Pre-Conference Workshops: Sunday, June 24, 2012 |
Workshop sponsored by:
Half-day Workshop
•Room: W196
JMP, R, and SAS:
the Beauty of Multiple Paradigms
Click here for a detailed workshop description
Instructor: Matthew J. Flynn, Director of Claims Research, Travelers Insurance
[ Top of this page ] [ Agenda overview ]
Conference Day 1: Monday, June 25, 2012 |
9:45am - 7:30pm
Exhibit Hall Open
Registration & Breakfast
[ Top of this page ] [ Agenda overview ]
9:00-9:45am • Room: W190A
Keynote
Persuasion by the Numbers: Optimize Marketing Influence by
Predicting It
Data driven marketing decisions are meant to maximize impact - right? Well, the only way to optimize marketing influence is to predict it. The analytical method to do this is called uplift modeling. This is a completely different animal from what most models predict: customer behavior. Instead, uplift models predict the influence on customer behavior gained by choosing one marketing action over another. The good news is case studies show ROI going where it has never gone before. The bad news? You need a control set... But you should have been using one anyway! The crazy part is that "marketing influence" can never be observed for any one customer, since it literally involves the inner workings of the customer's central nervous system. If influence can't be observed, how can we possibly model and predict it?
Speaker: Eric Siegel, Conference Program Chair, Predictive Analytics World
[ Top of this page ] [ Agenda overview ]
Platinum Sponsor Presentation
Raising the Bar for Predictive Analytics Deployment: The Newest Techniques
Although the use of predictive analytics has come a long way in recent years, it is clear that there are now much higher expectations for wider and more accurate deployment methods. So while more organizations see the value of analytics, few are comfortable with their current tools and abilities to create and deploy useful solutions.
In this session we'll explore the very newest techniques and capabilities that have emerged to help you ingrain predictive analytics into the DNA of your organization, and deploy solutions that empower your team to make the right decisions and consistently deliver the best results.
[ Top of this page ] [ Agenda overview ]
Platinum Sponsor Presentation
Predict who is Persuadable Before you Market
As you heard in the Keynote earlier "Persuasion by the Numbers: Optimize Marketing Influence by Predicting it," the only way to optimize marketing influence is to predict it. However, how can marketers actually predict the change they can have on customer behavior? It's called Uplift modeling.
Most direct marketing targets the wrong people. It wastes money by focusing effort on many who will act anyway, may not react positively, and even some who may react negatively. The most effective marketing instead focuses only upon those customers who can be persuaded to take action as a result of your message – saving marketing waste and improving results.
By leveraging uplift modeling to figure out which customers are actually persuadable before you market, your organization can:
- Improve campaign results by 30-300% versus traditional analytic methods
- Slash campaign spend by as much as 40%
- Eliminate the negative effects of marketing by weeding out the "sleeping dogs"
Leading analysts declare that "Uplift analysis is a must-consider concept for every organization with significant campaign management activities." This session will explore the principle of uplift modeling which can be used to optimize targeting to maximize the returns from direct marketing. Attend this informational webinar to learn about this latest technique.
[ Top of this page ] [ Agenda overview ]
Breaks / Exhibits
[ Top of this page ] [ Agenda overview ]
10:40-11:25am • Room: W192B
Track 1: Public Sector
Case Study: City of Chicago
Lessons from Year One: Predictive Analytics in Government
Government has long lagged behind the private sector's use of data, but the new administration in Chicago aims to change that. In the first year, the Mayor appointed Chicago's first Chief Technology Officer and the nation's first municipal Chief Data Officer. Beginning to use data meaningfully in Chicago has presented many challenges and opportunities. In this presentation, the CDO will outline the lessons from the first year of bringing analytics into government. Projects include data documentation, creation of a common operating platform using NoSQL geospatial capabilities, and development of a predictive framework to capture a neighborhood's quality of life.
Speaker: Brett Goldstein, Chief Data Officer, City of Chicago
•Room:W190A
Track 2: Fraud Detection
Case Study: USPS Office of Inspector General
Fraud Detection: Fraught with Frightful Modeling Hurdles
Building predictive models to find fraud, waste, and abuse can be an especially tricky application of data mining. Complicating factors include: frequent lack of training cases, ever-changing patterns as fraudsters adapt their schemes, high sensitivity to false positives, and the relative rarity of fraud. We describe approaches to tackling these modeling hurdles, and highlight them with examples from our consulting projects in the commercial and government arenas.
