Cloud Democratizes Access to Big Data Analytics Every organization can discover and act on analytic to deliver superior service Number 74 January 2014 In the quest to provide superior service to customers, a handful of competitors are rising above the rest. These companies have invested heavily in Big Data analytics and real-time decision management to better understand consumer behavior and its context not just what is happening but why. They re using this deeper insight to act with ever-greater speed, relevance and value to customers. Just look at the top two or three performers in the industry of your choice: retail, banking, insurance, health care. Invariably, they are leveraging state-of-the-art analytics in ways that distinguish them from the pack. Yet every organization from startups to multinationals to government agencies and non-profits can achieve this analytic sophistication. How do you compete with a company achieving a 30% response rate? Cloud-based solutions are expanding access to advanced analytics and decision management for organizations of all sizes. Without deeppocket investments, full-time data scientists or massive racks of servers, they can tap into powerful technologies that enable them to serve customers or constituents with keen perception and timely action. And soon, the most sophisticated analytics tools and applications, as well as increasingly diverse data sets, will be put to use to solve problems in ways we can scarcely imagine today. This paper will discuss: How cloud services are democratizing access to Big Data analytics, optimization and decision management. Why rapid application development (RAD) and cloud-based community marketplaces are changing the way we think about analytic investment. How to tap into the apps, tools and infrastructure you need to quickly turn into better operational decisions. How to turn traditional champion-challenger testing into a learning dynamo that drives continuous improvement. Ways to leverage cloud-based analytics for overcoming silos and other obstacles to collaborative problem solving. www.fico.com Make every decision count TM
Big Data Analytics for Everyone Top performers, from industry to industry, are improving performance by investing heavily in Big Data analytics. But cloud-based analytics is leveling the playing field. Take customer experience, a primary competitive battleground for many industries today. Top performers are using analytics to become customer-centric in everything they do. They can adjust their actions at the first signs of changing consumer behavior or attitudes. They re quick to innovate and tap into new markets because they learn fast and aren t afraid of making mistakes. And they re good at mobilizing internal and external resources to solve problems. Here are a few scenarios showing where these leaders are headed using Big Data analytics: Retail Marketing Bank Fraud Protection Health Care Collections What most companies do today Predict customer propensity to purchase product X and send coupon. Call customer to check on a suspicious pattern in credit card charges. Contact all delinquent customers to ask for payment. Where leaders are headed Predict customer will purchase product X next week and send coupon for precise amount (no more) most likely to ensure she buys from you. Deliver offer through the channel at day/time most likely to prompt action. Make companion point-of-sale offer mathematically optimized for balanced benefits in customer loyalty and profit. Send automated message (SMS, email, voice) if transaction is out of pattern and cardholder s mobile phone location is not the same as transaction location. Allow cardholder to verify legitimacy without embarrassment or fuss. If fraud is underway, provide automated resolution or option for human agent assistance. With analytics perceiving travel is with family and predicting needs based on behaviors of similar customers in this situation/locale, automatically offer access to funds plus third-party services so vacation can go on uninterrupted. Predict root causes of delinquencies (overlooked bill, financial stress) and adjust collection strategy accordingly. Know instantly when agents diverge from strategies or compliance rules. Analyze speech patterns indicative of customer attitude and intent. Use these to empirically derive contact/ conversation strategy rules and mathematically identify optimal next step. www.fico.com page 2
»» Figure 1: A new level of computing opens up far-reaching new prospects ICT INNOVATION 2005 2020+ ICT INNOVATION 1985 2005 Billions of Users CIOs LOBs Enterprises SMBs SPs Consumers Emerging Markets Hundreds of Millions of Users Millions of Users Trillions of things Intelligent Industry Solutions 3 rd PLATFORM Mobile Broadband Big Data/Analytics Social Business Cloud Services Mobile Devices and Apps 2 nd PLATFORM LAN/Internet Client/Server PC 1 st PLATFORM Mainframe Terminal Source: IDC Predictions 2013: Competing on the 3rd Platform, Doc # 238044, Nov 2012 Figure 2: FICO s approach to cloud-based services Businesses Entrepreneurs Academics Developers Researchers Governments Banking Solutions Retail Solutions COMMUNITY ANALYTIC MARKETPLACE Health Care Solutions Insurance Solutions Millions of... Apps Services Information Content Experiences Tens of Thousands of Apps Thousands of Apps Systems Integrators/ Consultants ISVs Other Solutions DECISION MANAGEMENT PLATFORM Application Development Productivity Tools Business Rules Analytics Optimization Templates & Frameworks Big Data Analytics Components Analytic Services Data Connectors Visualization INFRASTRUCTURE Compute Storage Big Data Infrastructure Networking Consortium Data Third-Party Data Internal Data With cloud-based analytics, organizations of all types and sizes across industries can aim for superior operational capabilities like