Wednesday February 18, 2009
8:00am-9:00am
Registration & Breakfast
Nikko Ballroom Foyer
9:00am-9:50am
Nikko II
Keynote
Five Ways to Lower Costs with Predictive Analytics
Question: How does predictive analytics actively deliver increased returns? Answer: By driving operational decisions with predictive scores – one score assigned to each customer. In this way, an enterprise optimizes on what customers WILL do.
But, in tough times, our attention turns away from increasing returns, and towards decreasing costs. On top of boosting us up the hill, can predictive analytics pull us out of a hole? Heck, yes. Marketing more optimally means you can market less. Filtering high risk prospects means you will spend less. And, by retaining customers more efficiently, well, a customer saved is a customer earned – and one you need not acquire.
In this keynote, Eric will demonstrate five ways predictive analytics can lower costs without decreasing business, thus transforming your enterprise into a Lean, Mean Analytical Machine. You'll want to run back home and break the news: We can't afford not to do this.
9:50am-10:10am
Nikko II
Platinum Sponsor Presentation: SAS
Analytics: The Art and Science of Better
Better is a relative term. Analytics is the art and science of effecting better-making the invisible visible, aiding judgment under uncertainty, allowing us to continuously learn and improve. This talk advocates science-and creativity-for analytics to have greater impact.
10:10am-10:30am
Morning Coffee Break – Exhibit Hall Open from 10:10am to 7:30pm
Nikko Ballroom Foyer
Delegates may choose to attend either session at this time.
Track 1: Thought Leaders
10:30am-11:20amNikko II
From Analytics to Competing on Decisions
Seeking out increasingly small margins has brought analytics into vogue. Organizations realize that they can gain analytic insights customers, products, channels, partners and much more. But some companies are already finding that analytics is only a part of the process – the intelligent application of the findings of these new insights can only pay off if the decisions that are made are correct.
Translating analytics into better decisions requires a new conceptual framework enterprise decision management. By applying this framework and becoming decision-centric, by using business rules to control those decisions and by leveraging analytics companies are increasingly Competing on Decisions.
|
James Taylor, CEO, Decision Management Solutions |
Track 2: Verticals
10:30am-11:20amNikko I
Case Study: MetLife Auto & Home(r)
Identifying Fraud with Predictive Analytics
Organizations are realizing they must become more competitive by using Predictive Analytics to improve one-to-one customer relationships – across any aspect of their business. Specifically, MetLife is embedding Predictive Analytics within their claims handling process to assist with uncovering instances of fraud and abuse. By seamlessly integrating Predictive Analytics data mining and text mining software with business rules and identification validation, MetLife has enhanced their ability to detect fraudulent claims at each stage of the claim lifecycle. Scoring claims as early in the process as possible is critical so they can improve a claim adjuster's productivity and save customers time and money. This session will illustrate how organizations define a problem, decide on appropriate metrics, work cross-functionally across the organization and deploy the results of Predictive Analytics to the business user to optimize the overall process.
|
Kathy Konkel, Manager, Product Marketing, SPSS Inc. David McMichael, Ph.D., Assistant Vice President, MetLife Auto & Home |
11:20am-12:30pm
Nikko II
Multiple Case Studies: Anheuser-Busch, Disney, HSBC, Pfizer, and others
The High ROI of Data Mining for Innovative Organizations
Data mining and advanced analytics can enhance your bottom line in three basic ways, by 1) streamlining a process, 2) eliminating the bad, or 3) highlighting the good. In rare situations, a fourth way – creating something new – is possible. But modern organizations are so effective at their core tasks that data mining usually results in an iterative, rather than transformative, improvement. Still, the impact can be dramatic.
Dr. will share the story (problem, solution, and effect) of nine projects conducted over the last decade for some of America's most innovative agencies and corporations:
- Streamline:
- Cross-selling for HSBC
- Image recognition for Anheuser-Busch
- Biometric identification for Lumidigm (for Disney)
- Optimal decisioning for a leading high-tech retailer
- Quick decisions for the Social Security Administration
- Eliminate Bad:
- Tax fraud detection for the IRS
- Warranty Fraud detection for a leading high-tech retailer
- Highlight Good:
- Sector trading for WestWind Foundation
- Drug efficacy discovery for Pharmacia & UpJohn (now Pfizer)
John F. Elder, Ph.D., CEO and Founder of Elder Research, Inc.