Speaker: Antonia de Medinaceli, Director of Fraud Analytics, Elder Research, Inc.
11:05-11:25am • Room: W190A
Track 2: Sentiment Analysis
Case Study: MTV Networks
Predictive Social Marketing – Sentiment Forecasting and Impact on Success
Every summer, music fans worldwide look forward to one of the biggest music events of the year—the MTV Video Music Awards (VMAs). This year's VMAs turned out to be one of the world's largest, simultaneous social viewing experiences ever. Leading up to this year's VMAs, MTV marketers set out to grow the brand's social media presence and drive awareness. With 85 million MTV Facebook fans and more than three million Twitter followers—the stage was set for a firestorm of conversation and sharing. In order to assist MTV in their primary goal of gaining deeper insights into relationships between social activity and engagement with digital content, we combined Twitter and MTV.com data streams with text mining and predictive analytics techniques. As marketing continues to develop, sentiment forecasting will be critical to success in optimizing published content for both publishers and advertisers.
[ Top of this page ] [ Agenda overview ]
Expert Panel
Wise Enterprise: Best Practices for Managing Predictive Analytics
Your company is trigger-happy for predictive analytics, and there's plenty of excitement, momentum and public case studies fueling the flames. Are you destined for success or disappointment? Is it a sure-fire win to gain buy-in for a promising analytics initiative, equip your most talented practitioners with a leading solution, and pull the trigger?
This panel of leading experts will address the holistic view. What are the most poignant and telling failures in the repertoire, and where is the remedy? Beyond the management of individual analytics projects, what enterprise-wide communication processes and other best processes provide best security against project pitfalls? Stay tuned for big answers to these big questions.
[ Top of this page ] [ Agenda overview ]
Gold Sponsor Presentation
Using Analytics to Inform Human Capital Decisions That Drive Business Outcomes
The emerging space of HR Analytics can help you unlock the value of your employees by linking people data to critical business outcomes. This session presents a case study of how predictive analytics was used to address a classic question of Build vs. Buy for selecting Store Managers within Sears and Kmart. You will see how data-minded HR professionals were able to utilize some basic techniques to answer questions such as, "How do internally promoted managers differ from externally hired managers on driving sales, profitability, and customer satisfaction" and "How long does it take externally hired managers to get up to speed and become profitable." Whether you are just beginning your HR Analytics journey or you have been on the path for a while, you will take away some ideas for application of tried-and-true analytic techniques to answer human behavior questions.
Speaker: Koren Ichihara, Sr. Analyst – HR Analytics, Sears Holdings Corporation
[ Top of this page ] [ Agenda overview ]
Lightning Round of 2-Minute Sponsor Presentations
[ Top of this page ] [ Agenda overview ]
Lunch sponsored by:
Lunch / Exhibits
[ Top of this page ] [ Agenda overview ]
Special Plenary Session
Multiple Case Studies: Anheuser-Busch, the SSA, Netflix
Data Mining Lessons Learned - Technical & Business - From Applied Projects
In the recounting of analytics projects, my favorite part is "the reveal": where the idea that turned things around is disclosed. Often disarmingly simple (in retrospect) it is virtually always preceded by waves of failure. Yet failure, or at least an environment shockingly tolerant of it, may be essential to the emergence of such breakthroughs.
I will tell tales of some favorite "reveals" that led to technical successes. But, a true win must also be a business success. This requires dealing well with idiosyncratic carbon-based life forms. So we'll also discuss the (painfully acquired) lessons in the parallel universe of business.
Speaker: John Elder, CEO & Founder, Elder Research, Inc.
