Conference Day 1: Wednesday, May 14, 2014
7:30-8:45am • Room: 205
Conference Chair Welcome
Speaker: Richard Boire, Conference Chair, Predictive Analytics World Toronto
Modernization and Analytics
More info coming soon.
9:10-10:00am • Room: 206AC
The Data Scientist and Value Architect: A Predictive Analytics Marriage Made in Heaven
A world exploding in data and information leads to this fundamental question: How do we make sense of it all? How can we achieve success. Software and technologies do help but they are only tools. Ultimately it is people that will determine success in this field. But what kind of people. Two clear roles emerge, one where the person deals with data (Data Scientist) and the other where the person has that critical domain knowledge (Value Architect.)
But what are the functions of these roles and more importantly how do they interact in building predictive analytics solutions. In this session, Richard Boire discusses the importance of both roles and their specific expertise and, more importantly, how integration of these roles is the real factor to success. The need to identify a business problem and the ability to work the data to solve this problem is the underlying objective of the data scientist and the value architect. This fundamental business need is common across all industries. In the world of big data, the importance of both roles is even more paramount. Yet, the traditional predictive analytics and data mining approach in building solutions has not changed. At the same time, though, new technologies need to be embraced that allow us to adopt these same analytical approaches in the Big Data world of semi-structured and unstructured data.
Speaker: Richard Boire, Partner, Boire Filler Group
Exhibits & Morning Coffee Break
Case Study: A Major Financial Services Firm
Data Science Approach to Reduce Call Center Attrition
This case study presentation by Talent Analytics Chief Scientist Pasha Roberts provides details of how a major financial services call center used data science to significantly reduce attrition by over 30%, yielding a multi-million dollar savings. Analytics managers will leave this session with:
- a concrete understanding of the business value of talent optimization, showing how correlating "raw talent" to attrition can yield unprecedented results
- a concrete understanding of the analytics approach and models used to reduce attrition.
Track 2: Risk Analytics
Case Study: Fifth Third Bancorp
Loss Estimation Models in CCAR: Comprehensive Capital Analysis and Review by FED
CCAR requires that US banks with assets of $50B to develop a capital plans assessing the capital adequacy. 18 major firms have participated and additional 12 might join in 2014. In CCAR, each firm is expected to use quantitative models as the basis to project expected losses. While there are a range of approaches to estimate losses, we would like to emphasize best practices for consumer loans. Different from market-to-market approaches used for wholesale portfolios based on Markov Rating Transition models, a default-only approach based on Expect Loss models is our choice to produce credit loss estimations for retail portfolios.
Gold Sponsor Presentations
Case Study: Workplace Safety & Insurance Board
Discrete Time Logistic Hazards Models for Workplace Safety Insurance On-Benefit Duration Models
The safe and timely return to work for an injured worker is a key priority for Workers' Compensation. Using predictive models, we understand the key risk factors affecting the time on-benefits and determine the probabilities being on benefits at different time points. Although most injured workers experience a single off benefit event, some may experience a recurrence for the same injury after returning to work. We use discrete time logistic hazards regression to model on-benefit and off-benefit durations. The probability of being on-benefit at a point in time is then calculated using the Markov transition probabilities from the above models.
My Five Predictive Analytics Pet Peeves
Predictive Analytics (PA) has become increasingly mature as a technical discipline over the past decade in part because it stands on the shoulders of the related disciplines of data mining and machine learning. However, there are recurring themes that permeate discussion boards and conferences that have become my personal pet peeves. This talk examines five of them and why they matter to practitioners, including why we must have humility in how far data science and algorithms can take us, and the value of business objectives, measuring "success," and measuring "significance."
Speaker: Dean Abbott, President, Abbott Analytics, Inc.
