Unleash SAP Hana Predictive Power. Alexis Fouquier BI & Data Warehousing, SAP EMEA November 2012

Transcription

1 Unleash SAP Hana Predictive Power Alexis Fouquier BI & Data Warehousing, SAP EMEA November 2012

2 Agenda The Business Case for Advanced Analytics SAP s Predictive Analytics Strategy Predictive Analytics with SAP HANA Customers Benefit from Predictive Analytics with SAP HANA Predictive Applications for SAP HANA Summary

4 What if... You could identify hidden revenue opportunities within your customer base through predictive analytics?... You could retain your high-value customers/employees/vendors/partners with the right retention offers?... Your call centre agents could delight customers with the best next-step recommendations?... You could increase cross-sell and up-sell effectiveness through cross-channel coordination?... You could build long-term customer/employee/vendor/partner relationships with intelligent interactions?

5 Competing in today s demanding marketplace means going beyond what happened MANUFACTURING l RETAIL l HEALTHCARE l BANKING l UTILITIES l TELCO PUBLIC SECTOR FINANCIAL SERVICES OPERATIONS l HR l FINANCE IT SALES l MARKETING What happened? How and why did it happen? What is happening now? What is the risk if it does/doesn t happen? What will happen? How do you prevent / ensure it happens again? Tom Davenport International Institute for Analytics

6 Today s analytics reality Specialization Predictive Modeling Data Mining Generic Predictive Analytics EIM Raw/Clean Data BI Reporting, Discovery, Analysis, Dashboards Standard Reports Sophistication/Skill Set There is a disconnect between the BI and Advanced Analytics world

7 Extend Your Analytics Capabilities Sense & Respond Predict & Act COMPETIVE ADVANTAGE Raw Data Cleaned Data Standard Reports Ad Hoc Reports & OLAP Generic Predictive Analytics What happened? Predictive Modeling Optimization Why did it happen? What will happen? What is the best that could happen? ANALYTICS MATURITY The key is unlocking data to move decision making from sense & respond to predict & act

8 Fresh Direct Our Food is Fresh. Our Customers Are Spoiled

9 From 91% to 99% on time

10 Coinstar Inventory Optimization Real time Offers Servicing

11 How Predictive Analytics is Used in Industry Business and Industry Use Cases where SAP has helped Healthcare Predict likelihood of disease to begin early treatment; identify clinical trial outcomes. CRM Sales Identify revenue forecast based on customer opportunities and pipeline execution. Banking Identify key behaviors of customers likely to leave the bank; improve credit risk analysis. Utilities Forecast demand and usage for seasonal operations; provide anticipated resources. CRM Marketing Identify potential leads among existing customers and intelligently market to them based on individual preferences and histories Government Predict community movement and trends that affect taxing districts; anticipate revenue. Retail Intelligent selection of store locations based on demographics; inventory planning. Telco Forecast demand on system load for capacity planning and customer scale.

12 Changing Landscapes and New Opportunities Data mining and predictive have been around for decades. But new market forces are changing the landscape and offering new opportunities Increased business interest Now that BI users know what happened, they are asking why and what s likely to happen next Explosive demand from sales, marketing, and call center analyses fraud, and government intelligence/security agencies Increased data value (e.g., Big Data) Exploding data volume Expanding data varieties Increasing technology performance Parallel processing, faster CPUs, and in memory technologies reduce time and cost of data processing

14 SAP s Predictive Analytics Strategy Empower the Business In time Actionable Insights In Context Extend the Business Intelligence competency to Advanced Analytics Embed Predictive into Apps and BI environments Lend expertise In-memory processing No data latencies Big Data ready Relevant to your business Within the context of your Industry and LOB scenario Custom applications Real time in memory predictive and next generation visualization and modeling

16 Predictive Analytics with SAP HANA Transforming the Future with Insight Today Unleash the value of Big Data through the power of SAP HANA Employ in-database predictive algorithms Access 3,500+ open-source algorithms via R integration for SAP HANA Intuitively design and visualize complex predictive models SAP Predictive Analysis software Bring predictive insight to everyone in the business Embed within business applications Extend into BI and reports Insight into events instantly delivered to dashboards, alerts, and mobile devices