[ Top of this page ] [ Agenda overview ]
2:25-2:45pm • Room: W190APlatinum Sponsor Presentation
Predicting Risks Using a Subjective Model: Utilizing Subject Matter Expertise and Clustering to Estimate Environmental Hazard
In the world of energy generation and natural resource extraction, the industry leaders own and maintain thousands of properties, both new and historical, used in the production of oil and gas products. Older retired sites can harbor undetermined risk from ill-documented or unknown past spills or accidents. The challenge is to identify potential risk from observable characteristics and determine where clean-up funds should be targeted to best mitigate potential damage to the local community and environment.
This presentation describes the creation of a model to evaluate environmental risk, the validation of that model through consultation with subject matter specialists and how factor analysis and clustering were applied in order to predict which sites were likely to conceal risks and to determine appropriate remediation funds.
[ Top of this page ] [ Agenda overview ]
Track 1: Crowdsourcing Predictive AnalyticsCase Study: Allstate & Wikipedia
Crowdsourcing Predictive Analytics: Why 25,000 Heads Are Better
Than One
Astronomical amounts of data -- measured in exabytes -- are being created each day. This has not only made it difficult for companies to cope with the deluge but has also made it harder for them to make sense of it all. While there has been a lot of development in recent years around tools to manage large volumes of data, there has been very little progress on efficient ways to tease out trends from the data and gain actionable insights. Crowdsourcing the analysis task through public data mining competitions addresses this critical need in the marketplace. In this session we will discuss how Allstate, a forward-thinking fortune-100 company, found an innovative way to leverage the power of crowdsourced analytics for predicting insurance claims based on vehicle data. We will also talk about how Wikimedia Foundation tapped into the collective talent of data scientists around the world to predict editor churn. These competitions have shown that this approach is not just efficient but also creates highly predictive models, which have beaten internal benchmarks on every occasion.
Speaker: Karthik Sethuraman, Head of Analytic Solutions, Kaggle
2:50-3:10pm • Room: W190A
Track 2: Social Media AnalyticsCase Study: Topsy Labs
Leveraging Social Media to Identify Predictive Behaviors
Millions of people communicate every day within social networks around the globe. If businesses can quantify and qualify what people are saying about their products and services in realtime, they can utilize this intelligence to adjust messaging, product and corporate strategies. Topsy Labs is a social search and analytics platform that processes social sources to enable businesses globally to apply social intelligence to realtime decisioning. This session will dive into the world of realtime social analytics and discuss the methods used along with benefits accrued.
Speaker: Jamie de Guerre, VP Product, Topsy
[ Top of this page ] [ Agenda overview ]
Track 1: Market Research
Case Study: Microsoft
Combining Customer Behavior & Attitudes – Newer Ways to Gather Data, Leverage Analytics & Take Smarter Product & Marketing Decisions
Traditionally, product and marketing teams have relied heavily on surveys to understand customer perceptions with little or no visibility into how customers use or interact with products. In reality, it is usage that influences perceptions that in turn leads to customer actions like repurchase and recommendation. Hence linking behavior and attitude enables organizations to better understand customers and thereby take more informed decisions.
In this session Jeff Ahlquist, Research Director at Microsoft with the help of a case study talks about how Microsoft is leveraging telemetry, market research surveys, big data technologies and predictive analytics to understand customers more intimately.
3:15-3:35pm • Room: W190A
Track 2: Social Data; Text Analytics
Case Study: British Broadcasting Company
Data Mining for Social Moderation
The BBC (the British Broadcasting Company), a known media force in Europe, implemented an in-database data mining solution for their public-facing website. The BBC encourages users to post comments and become part of their social media community. In this project, the BBC needed a way to improve the social moderation of millions of posts on thousands of forums. This presentation covers integration within a leading database platform, and final model results evaluation. We also discuss cost efficiencies. More organisations will use enterprise solutions to handle high-volume social networking.
Speaker: Mark Tabladillo, Mentor & Paco Gonzalez, Mentor, SolidQ
[ Top of this page ] [ Agenda overview ]
Plenary Session
Track 2: Social Data for Financial Indicators
Case Study: AlphaGenius
Sentiment Investing - Above Market Returns Extracting & Analyzing Twitter & the Social Internet
For the first time in human history a collective measurement of sentiment can be taken from Twitter and the social Internet. The data is often free and available for everyone to see. The challenges for prediction is collecting, normalizing, and analyzing the unstructured data. AlphaGenius can describe our process of doing this for building investing models.
