View stunning SlideShares in full-screen with the new iOS app!Introducing SlideShare for AndroidExplore all your favorite topics in the SlideShare appGet the SlideShare app to Save for Later — even offline
View stunning SlideShares in full-screen with the new Android app!View stunning SlideShares in full-screen with the new iOS app!
Real AnalyticsShifting from business hindsight to insight to The crunchy questions haunting the business requireforesight a combination of hindsight, foresight and insight. ByIn 2010, many organizations began to see information investing in a balance of information management,automation outweigh business process automation as performance management and advanced analytics,their highest priority area1. In the reset economy, analytics organizations can make small steps, smartly madeoffered improved visibility to drive operational efﬁciencies, to capture measurable results. These can span fromas well as a platform for growth by addressing heart-of- improving fragmented customer relationships by analyzingthe-business questions that could guide decisions, yield omni-channel interactions, to providing an integratednew insights and help predict what’s next. It seemed enterprise view of risk and ﬁnance. This is the essencelike a no-brainer. But companies quickly discovered of real analytics: delivering business value through thethat the journey is complicated – requiring a clear continuous build-out of core information disciplines.analytics vision aligned to the business strategy, severallayers of supporting capabilities and the fortitude to While getting the right answers is critical, manyembed analytical thinking across multiple facets of the organizations don’t normally know the right questionsorganization. In 2011, leading organizations are launching to ask to get there. Powerful new tools and supportingbroad initiatives with executive-level sponsorship, ready infrastructure have removed most technical constraints,and eager to achieve their vision via real analytics. but analytics initiatives continue to suffer from the lack of a clear vision and commitment to embed analytics-basedData volumes continue to explode, doubling every 14 approaches into how work is performed. Real analyticsmonths2. Regulators are demanding deeper insight into can add knowledge, fact-based predictions and businessrisk, exposure and public responsiveness. Public and prescriptions – but only if applied to the right problems,private organizations alike are feeling increased pressure to and only if the resulting insight is pushed into action.achieve proﬁtable growth. New signals are evolving thatcontain crucial information about companies and markets You can’t drive your car with only the rear view mirror –– including sensor-laden assets, unstructured internal data equivalent to historical reporting. You use the view outand external sentiments shared via social computing3. of the windshield and the dashboard gauges – EnterpriseCisco estimates that the amount of data ﬂowing over the Performance Management (EPM) and performanceinternet each year will reach 667 exabytes by 20134. The dashboards. In fact, many drivers take advantage ofmagnitude and complexity of global businesses have made navigation systems fed by GPS to see the road ahead andit even more difﬁcult for leaders to uncover hidden insight. direct the next turns. That’s like going from descriptive analytics to predictive and prescriptive. That’s moving from hindsight to insight to foresight for the business. 1
History repeating itself?Over the past two decades, companies have invested heavily in back-ofﬁce systems to automate their business processes.Information investments were typically siloed, static, historical and focused only on operational reporting for pockets ofthe business. Real analytics is focused on a more holistic, forward-looking approach, positioning information as an asset tosupport effective business decision and action. What were the challenges? What’s different in 2011? ERP-based information • Large-scale packaged technologies form the • ERP providers have invested in adding information repositories foundation of many organizations’ system platforms to their solution sets, including perform- footprints. Embedded reporting and perform- ance management and some advanced analytics ance management tools were leveraged to tools. These are largely integrated into the core try to meet information needs. However, process automation solutions. solutions were mostly backwards-looking, with minimal real-time dashboarding and • Integration between internal and external systems limited advanced analytics. has been eased by adoption of open architecture standards and advancements in transactional and • Most organizations have a hybrid applica- view-based integration tools. tion landscape, with multiple ERP instances and tens or hundreds of ancillary systems that execute end-to-end business processes. Visibility conﬁned to the ERP transactional store was insufﬁcient for true business insight, but integration to other systems was costly and complex. Business intelligence/ • Performance improvement has been a critical • Leading organizations have adopted a