Lunch in the Exhibit Hall
1:30-2:15pm •Room: 206AC
Case Study: Obama for America
Pinpointing the Persuadables: Convincing the Right Customers and the Right Voters
Marketing, political campaigning, and healthcare have one major thing in common: millions of per-person treatment decisions must be selected in order to drive positive outcomes. Prior to President Obama's reelection campaign, standard practices for persuading voters—that is, changing their minds—were unscientific and driven by long-standing assumptions and hunches. This mirrors outreach efforts by other companies and organizations, which know that a certain percentage of their marketing efforts will inevitably be wasted on people who are not going to be receptive to it. Daniel Porter of BlueLabs, who served as the Director of Statistical Modeling for the Obama Campaign, discusses his experience using the results from a large-scale randomized, controlled experiment to target persuadable voters for the Obama Campaign, as well as ways these cutting- edge statistical modeling techniques can be applied to influencing behavior in realms ranging from health outcomes to customer retention.
Lightning Round of 2-Minute Sponsor Presentations
2:30-3:15pm •Room: 206AC
The Peril of Vast Search (and How Target Shuffling Can Save Science)
It's always possible to get lucky (or unlucky). When you mine data and find something, is it real, or chance? The central question in statistics is "How likely could this result have occurred by chance?" Ancient geniuses devised formulas to answer this question for multiple special-case scenarios. Yet, their calculus only applies to quaint, handmade analyses, where only a few hypotheses are considered. However, modern predictive analytic algorithms are hypothesis-generating machines, capable of testing millions of "ideas". The best result stumbled upon in its vast search has a much greater chance of being spurious. Such overfit is particularly dangerous, as it leads one to rely on a model molded to the data noise as well as signal, which usually is worse on new data than no model at all. The problem is so widespread that it is the chief reason for a crisis in experimental science, where most journal results have been discovered to resist replication; that is, to be wrong!
The good news is an antidote exists! Dr. Elder explains the simple breakthrough solution -- still rarely employed, though newly being re-discovered in leading fields. Attend as John illustrates how to use the resampling method he calls "Target Shuffling" in multiple learning scenarios, from model fitting to data exploration, showing how it calibrates results so they are reliable...essentially providing an honest "placebo effect" against which to test a new treatment (finding).
Bottom line: Honest Data Science can save Experimental Science!
Exhibits & Afternoon Break
Case Study: Ameublement Tanguay
Predictive Analytics to the Rescue of E-mail Marketing
The ubiquity of information brought by the digital age made consumers more informed and demanding. This new paradigm stresses the importance of customer retention which lies in a deep understanding of their needs and behaviours. This situation is all the more significant in Email Marketing, where customer's inboxes are flooded with irrelevant offers making it increasingly difficult to catch their attention. This session shows you how this problem can be addressed with predictive analytics by anticipating customer behaviour and generating relevant content throughout the customer lifecycle.
Track 2: Internal Collaboration
Addressing the "Analyst/Marketer" Disconnect: 9˝ Steps to a More Successful Collaboration
Let's face it...we've all been involved in predictive analytics projects that didn't quite go smoothly. Even with the best of intentions, Analysts and Marketers do not always work well together. Maybe their objectives weren't aligned, perhaps they misunderstood key steps in the process, or it could be that they simply failed to communicate effectively. But a proper collaboration between the two can mean the difference between an organization that adopts predictive analytics as a key part of its tool-kit and one that doesn't. In this session, Colin shows you how to avoid being one of the latter.
Case Study: Liquor Retailer
Cross-selling Retail Liquor Products: A Segmented, Market-Basket Approach
This session ecamines how data analytics were used to design a holistic strategy around cross-sell and acquisition within the Sparkling Wine product category. To optimize spend, our objective was to ensure that each customer received the offer to which they were most likely to respond. To accomplish this, segmentation was performed to group customers into groups with similar preferences. Attend and see how we performed market-basket analysis within each segment to identify the products closely related to sparkling wine to achieve $7MM of new sparkling wine consumers, $19MM of sparkling wine up-sell, and $16MM of sparkling wine cross-sell.
Track 2: Sports Analytics
Designing Effective Hockey Teams through Physical Diversity
The production of winning in sports is somewhat unique. Physical characteristics potentially impact the production of winning (or losing) more than they affect production in many non-sport based industries. Different than most industries, the production of winning in sports is governed by a contest success function that represents the relationship between the inputs and winning.
Theoretical frameworks suggest that diversity can affect production through a number of different means. Relevant to physical diversity, research suggests that diversity can affect outcomes by influencing the strategies of individuals of the group, and increase production if the members of a team have relatively more disjointed and complementary relevant skills sets, or decrease production if they have relatively less disjointed skills.