Speaker: Randy Saaf, CEO & Founder, AlphaGenius
[ Top of this page ] [ Agenda overview ]
Break / Exhibits
[ Top of this page ] [ Agenda overview ]
4:35-5:20pm • Room: W192B
Track 1: Customer Retention; Financial ServicesCase Study: Paychex
Combat Client Churn with Predictive Analytics
In economic conditions such as this, it is critical for businesses to have a stronghold on their client retention efforts. Historically, it has been shown that businesses excelling in this arena are often better positioned for long-term success and possess a competitive advantage. To optimize the value of retained customers it's essential to understand which clients are a fit for retention campaigns so that the loss of time and resources is minimized. In this session, we will review how Paychex leveraged two existing models, Paychex Attrition Model and a custom built Lifetime Value Model, to create a Retention Tracking System.
Speaker: Frank Fiorille, Director of Risk Management, Paychex, Inc.
4:35-4:55pm • Room: W190A
Track 2: Insurance
Case Study: Travelers Insurance
Insurance and R
Connecting to additional software tools from R creates additional analytic bandwidth. Cross-tool communication takes the strengths from multiple tools and empowers the analyst with tools where the sum is more than the parts. More bandwidth means more analytic productivity! Combining tools provides a better match of analyst skill levels to tools and takes advantage of smarter automation to enable/empower the analyst to concentrate on creating more value.
My motivation - to help two groups of analysts here at Travelers, 1) experienced modelers who wish to explore sophisticated modeling tools in R, and 2) new grads who come to us with strong R backgrounds but limited skills with commercial enterprise analytics tools. In order to be productive, both groups need to access company data and high-performance software which is located on central UNIX servers (AIX and Linux) and access to company data repositories such as downstream databases, they need to manipulate and summarize that data, then bring it down to R on the desktop. To date, this is a very manual time-consuming process. Recent versions of enterprise analytics tools have added capabilities to access R. Taking another approach, an R package helps communicate with the enterprise commercial tools. The key component that makes this possible is other R packages that provide communication via MS Windows COM facilities to access the automation objects and integration technologies on the desktop.
Speaker: Matthew Flynn, Director of Claim Research, Travelers Insurance
[ Top of this page ] [ Agenda overview ]
5:00-5:20pm • Room: W190A
Track 2: Insurance & Big dataCase Study: Allstate
A Brave New World - Predictive Model Development with Hadoop, Rhadoop, & R
Big data is everywhere and new tools are emerging to harvest its predictive value. In this talk, we'll share our experience -- the good, the bad, and the ugly -- with moving into the world of Hadoop and R for predictive modeling. What questions did we have going in? How did they change along the way? What new problems did R and Hadoop create? We'll share what we learned and what we'd do differently if we were starting again.
[ Top of this page ] [ Agenda overview ]
5:25-6:10pm • Room: W192B
Track 1: Blackbox TradingAlgorithmic Trading: How to Avoid Fooling Yourself When Searching for an Investment Edge
Speaker: John Elder, CEO & Founder, Elder Research, Inc.
5:25-5:45pm • Room: W190A
Track 2: Advanced Predictive Modeling MethodsCase Study: Nielsen & a financial services enterprise
Finding Consumers More Accurately and Actionably Using Data Mining Tools
Being able to type consumers into specific segments is an important problem. Parametric models, like discriminant analysis, are often used to predict classification into consumer segments. Our experience is that these models suffer from predictability, inability to tolerate missing values, and lack of flexibility in trade-offs between predictability, purity, and complexity. Classification trees have produced better results. In addition to improved accuracy, they offer greater adaptability for business needs (e.g. number of variables included, the way the inputs will be collected). We provide a case study in which classification trees provided superior accuracy and actionability to a financial services client.
Speaker: Dimitar Antov, Project Director, The Nielsen Company
5:50-6:10pm • Room: W190A
Track 2: Analytical TractionWhat is the Analytic Maturity of Your Company and How Can You Improve It?