12:30pm-1:45pm
Birds of a Feather Lunch
Nikko I
Find your clan of like-minded colleagues with whom to dine and discuss, comparing your organization's stories and challenges.
- Business Applications and Analytical Process:
- Response modeling
- Churn modeling
- Product recommendations
- Online/Internet applications
- SAS Sponsored Topics:
- Getting from Data to Mining: Best Practices in Data Preparation
- Effective Strategies for Deploying and Monitoring Models
1:45pm-2:35pm
Nikko II
Keynote
New Challenges in Predictive Analytics: Social Networking, Direct-Response Marketing, and Understanding Customer Behavior
The rise of the interactive media represented by web, social media, search and behavioral targeting have created new challenges and opportunities for predictive analytics. While these new media offer a deeper mechanism for approaching the holy grail in marketing and advertising – understanding the customer's intent – the rich structure of available data, from social graph data to time-series from interactions to reputation and other behavioral traits, expand the complexity of prediction in dimensions where we have little experience and a poor understanding of the terrain.
Dr. Fayyad will present examples of such applications as well as challenges, and will relate some case-studies to illustrate the power of understanding and harnessing this data. However, the context will also be used to illustrate the stronger need to build up new sciences to help us better understand these new powerful dimensions. Building up the science underlying these new capabilities is necessary for the future of these new media and the success of predictive and descriptive analytics in these new fields.
| Usama Fayyad, Ph.D., CEO, Open Insights |
2:35pm-2:45pm
Gold Sponsor Presentation: SAP
Delivering Intelligence in Busines Intelligence – The I in BI
Business intelligence has traditionally provided insight into what has happened and what is currently happening. But how do you make good decisions and run your business when you have no insight into what might happen? Learn how critical it is for your business to combine the power of predictive analysis with a complete business intelligence infrastructure to stay competitive and succeed in today's market. See how to provide relevant predictive analytics to the business “BI User”. It's not traditional predictive analysis – it's analysis for the business BI end user.
2:45pm-2:55pm
Session Break – Exhibit Hall
Nikko Ballroom Foyer
Delegates may choose to attend either session at this time.
Track 1: Product recommendations and the Netflix Prize
2:55pm-3:45pmNikko II
Case Study: Netflix Prize (leading competitor and winner of the 2008 Netflix Progress Prize)
Advanced Approaches for Recommender Systems and the Netflix Prize
This presentation discusses a set of methods that form the basis for the recommender system that holds first place for the Netflix Prize competition. Research has shown that common methods like item – or user-based collaborative filtering are not ideal for either prediction accuracy or speed.
In fact, a combination of different approaches is necessary for accurate predictions within the Netflix competition. We also discuss recommendation systems beyond our Netflix solution, for which predictive models may be customized to integrate a large variety of implicit or explicit user inputs, e.g., ratings, clicks, and page views.
| Andreas Töscher, Research & Technical Development, Commendo Research and Consulting, GmbH |
Track 2: Verticals
2:55pm-3:45pmNikko I
Healthcare
Case Study: San Diego Supercomputer Center
High Performance Scoring of Healthcare Data
Predictive analytics enables the health care industry to accelerate critical decisions, automate complex data analysis, and ultimately to improve patient care. This session will highlight how data mining of large data sets combined with agile deployment of predictive models will help providers to effectively integrate intelligent decisions while managing the pressure to reduce the overall cost of healthcare. Our interdisciplinary approach leverages the leading-edge research cyberinfrastructure at the San Diego Supercomputer Center (SDSC) and is based on a collaboration between SDSC and Zementis.
|
Natasha Balac, Ph.D., Data Applications Group Manager, San Diego Supercomputer Center Michael Zeller, Ph.D., CEO, Zementis |
Delegates may choose to attend either session at this time.