combination reporting/data warehousing part of the real analytics journey, but it is not of performance improvement, information manage- sufﬁcient because it lacks vehicles to guide ment and advanced analytics to meet the needs of insight and foresight. the business. • Organizations often faced multiple isolated, • Enterprise-wide governance is a critical dimension competing initiatives buried within business of real analytics, allowing for visibility across and units, functions and geographies – creating beyond organizational boundaries. confusion and multiple versions of the truth. • Real analytics efforts are embedded in business • Results of information efforts were only loosely processes with executive and management linked to operations and decision making, support, with continuous feedback loops so limiting the amount of value realized. that actual performance can guide the next iteration of analysis. • Technical constraints forced the segmenta- tion of information repositories into discrete, • A combination of improvement in storage, federated views. Complex operations, inte- processing and network performance, as well as grated views or even traversing of data advanced new options for dealing with complex sets were compromised. calculations on large data sets (e.g., high-per- formance information appliances, column-based • Analysis was characterized by small datasets in-memory databases, distributed computing with variables between 10-20 and limited cases tailored for data processing). For example, a (<100), driven by unrealistic assumptions that large consumer credit card issuer recently datasets were linear, normal and independent. analyzed two years of data (73 billion transactions Data universes were restricted to static data across 36 terabytes of data) in 13 minutes. In the snapshots interred by a handful of tools (e.g., past this transaction would have taken more than SQL, SAS). one month5. • Analysis now routinely handles massive datasets with millions of variables and billions of cases, increasingly in real time. Tools such as PMML, DMQL, SPSS and DMX allow the focus to be on exploratory analysis to discover relevant patterns, trends and anomalies in data, without having an explicit goal in mind. 2
Technology implicationsReal analytics represents a combination of information management, performance improvement and advancedanalytics. Each of these capabilities has a number of critical underlying technical implications, with interdependenciesrequiring an enterprise information architecture spanning the entire stack. Topic Description Information Tools for establishing trusted foundational data are essential. These include master data management management for maintaining data correlation, consistency of semantic meaning, providing matching services to identify and link identical entities and enabling bidirectional updates across systems of record. Data quality is also a concern, requiring tools to monitor, analyze, report and scrub. Finally, tools to manage data governance are needed, and should be tightly linked with master data and data quality solutions. Performance This drives to the heart of monitoring, reporting and recommending action by combining historical improvement reporting, business intelligence and dashboards. Technical implications include report design, business rule development, business process integration and the development of dashboards and scorecards. Increasingly, performance management solutions also include mobile delivery channels, either through Web-based outputs or dedicated applications6. Advanced analytics Advanced analytic tools enable predictive modeling, embedding analytics into business processes, discovery and information visualization7. This work typically involves advanced statistical modeling and correlation of widely disparate data sets, requiring access to internal and external data. Infrastructure Complex analysis on large data sets requires a high performance computing environment. Options in 2011 include on-premise appliances, in-memory column-based databases and cloud-based options for elasticity and distributed processing. 3
Lessons from the frontlines They ﬁrst analyzed over 16 million claims and mergedShining a light on addressable markets data from multiple sources which had never beenWhile a specialty insurer’s sales were growing 20 percent harmonized – both within the company and fromannually with 80 percent customer retention, the external business credit and “ﬁrmographic” sources.company had captured only six percent of the annual Then, using advanced modeling techniques such aspotential in its market. In order to grow the business, decision trees, association rules, logistic regression andthe company wanted to better understand its customer Benford’s analysis, the company developed ﬁve setsbase with the goals to improve retention among current of new rules capable of identifying potentially false orpolicyholders using targeted communication and cross- inappropriate claims. The organization added the newselling, as well as identify potential new customer rules to their existing process in order to improve bothsegments that were more likely to purchase