The Software Capability Maturity Model has proven its value over time and allows companies to assess their software development processes and to improve them over time. Today, there is no similar model for evaluating the analytic maturity of an organization. In this talk, we describe an Analytic Maturity Model that we have developed and tested over the past four years that enables companies to build better analytic models. We discuss eight steps that your company can take to improve its analytic maturity. We also share the results of a survey of over 20 companies that assessed their analytic maturity.
[ Top of this page ] [ Agenda overview ]
Reception / Exhibits
[ Top of this page ] [ Agenda overview ]
Conference Day 2: Tuesday, June 26, 2012 |
9:45am-4:30pm
Exhibit Hall Open
[ Top of this page ] [ Agenda overview ]
Registration & Breakfast
[ Top of this page ] [ Agenda overview ]
Keynote
Predictive Analytics and Business Performance
In this session, Bruno Aziza will discuss the challenges organizations face with Analytics and Performance. This participative session will provide first-hand accounts from Fortune 500 companies who are winning by building accountability, intelligence, and informed decision-making into their organizational DNA.
[ Top of this page ] [ Agenda overview ]
Platinum Sponsor Presentation
Managing Forward: Analytics For Today's Multi-Channel, Multi-Device Consumer
When done right, customer satisfaction measurement can yield more than just insights into how well your company, brand, or channel (e.g., web, mobile, store) is performing today. It can also predict the likelihood of customers to engage in critical future behaviors. However, not all methodologies are created equal. They must answer three essential questions of management while demonstrating success not only in theory but in the marketplace.
Speaker: Larry Freed, President & CEO, ForeSee
[ Top of this page ] [ Agenda overview ]
Gold Sponsor Presentation
Mu Sigma – A Brief Introduction
An elevator pitch on Mu Sigma
Speaker: Prathap Venkatesan, Regional Head, New Business, Mu Sigma
[ Top of this page ] [ Agenda overview ]
Gold Sponsor Presentation
Big Data Insights in a Rush
The data avalanche has only just begun. To avoid getting crushed, business analysts and data scientists need to arm themselves with the platforms and techniques to support deep mining and machine learning, on a scale that will break today's systems. In this short session, learn about how to prepare for and tackle really, really, really big data challenges, including how to compose and execute analytics on platforms like Hadoop, without writing any code.
[ Top of this page ] [ Agenda overview ]
Breaks / Exhibits
[ Top of this page ] [ Agenda overview ]
Track 1: Thought Leadership
Case Study: An Insurance Company
Business Friendly Data Mining
Organizations that are more sophisticated with analytics are more likely to be top business performers. But too often there are gaps in understanding between business, IT and analytics teams. Gaps in building business understanding lead to analytic models that don't have business value and gaps in implementation lead to models that take months or years to adopt or never get adopted at all. Wasted effort and "good" models that just sit on the shelf gathering dust are the result. In this session, James Taylor outlines some proven techniques for improving business/analytics/IT collaboration, clarifying data mining goals and enabling rapid deployment of models in systems.
Speaker: James Taylor, CEO, Decision Management Solutions
10:45-11:05am • Room: W190A
Track 2: Telecommunications
Case Study: Nokia-Siemens Networks
Understanding Mobile User Outages: Predictive Analytics in Wireless Broadband Networks
Today's Telecom service providers are rapidly evolving their network, adding 4G/broadband capabilities to satisfy the ever increasing data demand from an exponentially growing smart phone user population. In this work, we focus on case study relating to service availability and continuity for a major 4G customer in a live mobile broadband network. We set up a supervised learning problem to predict customer outages utilizing field data in such a mobile network. Results using non-linear regression and ensemble techniques are shown to provide key variable importance measures and further suggest concrete recommendations for improving network performance and user experience.
[ Top of this page ] [ Agenda overview ]
Track 1: Text Analytics
Cross-Language Text Analytics: Overcoming Language Barriers
A brand manager wonders what consumers throughout Asia think about soft drinks. An intelligence analyst suspects a terrorist group is organizing a bombing online. Engineers must investigate quality problems leading to complaints around the world. Each of them needs information locked in some form of text, but the person who needs the information doesn't know the language of the source text.
In this session you will learn:
- Steps for deriving subject and sentiment categorizations from foreign language text
- How performing text analytics on automatically translated text wastes resources and produces poor results
- Why cross-lingual text analytics offers unique opportunities to support business growth.