Track 1: Product recommendations and the Netflix Prize
3:45pm-4:30pmNikko II
Visualization and the Netflix Prize
Two of the most popular recommendation techniques—user-oriented and item-oriented collaborative filtering—have underperformed in the Netflix Prize competition. This session will look at visualizations which can help us understand the behavior and shortcomings of these techniques, as well as provide insight into the movie ratings dataset itself. We will look at the creation of these visualizations, and in particular the challenges presented by the very large scale of the dataset. Finally, we will briefly discuss visualization in other recommender system domains.
Track 2: Verticals
3:45pm-4:30pmNikko I
Consumer Services (Online)
Case Study: TaxBrain
Completing the Visitor Targeting Cycle: Predictive Behavioral Targeting and Audience Maps
Behavioral Targeting means many things to different people. Joshua Koran will review the evolution of behavioral targeting from simple retargeting to predictive algorithms that adapt segments to marketers' goals, in addition to how online marketing is benefiting from the integration of performance-driven reporting solutions that enable marketers to more easily navigate the high-dimensional landscape of online targeting. TaxBrain senior marketing manager Thomas Rose-Bolden will present a real-world case study demonstrating how predictive BT and visual data analysis was used to identify key audience segments that improved their campaign performance.
|
Thomas Rose-Bolden, Senior Marketing Manager, TaxBrain Joshua Koran, VP, Targeting and Optimization, ValueClick |
4:30pm-4:55pm
Afternoon Coffee Break – Exhibit Hall
Nikko Ballroom Foyer
Delegates may choose to attend either session at this time.
Track 1: Product recommendations and the Netflix Prize
4:55pm-5:40pmNikko II
It's the Data, Stupid!
Many sophisticated machine-learning and data mining algorithms (e.g., support-vector machines, singular-value methods) don't scale well to really large datasets. In such cases, a common approach is to reduce the data set to a manageable size through sampling. Anand's experience, validated by many real-world examples such as Search, Ad Targeting, and the Netflix Challenge, proves that it's often better to use really simple algorithms to analyze really large datasets, rather than complex algorithms that can only work with smaller datasets. In other words, more data usually beats better algorithms. In this talk Anand will explain this assertion and provide examples of its application.
Track 2: Verticals
4:55pm-5:40pmNikko I
Non-Profit
Case Study: National Rifle Association
How to Improve Customer Acquisition Models with Ensembles
Model ensembles have been used increasingly by data mining practitioners to increase model accuracy and routinely appear as top performers in predictive modeling competitions. However, they also provide significant robustness advantages over single models, particularly when data is noisy and has significant levels of uncertainty.
Insightful Miner was used to create an ensemble of models to improve the ROI of a direct marketing campaign to members of the National Rifle Association. This session will highlight a systematic methodology to use in building bootstrapped ensembles of logistic regression models, similar to the Bagging approach first advocated by Leo Brieman. The model ensembles are more reliable and out perform single models by exploring the results of applying both modeling techniques to the NRA membership file.
Delegates may choose to attend either session at this time.
Track 1: Product recommendations and the Netflix Prize
5:40pm-6:30pmNikko II
Marketing Operations: The Engine Behind Predictive Analytics
The holistic application of predictive analytics to create a smarter, more agile enterprise is unquestionably attractive, even sexy. But predictive analytics cannot fulfill its considerable promise without a strong Marketing Operations (MO) organization behind it to give it direction, meaning and opportunities for practical application.
What the heck is Marketing Operations and why should I, as an astute modeler and analyst, give a damn? While the definition and perceived value of MO varies from one organization (and one CMO) to another, one truth is self-evident: Those organizations that choose to ignore the need for MO will continue to operate in functional silos that constrain their ability to be proactive, consistent, scalable and to deliver sustainable results.
On the other hand, those serious Marketing organizations that embrace a new MO can overcome unmanageable complexity and the perception of being a cost center to deliver Marketing excellence. By leveraging process, technology, guidance, metrics and powerful tools like predictive analytics, Marketing Operations enables truly integrated and aligned Marketing that serves the enterprise as a strategic asset, a growth driver, an accountable business function and a navigator of change.