its policies. the accuracy of its warranty claim reviews and the return on investment from the operation. The post-analysisTo supplement what the insurer already knew about predictive beneﬁts were more than three times largerits customers, third-party market segment information than initial expectations. Also, as an unexpected beneﬁt,(biographic, demographic, psychographic), enhanced the company was able to create a list of pre-approvedcensus and other external data were used to append labor operations and parts for a given repair accordingnearly 300,000 policies and 150,000 customers. to speciﬁc make, model and year – drastically reducingCluster analysis was used to identify primary customer the need for reviewing and interpreting every claim.groupings and segments, and, ﬁnally, decision treeanalysis was completed to differentiate those segments Delivering on the premiumthat produced the highest value for the company. A leading insurance provider saw its core business beingThey added a market penetration study to compare pushed to deliver more personalized services at lowerexisting and potential market share by segment. costs – while facing increased transparency for demands, growing commoditization of its product offerings andFor the ﬁrst time, the insurer had a uniﬁed view of its overall slowing industry growth. Their response? A multi-customer bases as well as insights on customer behavior, year analytics program to increase sales effectivenesspreferences and lifestyles – all useful in creating new and operational efﬁciency, increase customer retentionup-sell, cross-sell and retention strategies and focusing and better support executive decision making. Predictivegrowth on speciﬁc consumer segments and regions. modeling driven by online, agent and customerAs a result, the company was able to increase product feedback was the cornerstone of the effort – with apurchase loyalty and growth among key customer core analytics competency center built to support needssegments and attained measurable improvements, across business units. The results speak for themselves:including migrating core segments to higher proﬁtability. improved policy retention by 300 basis points, increased acquisition rates on abandoned quotes by 200 basisAdvanced auto(motive) analytics points and advances in customer satisfaction rates.By replacing manual, rule-based warranty claim reviewswith scientiﬁc, automated methods, an automakersigniﬁcantly improved its warranty claims adjudicationprocess with the ability to preemptively identifypotentially false or inappropriate (improper, exaggerated,embellished) claims. The legacy warranty claimsadjudication process relied heavily on manual reviews,exception reports and static rules. The automaker neededa more scientiﬁc approach to adjudication that couldhelp it identify potentially fraudulent or false activity in amore efﬁcient, automated manner and also enhance itsrule set with more sophisticated statistically groundedrules that are too complex for manual processes. 4
Where do you start? • Right ﬁt analytics. Match statistical models andFew organizations are starting from scratch when it comes analytics techniques to the job at hand. Overpoweredto real analytics. Many organizations have decades of solutions waste time and money. Underpoweredexperience with information-related initiatives in various solutions can miss important insights. Buy what youforms. Because of its wide scope, however, real analytics need and use what you buy – across tools and services.initiatives require special attention to dependencieson in-ﬂight efforts. While speciﬁc steps vary company • Accelerate insights. Automate delivery of theby company, some fundamental principles apply: information people need to do their work and automate responses whenever possible, so that• Crunchy questions. Start by laying out speciﬁc, action is taken with more certainty and at the heart-of-the-business questions. Prime your business lowest possible cost. leaders with ideas from other industries showcasing how unstructured and external data can be applied in • Behavior change. Recognize that a big part of practical terms. Then prioritize according to what drives the impact of real analytics will be creating a fact- value, where returns will likely be higher and the degree based culture that embraces its repercussions, to which results can be made actionable. allowing analytics capabilities and outputs to be embedded into operational processes across the• Start where you are. Assess your current capabilities enterprise and up and down the organization. and get a clear picture of the gap between what your organization can do and what it needs to do. Think in • New talent. Institutionalizing real analytics will require terms of both technical capabilities and organizational new skills, including pockets of creative design, deep depth. Grade yourself, prioritize projects aligned to mathematics, statistics and behavioral change skills. crunchy questions and ﬁll the cracks – both with Develop a strategy for how to locate it, develop it and small, focused efforts and with some cross-functional retain it. investments (e.g., enterprise data management). 5