Speaker: Meta Brown, General Manager of Analytics, LinguaSys
11:05-11:25am • Room: W190A
Track 2: HR Analytics
Case Study: U.S. Special Forces
Hiring and Selecting Key Personnel Using Predictive Analytics
Hiring and selection of personnel in specialized work environments incurs huge direct and opportunity costs for organizations. One of the largest challenges is that the selection process is often left in the hands of those with either high experience in the domain area but little experience in selection or vice versa.
Predictive Analytics and statistics can play a critical role in formalizing and automating much of the selection process. This session provides an overview of the selection processes using both measures of skills and psychological measures to quantify IQ, domain knowledge, grit, and determination. Examples will be drawn from hiring practices for Special Forces (such as Army Rangers and Navy SEALs) and predictive analytics teams.
Speaker: Dean Abbott, President, Abbott Analytics
[ Top of this page ] [ Agenda overview ]
Track 1: Sponsored Lab
Sponsored Lab: Live Topical Demo
Using Spatial and Predictive Analytics to Supercharge Location-Based Marketing Decisions
In this session we present a case study that shows how R-based predictive analytics can be combined with Alteryx's spatial analytics capabilities, in particular its drive time engine, to optimize a direct marketing campaign in a way that explicitly considers a prospect's location relative to the retail outlets of both the company and its main competitors.
Track 2: Sponsored Lab
Sponsored Lab: Live Topical Demo
Improving Sales Forecasts Using IBM Predictive Analytics
Today accurate sales forecasts are critical for creating company budgets and aligning resources effectively. Many companies have moved beyond sales force predictions to also include predictive analytics like Time Series Forecasting. In this session we will discuss how United Stationers (NASDAQ: USTR) a fortune 500 wholesale distribution company uses predictive analytics to forecast sales for three of its divisions. This method of forecasting has helped United Stationers improve the overall accuracy of forecasts by 30%. We will also spend time discussing the predictor variables that are most accurate, tips for preparing data prior to forecasting and methods for working with seasonal data trends.
Speaker: John Hassman, Director, Marketing Analytics, United Stationers, Inc.
[ Top of this page ] [ Agenda overview ]
Lunch / Exhibits
[ Top of this page ] [ Agenda overview ]
Keynote
Predictive Analytics For the Win: Data Science Meets the Quiz
Show Jeopardy!
This keynote will provide a first-hand account of how predictive analytics techniques were applied to the problem of preparing for Jeopardy!, a television quiz show. Data analysis of both the problem domain, Jeopardy!, and sampling of human learners to pinpoint strengths & weaknesses were used to develop a plan for preparing for the show. This analysis and plan allowed Craig to set many records on the show, including the highest one day winning total ever & the fourth highest overall total ever. Comparisons to the contemporaneous IBM's Watson project will also be presented, chiefly the difference between a 100% AI solution and a solution which involves augmenting a human player with insights gleaned from predictive analytics & machine learning techniques.
Speaker: Roger Craig, CEO, Cotinga LLC
[ Top of this page ] [ Agenda overview ]
Lightning Round of 2-Minute Sponsor Presentations
[ Top of this page ] [ Agenda overview ]
Break/Exhibits
[ Top of this page ] [ Agenda overview ]
Track 1: HR Analytics
Case Study: Hewlett-Packard
An Innovative Approach to Analyze Employee Satisfaction Response in Light of Customer Satisfaction Response
HR teams often face difficulty in prioritizing the areas of improvement around employee satisfaction mainly due to lack of benchmark figures available. So, we linked employee satisfaction with customer satisfaction and see the effect of former on latter in helping business to identify opportunity areas. Here an innovative method, surrogate regression analysis, was developed where the entire dataset was transformed to surrogate values keeping the essence of original data by running multiple simulations. This helped us in avoiding the trouble of utilizing the ordinal survey data for causal analysis required for this purpose.