Track 2: Verticals
5:40pm-6:30pmNikko I
Publishing
Case Study: Reed Elsevier
Profitably increasing retention – and convincing the business that you are
A high level walk-through of a retention modelling project undertaken for the publisher Reed Elsevier. Focussing on the process, approach and outcomes. With a particular emphasis on the collaboration with the Subject Matter Expert and how he, in turn, planned the deployment and convinced other stakeholders of its effectiveness.
6:30pm-7:30pm
Reception Sponsored by SAS
Nikko Ballroom Foyer
7:30pm-9:00pm
Nikko II
pRediction – A quick survey of prediction methods in R 
Meeting held by the use R Group of San Francisco Bay Area
Panel Discussion - The R and Science of Predictive Analytics: Four Case Studies in R
- Bo Cowgill, Google
- Itamar Rosenn, Facebook
- David Smith, Revolution Computing
- Mike Driscoll, Dataspora
- Jim Porzak, The Generations Network
They will begin with an overview of prediction methods in R, and each of the panelists will discuss a case study of how R is used within their organizations. They will follow with a question and answer session for the panelists.
For existing R users, this panel will show how R is being used for real world problems. For analysts who may not be familiar with R, they will illustrate R's features and demonstrate its value as a business tool.
Audience participation encouraged!
Thursday February 19, 2009
8:00am-9:00am
Registration & Breakfast
Nikko Ballroom Foyer
Delegates may choose to attend either session at this time.
Track 1: Financial Services and Insurance
9:00am-9:50amNikko II
Case Study: Wells Fargo Internet Services Group
New Challenges for Developing Predictive Analytics Solutions
The explosive growth of data combined with the ever increasing need for analyzing this data has introduced new challenges in predictive analytics. Some business problems benefit from analytics solutions that require large number of predictive models or large number of variables. In some cases, analysis of network data resulting from the interactions among customers and free text data can enhance the lift performance of the models. In addition, predictive models should be immediately deployable in target environments where the data resides. Organizations may desire their traditional BI analysts to perform a portion of predictive analytics tasks using easier-to-use tools. To face these challenges, appropriate infrastructure, tools, processes, and skills will be required. A few real-world examples will illustrate the extent of these challenges and the need for better tools/methodologies. A case study focuses on challenges faced by Wells Fargo Internet Services Group in incorporating predictive analytics into their online marketing services covering data collection through scoring.
|
Khosrow Hassibi, Ph.D., Senior Technical Director, KXEN William Tangalos, ISG (Internet Services Group), Wells Fargo |
Track 2: Scoring Sales Leads
9:00am-9:50amNikko I
Case Study: Sun Microsystems
Marrying Prediction and Segmentation to Drive Sales Leads
Sun Microsystems open-sourced their Solaris operating system in 2005 and provides free downloads for use on Sun SPARC and Sun x86 as well as non-Sun x86 hardware systems. The challenge is monetizing that effort! Sun used a combination of predictive analytics and prospect segmentation methods to put the mechanisms in place to more efficiently identify high value sales prospects.
The case study will cover:
- Describing the business challenge
- Building a purchase model using random forests
- What the purchase model told us
- Creating prospect persona segmentation using cluster analysis
- Learnings from the persona segments
- Combined impact on marketing strategy and tactics
|
Jim Porzak, Senior Director of Marketing Analytics, The Generations Network Alex Kriney, Distinguished Marketing Director, Lifecycle Marketing, Sun Microsystems |
Delegates may choose to attend either session at this time.
Track 1: Financial Services and Insurance
9:50am-10:40amNikko II
Case Study: Wells Fargo Card Services
Customer Marketing – Predictive Modeling & Today's Growing Data Challenges
Many marketers are unclear regarding what it takes to begin predictive modeling, the steps involved in model development, and the struggles encountered in creating models that will meet corporate & regulatory compliance guidelines. This session will include relevant case studies co-presented by a partner at Wells Fargo.
Attendees will better understand:
- The value of predictive modeling
- How modeling can aid in meeting objectives
- The initial investment required
- The steps involving in sample creation and model development
- Implementation and ongoing validation
- Challenges evolving in the marketplace
|
Matt Kramer, Global Consulting and Analytics Group, Acxiom Jun Zhong, VP Targeting and Analytics, Card Services Customer Marketing, Wells Fargo |
Track 2: Scoring Sales Leads
9:50am-10:40amNikko I
Case Study: 
Scoring Leads from Salesforce.com Data
The business objective was to rank potential brides' propensity to set an appointment to discuss wedding photography. The resulting scorecard was deployed in the operational CRM systems, Salesforce.com and Eloqua.
