! Discoveries are worthless if they aren t utilized & implemented Analytics is going through its own industrial revolution! Image source: Walt Stoneburner on Flickr Image source: Target
To Drive Value From Big Data Think differently Invest differently Analyze differently Act differently
If You Have Further Questions Or Comments Bill.Franks@Teradata.com www.bill-franks.com Twitter: @BillFranksGA Coming Fall 2014!