The current article estimates the impact of diversity among the physical characteristics of height, weight, and age of professional hockey players on both game and game-goal outcomes. The analysis is completed at two levels to determine whether managers or coaches (or both managers and coaches) can benefit from the physical diversity effects, in terms of designing their rosters and within game line-ups respectively, to increase the likelihood of game wins.
Results suggest that managers that employ a less and more diverse group of players in terms of height and weight respectively, have a greater likelihood of winning games. However, physical diversity does not affect game-goal outcomes. Therefore, managers should consider the effects of diversity when designing team rosters, but it is not necessary for coaches to alter within game line-ups to account for their effects.
Mike Boyle, Co-Founder, The Sports Analytics Institute
Conference Day 2: Thursday, May 15, 2014
Conference Chair Welcome
Speaker: Richard Boire, Conference Chair, Predictive Analytics World Toronto
Using Predictive Analytics to Predict Employee Performance and Attrition in the Knowledge Economy
Predictive Analytics has been used to great effect for predicting human behavior in groups as diverse as customers, voters and prisoners with successful examples discussed here at PAW. Just as the business case for predicting behavior has long been established and remains compelling, innovative organizations now understand the benefits and urgency of applying the same analytic techniques to optimize the performance and churn (attrition) of their greatest asset and expense - their employees.
This keynote will outline the business imperative for using data science to solve "the other churn" problem - employee churn - with the same powerful and proven analytic techniques that have been applied predictively in other human behavior domains for years.
Business and Analytics Leaders that attend this presentation will learn:
- Why it's important to apply Churn Modeling techniques to Employees
- Why Employee Churn is an expensive -yet controllable - problem
- Why Employee Churn dramatically affects brand, service levels and product perception and thus the bottom line
- Why Businesses will benefit from applying the value of predicting behavior in other domains to Employee Performance & Churn
Exhibits & Morning Coffee Break
Track 1: Sales Analytics
Case Study: Paychex
Shaping Sales Strategy with Predictive Analytics
As we know, predictive modeling brings art AND science together. Without it, many strategic decisions are left to the "gut," leaving enormous opportunities in the age of big data. Paychex leveraged expertise in Predictive Analytics to add an empirical layer to sales strategy decisions. With the addition of models to predict likely sales units and establish a yardstick to measure sales value by zip code, sales management became statistically informed as they made decisions regarding quota setting, territory alignment and market expansion. This session describes how Predictive Analytics at Paychex was granted a seat at the strategic table.
Speaker: Tom Kern, Risk Modeling Manager, Paychex
Track 2: Survey Prediction
Case Study: CIBC
Survey Response Prediction and its Implications
The digital revolution has enabled frictionless participation of customers in surveys and reviews. Despite the growth in feedback, organizations struggle with deriving actionable insight from these rich sources. We demonstrate how a clever combination of Association Analysis with Classification models can be used to predict feedback. These predictions relieve customers from providing redundant information, and in turn help reduce the cost of surveys and boost the richness of responses. The implications of this approach are revolutionary; it can be seen as a mechanism for continually increasing signal and decreasing noise in feedback systems.
Track 1: Customer Focus
Case Study: Capital One/Bank of America
Data, Data Everywhere: Leveraging Predictive Analytics to Unlock Consumer Concerns and Eliminate Dissatisfaction
Based on widespread adoption of social media and the introduction of the CFPB's consumer complaint database, financial institutions are literally swimming in customer data. However, many institutions aren't sure how to leverage this information to not only benefit their customers, but their overall business, as well. Based on in-depth analyses of the consumer complaint database, this session provides insight on:
- Successfully tracking and managing useable data;
- Best practices for analyzing various data channels to identify valuable information; and
- Leveraging predictive analytics to stay ahead of emerging issues and prioritize responses
Speaker: Steven Ramirez, CEO, Beyond the Arc
11:20am-12:05pm • Room: 206AC
Track 2: Telecommunications
Case Study: Blackberry
Unlocking the Power of Big Data to Understand and Adapt to Your Customers
You've heard the buzz around "big data" but what should it mean to you? As an exponentially increasing amount of data spews from your company's customer touch points, products and connected devices, more and more pressure has been put on marketers and businesses to turn that into data and information and opportunity. This session walks through the high-level steps and real-world examples required to move from a fractured and disparate data environment to an advanced analytical ecosystem that allows you to drive revenue and personalize the experience of your customers in multiple ways.