Speaker: Jyotirmay Nag, Business Analyst of RnD Analytics, Hewlett-Packard
2:35-2:55pm • Room: W190A
Track 2: Healthcare Analytics
Case Study: Pfizer
Right Medicine, Right Patient
Can predictive modeling improve patient care? A wealth of data exists in large healthcare databases on patient disease characteristics and their response to specific treatments. Max will discuss some of the technical and non-technical issues in providing care providers with quantitative results related to how individual patients might response to therapies.
Speaker: Max Kuhn, Director of Nonclinical Statistics, Pfizer
[ Top of this page ] [ Agenda overview ]
Track 1: Workplace Behavior Modeling; Enterprise Dynamics
Modeling Project Leaders' Perceptions of Their Clients
Based on a survey of project managers across several industries, this presentation will summarize various models that were developed in order to gain insight into certain project manager perceptions of project sponsor behaviors.Specific project sponsor related behaviors addressed include: project manager trust in the project sponsor, perceived project sponsor openness to information, and perceived project sponsor use of power. Use of distribution fitting and simulation software will be highlighted. A key lesson learned from this study is that distribution fitting and simulation techniques can be used to gain greater insight into workplace behaviors.
Speaker: David Perkins, Associate Professor of Business, Grand Canyon University
3:00-3:20pm • Room: W190A
Track 2: Telecommunications
Case Study: Accenture
Next Generation Mobile Analytics - Combining the Power of Real Time Data with Predictive Analytics
As the mobile markets become more competitive, it becomes much harder to fulfill demanding subscribers who have complex needs. This has caused operators losing millions of dollars due to high churn issue. This session shows how we combine the power of traditional predictive modeling with real time data. Predictive modeling is used to understand the probability a subscriber is going to churn based on historical data while real time data is used to trigger and enhance the predictive power. To ensure the right level of investment, offer optimization is applied to offer the most relevant treatments to the subscribers.
Speaker: MeiMei Lim, Principal, Accenture
[ Top of this page ] [ Agenda overview ]
3:20-3:55pm • Room: W196
Breaks / Exhibits
[ Top of this page ] [ Agenda overview ]
3:55-4:15pm • Room: W192B
Track 1: Financial Services
Case Study: Silicon Valley Bank
Advanced Analytics inside the Banking Industry
One of the critical goals of a financial institution is revenue growth. The mantra being to reduce leakage/attrition, expand and strengthen relationships with current clients by cross-selling additional products/services, and to acquire new clients. Data mining, advanced analytics help in each of the three areas mentioned above. In this session, I would like to share the type of information regarding current clients that help profile them and apply predictive modeling to understand: - Who is likely to purchase additional products? - What they might need next? - More importantly, how can we transform these insights into actions? A clear vision, efficient planning and effective implementation to translate vision into reality along with appropriate actions helps move the needle!
Speaker: Kirtida Parikh, Director & Head, Enterprise Business Analytics, Silicon Valley Bank
3:55-4:15pm • Room: W190A
Track 2: Data Visualization
Case Study: eBay
Experimentation - From visual data exploration to decisions
The online experimentation lifecycle at eBay is a collaborative effort involving several well-defined steps. Visualizations are used both for hypothesis development as well as for communication of the final results. eBay's experimentation platform is used to facilitate the design, launch and analysis of experiments. In this session, we look at the online experimentation lifecycle at eBay and use some case studies to look at related tools and techniques.
Speaker: Vijay Madhavan, Senior Product Manager, eBay Inc.
[ Top of this page ] [ Agenda overview ]
4:20-4:40pm • Room: W192B
Track 1: Targeting Marketing
Case Study: Microsoft
Getting Management to Act on Driver Models
As a Marketing Director, you committed to increasing the top-2 box satisfaction number by 5%. The satisfaction attributes in a survey are heavily inter-correlated and many standard techniques will not give stable and valid priorities leading to missed opportunities and comprised willingness to act on the insights. This problem is very widespread. A case study of a major Technology firm shows how a Bayesian statistical framework leads to stable and valid identification of management's priorities and leads to insights that are more easily accepted and acted on as indicated by feedback from 15 marketing directors.