these. The cloud reduces the time and money required to develop and deploy advanced data-driven decisioning by up to 80% in many cases. These efficiencies, along with cloud-based opportunities to collaboratively build on the work of others, are changing the economics of analytics investment. There is far less expense and risk involved in trying new ways to solve business problems and deliver better service to customers. The cloud also makes it easier to share learning and replicate successes across organizations, raising return on investments. Moreover, organizations may be able to monetize their work by making it available for fellow cloud participants to build upon in solving other problems and developing unanticipated applications. What has brought about these huge efficiencies and opportunities? The combination of inexpensive standardized hardware components, service-based software architectures, Big Data processing technologies, ubiquitous broadband networks, social media, and intelligent mobile business and consumer devices has created a new evolutionary computing plateau. Solutions built on what IDC calls the 3rd platform (Figure 1) typically offer advantages like elastic on-demand computing power, scalable service-based infrastructure, standardized adapters for accessing data sources, reusable solution components and RAD tools. This is the foundation for cloud-based analytics. FICO s approach to cloud services focuses on faster, smarter, analytically driven customer decisions. So on top of the fundamentals, we have layered a wealth of business-critical technology services (Figure 2). These are focused on increasing data-driven and bringing them rapidly into operational decisions and actions. www.fico.com page 3
»» In the rest of the paper, we ll look at what it takes to join the ranks of the top performers and how the cloud makes the means available to all. We ll focus on three key areas, specifically how to: drive ever-greater from Big Data, accelerate analytic learning and innovate through collaboration.»driving» Intelligent Actions from Big Data Insights CLOUD ADVANTAGE Building a global solution in seven weeks A global express delivery/courier company needed to enhance knowledge sharing across customer accounts, and improve management visibility and control over its various transport solutions. It solved the problem in just seven weeks by using cloudbased RAD tools to develop and deploy an enterprise workflow and supply chain management system. The application, accessible to the organization s entire global workforce, incorporates a knowledge base of customer-centric best practices and structured problem solving for customer requests. Low-cost, high-agility solutions An online European bank needed operational processes that would enable it to fully exploit the cost advantages of being an online-only bank. Using a cloud-based FICO originations solution, the bank implemented lean, automated processes that put control over lending policies in the hands of the bank s business users. The system also incorporates parameter-based functions that make the process of adapting to conditions in new regional markets largely automatic. What are the baseline capabilities required to rapidly pull from Big Data and instigate actions from them? Here are the six most important technology enablers: 1. Big Data infrastructure and analytics. Top performers have implemented the infrastructure and technologies to analyze huge volumes and wide varieties of data. The cloud makes Big Data infrastructure available as a service on a consumption or subscription basis. Any organization can rapidly harness the processing power required for a new initiative in customer centricity, scaling up with growth. A wide variety of analytic technologies are also available, including machine learning algorithms for automated data mining, text and speech analytics, predictive models, and economic impact models. 2. Frequent additions of new data sources and analytic types. Top performers are always looking for ways to deepen and improve perception of customer behavioral context. The cloud makes it easier for all organizations to do what they do tap into new data sources. For one client, analyzing text and combining the extracted from it with from traditional structured data improved predictive model performance by 8%. Top performers also analyze their existing data in new ways. For example, a time-to-event model can examine time log data from point-of-sale systems (cash registers), which many companies have on hand, to predict how likely a customer behavior is to occur within a specified period of time. A retailer that used this technique doubled customer response rates. The cloud will provide access to this and many other analytic technologies. It will also provide tools for importing virtually any type of existing model into operational decision/process flows, as well as for building, testing, deploying and managing new models. 3. No gap between insight and action. As analytics pull context and meaning from incoming data streams, top performers bring this intelligence immediately to bear on operational decisions. To close the gap, they re increasingly embedding analytic decision making into all their customer-facing applications. Cloud analytics provide access to domain-specific applications with embedded decisioning. Or cloud users can tap into decision engines that will execute virtually any combination of analytic models and business rules to perform decision services for other applications, including legacy systems. By leveraging cloud-based analytics, business can ensure consistent customer treatment, even if that treatment is ultimately delivered through diverse, distributed customer touchpoints and legacy applications. The cloud allows critical decisions to be centralized and coordinated, even when customer interfaces are distributed and minimally connected. The cloud also provides tools for business users (rather than back-logged IT shops) to easily control and change decision processes and workflows. www.fico.com page 4