This project used traditional data preparation and model estimation techniques. The model had to be compatible with the operational scoring systems. The data prep, estimation, and model were built with a strong collaboration between consultant and business stakeholders.
|
Traci Chu, Director CRM, Bella Pictures John Wallace, Principal Consultant & Founder, Business Researchers |
10:40am-11:00am
Morning Coffee Break – Exhibit Hall Open from 10:40am to 4:15pm
Nikko Ballroom Foyer
Keynote
The Unrealized Predictive Power of Data
Technology affords companies unprecedented opportunities to interact with customers and employees. In any of these interactions, data is created. Yet most firms neither capture nor fully utilize those data to impact their bottom line and strengthen relationships with their customers. Product recommendations and behavioral targeting are early examples of leveraging new sources of data to predict customer behavior and preferences. The next iteration of these interactions, for example mobile phones, empowers owners to access richer data and discover new opportunities – with the possible inclusion of location data that enables companies to predict mobility patterns for marketing and planning purposes. Learn from the former Chief Scientist of Amazon.com how to create a comprehensive data strategy through:
- Leveraging the data you're already collecting, but not using
- Identifying data that you could and should be collecting
- Transforming data to next-generation predictive intelligence
Andreas S. Weigend, Ph.D., people & data, and former Chief Scientist, Amazon.com
11:50am-12:00pm
Nikko II
Gold Sponsor Presentation: WNS
Leveraging Global Analytical Resources in a Challenging Economic Environment
From corporate strategy to product planning, from engineering to marketing and sales, most successful companies have learned that predictive analytics is essential for success in today's marketplace. But, very few companies are satisfied with the extent to which they leverage predictive analytics. They all know that they can get more out of every dollar of investment in product development, marketing or simply working capital employed along their supply chain, if only they had more predictive analytics to support decision making. Leveraging global resources to create an analytics Center of Excellence enables a company to tap low cost, high quality resources, while standardizing analytical tools and sharing insight across their organization. Leading companies have already embarked upon this journey, creating the world's largest pool of trained analytics professionals in India. The key question is, where do you get started?
12:00pm-1:25pm
Birds of a Feather Lunch
Nikko I
Find your clan of like-minded colleagues with whom to dine and discuss, comparing your organization's stories and challenges.
- The Data Mining Process:
- Project management and organizational process
- Data preparation
- Core predictive modeling methods
- Deployment and model integration
- SAS Sponsored Topics:
- Best Practices in variable selection/dimension reduction
- Best Practices in automating steps in the modeling process to create analytic bandwidth
1:25pm-2:25pm
Nikko II
Expert Panel Session
Cross-Industry Challenges and Solutions in Predictive Analytics
In order to benefit from predictive analytics, there are hurdles to jump and barriers to avoid. As a cross-industry event, PAW provides the opportunity for verticals to learn from one another how best to get the job done. In this panel discussion, three senior experts with experience across complementary ranges of verticals will compare experiences and lessons-learned, discussing the toughest challenges and brightest solutions as they compare across industries.
|
Dean Abbott, President, Abbott Analytics Vijay Desai, Ph.D., Principal Scientist, SAS Richard G. Vlasimsky, Co-Founder & Vice-President, Valen Technologies |
2:25pm-2:40pm
Session Break – Exhibit Hall
Nikko Ballroom Foyer
Delegates may choose to attend either session at this time.
Track 1: Financial Services and Insurance
2:40pm-3:20pmNikko II
Case Study: Pinnacol Assurance
A Day in the Life: Predictive Analytics in the Insurance Industry
The application of predictive analytics is becoming more and more prevalent in the insurance industry. Those insurance carriers that use predictive analytics and use it well become more competitive through more accurate risk based pricing, more financially sound risk selection, more effectively managed claims, and more accurate audit targeting.