Lunch in the Exhibit Hall
Predictive Analytics in the Big Data World: Myths and Realities
This panel of insurance industry leaders examines what works and what doesn't in the new data paradigm. Attend as they identify and discuss "best practice" data analysis approaches, regardless of whether the data being analyzed is "Big" or "Small".
With new data exploding across the digital ether -- such as mobile- and app-generated information -- this discussion focuses on how that data can be used to solve traditional business problems and address new business challenges. The panelists also explore what will change -- and what will not -- with respect to the roles of the data scientist and value architect.
Panelists: Emma Warrilow, President, Data Insight Group
Daymond Ling, Senior Director, Advanced Analytics, CIBC
Tracey Jarosz, Senior Director, Business Intelligence and Analytics, CIBC
Gary Saarenvirta, Chief Executive Officer, makeplain
The Future of Predictive Analytics in Insurance
The insurance industry has always embraced analytics. Through the actuarial discipline, techniques have been employed that allow companies to charge policy premiums which are more accurately reflect risk. But advancements in predictive analytics methods and technologies are rapidly raising the stakes. Here, our panelists discuss how they're using predictive analytics in their pricing strategies. Data challenges and how organizations have overcome them to more fully leverage predictive analytics will be discussed. Attend as our experts examine what lies ahead for organizations that truly embrace predictive analytics as part of their corporate DNA.
Panelists: Barb Addie, President, Baron Insurance Services
Jamie McDougall, Vice President, Personal Insurance Solutions, Gore Mutual
Jim Sulston, Manager, Analytics, North Waterloo Farmers Mutual Insurance Company
Carlos Coutinho, Vice President, Orion Travel Insurance
3:00-3:30pm •Room: 205
Exhibits & Afternoon Break
Track 1: Retail Analytics
Case Study: A Retail Company
Predictive Analytics and Retail: Shaping the Mobile Experience
Attend and learn how predictive analytics is helping one of the world's largest retailers better understand and serve its customers. This presentation details this retailer's challenges, how predictive analytics was employed to solve them, the results they achieved, and how this methodology and information can be applied to other industries and businesses.
3:30-4:15pm • Room: 206AC
Track 2: Regulatory Oversight
Case Study: The British Columbia Securities Commission
Regulatory Oversight Using Predictive Risk Models
The British Columbia Securities Commission worked with an analytics consultant to design and build four Risk Models that prioritize regulatory reviews for major processes, using predictive math. Come hear how a small regulatory organization overcame major hurdles with data and custom user interfaces to implement these models and adopted new ways of overseeing securities market participants.
Track 1: Operational Analytics
Case Study: Canada Post
Realizing the Value of Big Data: Turning Operational Data into Commercial Solutions through the Power of Predictive Modeling
Canada Post, who manages one of Canada's most sophisticated logistics data environments, realized that the operational address data that powers delivery may very well be converted into products and services that could enable not just marketers, but also power e-commerce platforms, MDM solutions, GIS-engines and Risk management. Hear the journey taken as Canada Post transformed an address database into a highly-valuable commercial asset through predictive modelling; segmentation, augmenting of internal data with third party partnerships; and compressive governance rules. Attend as James shares real-life examples of e-commerce, marketing acquisition and risk solutions that were created through Predictive analytics and guided by companies Product Data Architects.
4:20-5:05pm • Room: 206AC
Track 2: Customer Analytics
Integrating Analytics and Research for Improved Consumers Insights
Most organizations conduct research and most companies conduct analytics, but integrating these two practices will drive improved consumer insights. In the session you will see examples of how Customer Satisfaction Research can be validated through Customer Analytics and how the outcome may change the way one views Customer Satisfaction. You will also see examples of how research data and customer analytics have been integrated together to improve insights, improve segmentation and enhance targeting techniques. The session will wrap up with a discussion on how research and analytics departments can work better together in the future.