4. Communications-enabled processes. Automated, intelligent communications are at the heart of how top analytic performers work and interact with customers. Analytically driven trigger decisions, which trigger actions that are executed via omni-channel strategies that ensure a seamless customer experience. The cloud provides access to this kind of intelligent customer interaction management. As a result, every organization can gain the benefits of reduced latency in operational processes, increased responsiveness to incoming signals and lower costs. For a European banking client, 78% of customers receiving automated communications said it improved the overall level of service they were experiencing. The bank also saw a 23% rise in ticket value per debit card spend. A retail finance company using automated communications for collections increased right-party contacts by 42% and immediate payments by 30%. CLOUD ADVANTAGE Rapid delivery of award-winning apps A leading South American insurance company needed more agility to efficiently conduct business through a network of 30,000 brokers as well as employers, joint ventures and strategic partners. Cloud access to FICO s business rules management system provides the development environment and decision engine for an award-winning auto insurance underwriting application that enables these diverse users to input data and check premium amounts. Previously the company employed the same lean, cost-effective approach in underwriting applications for both its health and dental insurance businesses. 5. Rapid development of new analytically driven applications. Top analytic performers are innovative in how they serve customers because they can go from good idea to deployed application in just weeks or days. As a result, they can try new things with very low cost, and therefore with much lower project risk. Cloud analytics coupled with RAD tools enable virtually any organization to quickly try new things, at a cost that s up to 80% less than traditional development methods. Correctly coupled, these tools are focused on creating applications that have analytics-driven decisioning at their core. They include domain-specific application frameworks with libraries of predictive customer characteristics, extensible data models, process templates and rapidly adaptable user interfaces. 6. Decision optimization mathematically finding the best win-win for customer and organization. Top analytic performers use science to focus faster on winning decision strategies. Increasingly, they are employing decision modeling with mathematical optimization to identify the best strategy for achieving a goal, such as maximizing profit, while balancing multiple objectives and constraints (time, budget, physical capacity, human resources, etc.). The cloud provides access to optimization engines as well as to decision modeling/optimization templates. Simulation tools clearly show the impact of optimization and aid in the exploration of objective/constraint trade-offs. Performance improvements across a wide range of decisioning use cases new account originations, marketing, collections are generally 15 20%. One retailer used optimization to identify offers that achieved up to 30% response rates. A bank that has developed an optimization culture is driving round after round of improvement in its credit line management strategies. The initial round of optimization sent incremental profit per account up by more than $7 within 12 months of execution. The latest round has already raised it by $5 per account after just four months. The bank is also applying optimization to collections and settlement decisions, as well as to pricing. www.fico.com page 5
»Learning» with Relentless Zeal Along with exceptional abilities to analyze data and take timely action, top performers are also exceptionally good at learning from what happens. Cloud-based analytics enables any organization to learn with this same speed and precision by providing an experiment-friendly environment for accelerated learning. FICO has long advocated systematic test-and-learn cycles using the champion-challenger method of comparing a current best decision strategy with one or more alternatives. Top performers rely on this approach but have thrust it into hyper-drive. They re conducting frequent, well-designed experiments. They re measuring operational outcomes against expectations without delay, assessing impact on detailed key performance indicators (KPIs). And as these leaders continue to accelerate test cycles, they re going to be able to drive learning down to the level of individual customers improving their understanding of not just populations, but people. That will make it tough for competitors that learn more slowly. How can the cloud help you learn faster? First, with cloud-based analytics, the cost of learning designing/implementing experiments and tracking/analyzing results is coming way down. Second, cloud services for decision modeling, simulation and optimization make it much easier to balance all the factors that go into complex strategies and forecast the impact on KPIs like response, purchase and revenue. Another advantage is the ease with which analytic data marts can be set up to capture decisions and outcomes, enabling swift assessment of strategy performance. Without having to wait for full production results, organizations can compare early actual results on KPI metrics against simulations to extrapolate longer-term outcomes. Figure 3: Closing the feedback loop for faster analytic learning Assess Variances between what happened and what was expected present