These improvements provide insurance carriers critical competitive advantages. This presentation will walk the audience through a “day in the life” of predictive analytics in the insurance industry, from predictive underwriting to audit, and provide case studies detailing implementation experiences, lessons learned, and benefits realized by two leading insurance carriers.
Track 2: Online Applications
2:40pm-3:20pmNikko I
Case Study
Predictive Modeling for Email Marketing
Most email marketing analytics focuses on opens, clicks and conversions. This is failing today because open rates are falling. A better approach is to analyze the behavior and demographics of email opt-in subscribers. The talk focuses on what e-Dialog has done on utilizing CHAID and regression techniques to examine how past behavior predicts future behavior. These advanced approaches help separate email blasts from the finely-tuned relevant offerings necessary for success in today's email environment sending the right emails to right people, at the right time. The cases involve three major clients: a retailer, a major airline and a major computer manufacturer.
|
Arthur Hughes, Senior Strategist, e-Dialog.com Anna Lu, e-Dialog.com |
3:20pm-3:35pm
Afternoon Coffee Break – Exhibit Hall
Nikko Ballroom Foyer
Delegates may choose to attend either session at this time.
Track 1: Financial Services and Telecom
Uplift Modeling
3:35pm-4:15pmNikko II
Case Study: Telenor
Applying Next Generation Uplift Modeling to Optimize Customer Retention Programs
In today's down economy, all organizations are being challenged to drive greater retention and revenue whilst spending less money and using fewer resources. In this session, hear how the world's 7th largest mobile operator has applied next generation “uplift” modeling (i.e. “net lift” modeling) to optimize retention programs and seen results 36% better than those possible using traditional analytic practices. Impressively, these results were reached while, at the same time, slashing the cost of retention programs by a staggering 40% – making this an ideal fit for today's recessionary marketing requirements.
Uplift models are different from traditional modeling in that the approach measures and predicts the true incremental impact of marketing activity. Whereas traditional models only aim to predict “behavior”, uplift models actually predict the incremental “change in behavior.” Telenor's novel approach and application was recently featured in Forrester Research's popular new report “Optimizing Customer Retention Programs”, where the approach was shown to achieve an 11-fold increase in campaign ROI when compared with existing programs.
Track 2: Online Applications
3:35pm-4:15pmNikko I
Case Study: 
Predicting the Quality of Advertisements on AdWords
AdWords provide a primary source of revenue for Google, so it's critical that users are satisfied with advertisements they click on. This talk will present a recent initiative generating models that predict certain quality metrics of ads in AdWords. This work focuses on the “cold-start” problem: how to handle new ads not yet evaluated and “long tail” ads that are rarely clicked. Analytical design choices and modeling trade-offs will be discussed, and results on different experiments will be presented.
Delegates may choose to attend either session at this time.
Track 1: Financial Services and Telecom
Uplift Modeling
4:15pm-5:05pmNikko II
Case Study: Charles Schwab & Co
Maximize Marketing Impact with Net Lift Models
The true effectiveness of a marketing campaign is measured by the incremental impact. That is, additional revenue directly attributable to the campaign that we would not otherwise have received. Yet traditional targeting criteria are often designed to find clients that are interested in the product but would have bought it whether or not they received the promotion. In such cases, the incremental impact is insignificant and the marketing dollars could have been spent elsewhere.
Net Lift Models are designed to maximize the incremental impact by targeting the undecided clients that can be motivated by marketing.
Track 2: Online Applications
4:15pm-5:05pmNikko I
Case Study: 
Predictive Scores to Optimize Pay-Per-Click Campaigns
In the context of pay-per-click advertising campaigns (Google, Yahoo), advertisers purchase keywords, create ad groups and automatically set a separate maximum bid per keyword / match type. Bids and keywords are optimized weekly, daily or sometimes in real time, based on past performance. Keywords that have generated very few clicks – and new keywords with zero click – require more advanced predictive models to assess their performance and generate optimum bids. These keywords represent sometimes as much as 50% of the potential ad spend, for large advertisers interested in the “long tail”. A scoring methodology based on text mining techniques will be discussed to successfully address this challenging problem.
Wrap Up / Conference End