Speaker: Marco Vriens, Senior Vice President, The Modellers LLC
4:20-4:40pm • Room: W190A
Track 2: Customer Retention
Case Study: Hewlett-Packard
Modeling Consumer Attrition in a Non-Contractual Setting
Consumer attrition is a poignant issue in a non-contractual setting like HP where the relationship duration with a customer is not observable. Modeling consumer attrition using purchase behavior patterns would help in devising differential targeting strategies based on the risk of attrition. We allow a customer twice her maximum inter-purchase time from her last purchase to return, else we consider her as attrited. Usual classification methods like Logistic Regression, fail to capture the complex, non-linear relationship in the data. We use an adaptive learning algorithm - artificial neural network - to address the unique traits of HP customers.
[ Top of this page ] [ Agenda overview ]
4:45-5:05pm • Room: W192B
Track 1: Forecasting Demand
Case Study: Large Movie Studio
Using Big Data to Optimize and Predict Opening Week at the Box Office
While motion pictures can be evaluated based on acting, directing, or popular awards, the key performance indicators important to industry executives are the opening box office numbers. There are a range of factors determining opening week performance, however, from the genre, actors/actresses/director of the movie to the marketing budget, its channel mix, and timing of the marketing activity prior to opening week. Seasonal and economic factors can also have an impact as overall movie attendance has a strong seasonal component with a summertime and end-of-year skew; economic factors such as gas prices also appear to impact overall movie attendance. The available levers to market movies has also expanded with the growth of social media and the online viewing and sharing of movie trailers. With this changing landscape, GroupM Business Science has developed models within the motion picture industry using both macro and micro-level data to understand the macro-economic factors impacting overall box office sales as well as movie-specific models which take into account marketing support levels and timing/flighting of owned media support. The models also include online movie trailer viewership and sharing (earned media), as well as website traffic of the movie (owned media) to estimate opening week performance. The model can be used to estimate opening week with a planned budget and flighting as well as uncover opportunities for optimization based on media mix and timing/flighting for a larger opening week with a similar marketing budget.
4:45-5:05pm • Room: W190A
Track 2: Real Estate Market Scoring
Case Study: Altos Research
There & Back Again: Model Interpretability in Real Estate Market Scoring
Seasoned predictive analytics practitioners understand that simple "accuracy" is the beginning of model validation not the end. Perfect accuracy on your own training data is trivial. How confident are we in our predictions during truly unprecedented scenarios? The business builds confidence and optimizes "variance" by involving itself in the gritty modeling process. Black boxes are difficult for the business to interpret so improving robustness often means going back to more transparent models. Ben will present a case study in local residential real estate market scoring when non-parametric ensemble methods were left behind for marginally less accurate but interpretable linear models.
Speaker: Ben Gimpert, Chief Technology Officer, Altos Research
[ Top of this page ] [ Agenda overview ]
5:10-5:30pm • Room: W192B
Plenary Session
Track 1: Reliability Modeling
Case Study: Social Media Analytics
The rise of social media platforms, such as Facebook and Twitter, is providing unprecedented opportunities for marketers to target and engage with their customers. However, marketers are also simultaneously facing numerous challenges in maximizing the returns on their social media spend. In this talk, we present an overview of the methodologies we developed to optimize advertising campaigns on Facebook. These methodologies, applied for more than 100 brands, have resulted in significant improvements in click-through-rates, conversion rates and subsequently drop in cost-per-acquisition.
[ Top of this page ] [ Agenda overview ]
Post-Conference Workshops: Wednesday, June 27, 2012 |
Full-day Workshop
• Room: W192A
The Best & the Worst of Predictive Analytics: Predictive Modeling Methods & Common Data Mining Mistakes
Click here for a detailed workshop description
Instructor: John Elder, CEO & Founder, Elder Research, Inc.
Full-day Workshop
• Room: CC21C
R for Predictive Modeling: A Hands-On Introduction
Click here for a detailed workshop description
Instructor: Max Kuhn, Director, Nonclinical Statistics, Pfizer
[ Top of this page ] [ Agenda overview ]
Post-Conference Workshop: Thursday, June 28, 2012 |
Workshop sponsored by:
Full-day Workshop
• Room: W195
Advanced Methods Hands-on:
Predictive Modeling Techniques
Click here for a detailed workshop description
Instructor: Dean Abbott, President, Abbott Analytics
[ Top of this page ] [ Agenda overview ]