diagnostic learning opportunities. Cloud-based business rules management enables organizations to act on what they learn (and diagnose further) by modifying decision strategies in minutes, without need for IT assistance. Resolve Act Decide Learning is fed back into analytics in multiple ways. Whenever customer activities or interactions yield new data, the cloudbased transactional analytic infrastructure refreshes customer predictive characteristics and profiles. It can generate new customer scores in real time. Analytic management facilities also make it easy to update predictive models from learnings every month, or even every week. In many cases, models can largely update themselves using adaptive or self-calibrating technologies, becoming more accurate with every transaction and adapting in real time without human intervention. Cloud resources also help organizations design better champion-challenger contests. For instance, one way to increase learning yield is to create some challengers that aren t too close to the current champion. Testing challengers www.fico.com page 6
outside the bounds of business as usual introduces controlled variation into the data, expanding what can be learned. It s also helpful to design experiments that yield data revealing not just correlative relationships (when A occurs B also occurs), but causal relationships as well (A affects B in this specific way). This kind of experiment improves understanding of customer sensitivities to offer features such as price, brand status or flexible terms, improving predictions of how individual customers are likely to respond to a subsequent offer or treatment. Cloud-based machine learning tools can generate challengers that produce data containing more controlled variation, as well as data that supports the discovery of causal relationships. At the same time, these tools constrain testing risk and cost by controlling the amount of deviation from the current champion strategy. Using simulation, organizations can therefore find their own sweet spot balancing learning speed and investment for high ROI. Organizations will also be able to access expertise and tools from the cloud, including executive dashboards, for understanding the wider, longer-term P&L implications of decisions. P&L driver assessments, for example, can analyze profit potential across the entire customer lifecycle for a segment or whole portfolio of customers or accounts. They show untapped profit potential and provide an action roadmap for achieving it. For one FICO client, a large retail bank in South America, this process uncovered revenue peaks and valleys. This visibility pointed to opportunities to use decision strategy optimization to offset impending P&L pressures and thereby maintain profit-per-account at target levels. Improvements achieved by other banks include lifts of 6% in total balances, 8% in interest revenue and 7% in riskadjusted return.»»innovating by Community to Solve the Unsolvable Cloud-based analytic services can also help organizations of all sizes eliminate barriers to problem solving. Top performers are very good at marshaling internal and external resources to accomplish an objective. This ability to concentrate knowledge and effort often enables them to break through perceived barriers to solve problems in new ways. Cloud-based analytics will help other organizations do the same. For small and medium-size organizations, barriers to achieving objectives are often lack of technology resources or expertise. Cloud services provide ready access to both. Moreover, the cloud community will make it easy to find and collaborate with other organizations or individuals who have relevant ideas or missing pieces of the solution. For larger organizations, barriers may have to do with lack of agility and speed caused by internal data silos and divisional boundaries that make internal collaboration difficult. Yet organizations may not be large or deep-pocketed enough to invest in the enterprise infrastructures and integration to overcome these impediments. With cloud analytics, they have a new opportunity to overcome undesired organizational impediments, without having to marshal large investments or wait for slow-moving organizational change. www.fico.com page 7
Consider the hub and spoke infrastructure built out by a top-performing North American banking group. This architecture (Figure 4, left side) required considerable up-front investment, but it delivers continuing advantage. As new banks are acquired, they re connected via another spoke to the company s world-class capabilities and best practices. Banks within the group can share decision strategies across geographic markets, comparing results and propagating learning. In fact, strategies are often worked on collaboratively by local and central teams. With cloud-based access to infrastructure as a service (IaaS), organizations can achieve similar cohesion without the upfront investment (Figure 4, right side). Standardized adapters for a wide range of data sources make it easier to overcome organizational silos by bringing the data needed for the decision process into an analytic data mart. Shared repositories (business rules, predictive characteristic libraries, predictive and optimization algorithms, decision model templates, etc.) facilitate learning and help organizations replicate and build on successes. Figure 4: Build collaborative infrastructure or consume it as a service GLOBAL HUB Connected apps Originations Customer management Collections Shared components Interfaces behavior scoring datamart Enterprise architecture High availability disaster recovery 24/7/365 Country spoke Real-time connection Decision Country apps spoke Users Decision Country Tools apps spoke Users Product Tools Decision apps lines Hosts Users Product lines Tools Hosts Product lines Hosts Connected decision apps Analytic services Shared components, frameworks, templates High-availability infrastructure as a service Country spoke Real-time connection Country spoke Country spoke The solution frontier As collaboration barriers come down, the possibilities for problem solving expand in all directions. Cloud-based analytics will give many more organizations and individuals access to analytic algorithms and methods. This wider sharing of both best practices and cutting-edge techniques will help drive huge productivity gains in the exploration of new technologies and applications for it. www.fico.com page 8
HOW FAR CAN WE GO? Let s indulge in a little blue-sky thinking. As it becomes easier to share data and combine what we know with what others know, barriers to finding solutions will fall. Greater social impact from charitable spending. Grant-making foundations currently struggle with how to allocate funding to maximize success. Cloud-based data sharing and analytics could help them make better-informed grant decisions, replicate proven successes and accelerate progress against their missions. Cures for rare diseases. To justify the large expense of searching for cures, medical and pharmaceutical researchers focus on diseases that afflict sufficient numbers of people. But shared data and access to new kinds of analytics may bring down costs for some kinds of research. That could get researchers thinking about new ways to approach smaller-scale disease problems, which may not have attracted yesterday s investment. And collaboration isn t just shared work processes; it s also building on the work of others. As such, collaboration may take place asynchronously and long after the original work. There s a wonderful example of this broader kind of collaborative power in the way Dr. Max Little, a Postdoctoral Research Fellow at the MIT Media Lab, has extended mathematical algorithms for detecting voice disorders to the problem of diagnosing Parkinson s Disease. Using voice signal processing and machine learning algorithms developed by others for unrelated purposes, ParkinsonVoice.org is now providing an online application that enables people all over the world to self-diagnose early signs of the condition. But even on a more routine business level, the potential for collaboration is encouraging. Every technology company, of course, has a finite set of markets and domains it knows a lot about, and FICO is no exception. Yet the potential applications for our technology extend much farther, and the openness and community of the cloud encourages other organizations to use what they know to realize these opportunities. Take, for example, the FICO LiquidCredit Service. We originally developed this web-based application to help small and medium-size banks make faster, more accurate originations decisions. A partner company later adapted the application to the needs of the auto finance industry. In the FICO Analytic Cloud, LiquidCredit Service will be one of many available on-demand services. Members of the Cloud community will have the opportunity to combine the application with their own or third-party technologies to solve other problems in new ways. FICO s aim is not only to allow extension and repurposing of applications, but also of the technology components they comprise. For instance, adaptive analytics is a key technology in FICO Falcon Fraud Manager. When an adaptive model is layered onto the fraud model, it enables fraud detection to dynamically learn from behaviors being currently observed in the production environment. This technology could be applied to other kinds of decisions in rapidly changing environments. Wherever decision outcome data is being captured for analysis Did the collections call result in a payment? Did the discount offer result in a purchase? adaptive analytics could enable models to automatically learn from these current results. This is a glimpse of the potential for extending analytic decision making. As FICO, our partners and other members of our FICO Analytic Cloud community make technologies available for others to build on, expect the rate and range of innovations to multiply. One member of the community has some data, another has an idea for how to use it to solve a new problem, and others have just the analytic or software components needed to do it. Stay tuned it gets really exciting from here. www.fico.com page 9
»Conclusion» Customer expectations for products and services are being set by a handful of leaders who have invested heavily in analytics-driven decision management. Increasingly, your customers and constituents simply take it for granted that you know them just as well as these top performers do, and are acting in accordance with their individual needs and situations. Soon, cloud-based analytics will mean every organization can meet these expectations. Learn more: Get further details about the FICO Analytic Cloud Read other Insights white papers on Big Data analytics: Extracting Value from Unstructured Data (No. 71) Is It Fraud? Or New Behavior? (No. 69) When Is Big Data the Way to Customer Centricity? (No. 67) Subscribe to the FICO Labs Blog for the latest on Big Data, cloud computing and other technology innovations The Insights white paper series provides briefings on research findings and product development directions from FICO. To subscribe, go to www.fico.com/. For more information North America toll-free International email web +1 888 342 6336 +44 (0) 207 940 8718 info@fico.com www.fico.com FICO, LiquidCredit, Falcon and Make every decision count are trademarks or registered trademarks of Fair Isaac Corporation in the United States and in other countries. Other product and company names herein may be trademarks of their respective owners. 2014 Fair Isaac Corporation. All rights reserved. 3053WP 1/14 PDF