17 SAP HANA In Memory Predictive Analytics Combine the depth and power of in-memory analytics within SAP HANA with the breadth of R to support a variety of advanced analytic and predictive scenarios Predictive Analysis Library (PAL) Native predictive algorithms In-database processing for powerful and fast results Quicker implementations Support for K-Means, K-Nearest Neighbor, C4.5 Decision Tree, Multiple Linear Regression, Apriori, ABC Classification, Weighted Score Tables R Integration for SAP HANA Enables the use of the R open source environment (> 3,500 packages) in the context of the HANA in-memory database R integration enabled via high performing parallelized connection R script is embedded within SAP HANA SQL Script Y X Z

18 Predictive Analytics with SAP HANA The Predictive Analysis Library in SAP HANA The Predictive Analysis Library (PAL) is a built-in C++ library to perform in-database data mining and statistical calculations, designed to provide excellent performance on large data sets. The Predictive Analysis Library uses tables as the data interface. Three types of tables are used for each PAL function: Input data table supplies input data to PAL functions. Each function can have one or more input data tables. Parameter table supplies parameters to PAL functions. Parameters are special inputs that control what the functions do. Each function has only one parameter table. Output data table holds the output results calculated by PAL functions. Each function can have one or more output data tables.

19 SAP HANA In Memory Predictive Analytics Predictive Analysis Library (PAL) Supported algorithms include K Means Moving Average - Centre and Trailing C 4.5 Decision Tree Moving Median K Nearest Neighbour Single Exponential Smoothing Multiple Linear Regression Double Exponential Smoothing Apriori Triple Exponential Smoothing ABC Classification Exponential Regression Weighted Scores Table Bi-Variate Geometric Regression CHAID analysis Bi-Variate Logarithmic Regression Logistic Regression Inter-Quartile Range Test - Tukey s Test Seasonal Decomposition Y X Z

20 R Integration for SAP HANA What is R? R is a software environment for statistical computing and graphics Open Source statistical programming language Over 3,500 add-on packages; ability to write your own functions Widely used for a variety of statistical methods More algorithms and packages than SAS + SPSS + Statistica Who s using it? Growing number of data analysts in industry, government, consulting, and academia Cross-industry use: high-tech, retail, manufacturing, CPG, financial services, banking, telecom, etc. Why do they use it? Free, comprehensive, and many learn it at college/university Offers rich library of statistical and graphical packages

21 R Integration for SAP HANA Adoption R is used by the most data miners

22 R Integration for SAP HANA Functionality Overview R integration for SAP HANA enables the use of the R open source environment in the context of the HANA in-memory database Establishes a communication channel between HANA and R for fast data exchange Improved data exchange mechanism supports transfer of intermediate database tables directly into vector oriented data structures of R. Performance advantage over standard tuple-based SQL interfaces with no need for data duplication on the R server Supported by providing the R-operator as a custom operator in the calcmodel of the calculation engine This allows the application user to embed R script within SQL script and submit entire query to the HANA database As the plan execution reaches R-node, a separate R runtime is invoked using Rserve and input tables of R node passed to R process using improved data transfer mechanism

23 R Integration for SAP HANA Functionality Overview Embedding R scripts within the SAP HANA database execution Enhancements are made to the SAP HANA database to allow R code (RLANG) to be processed as part of the overall query execution plan This scenario is suitable when the modeling and consumption environment sits on HANA and the R environment is used for specific statistical functions Sample Code in SAP HANA SQLScript DROP TABLE "spamclassified"; CREATE COLUMN TABLE "spamclassified" LIKE "spameval" WITH NO DATA; ALTER TABLE "spamclassified" ADD ("classified" VARCHAR(5000)); DROP PROCEDURE USE_SVM; CREATE PROCEDURE USE_SVM( IN train "spamtraining", IN eval "spameval", OUT result "spamclassified") LANGUAGE RLANG AS Send data and R script 1 3 Get back the result from R to SAP HANA 2 Run the R scripts BEGIN library(kernlab) model <- ksvm(type~., data=train, kernel=rbfdot(sigma=0.1)) classified <- predict(model, eval [,- (which(names(eval) %in% "type"))]) result <- as.data.frame(cbind(eval, classified)) END; CALL USE_SVM("spamTraining", "spameval", "spamclassified") WITH OVERVIEW; SELECT * FROM "spamclassified";

24 Demo - Dataset used is hosted by SAP Hana - Predictive analysis is performed by SAP HANA Predictive Analysis Library (PAL).

26 Mitsui Knowledge Industry Healthcare Speed Research & Improve Patient Support 408,000x faster than traditional diskbased systems in a technical PoC 216x faster by reducing genome analysis from several days to only 20 minutes making real-time cancer/drug screening possible Business Challenges Reduce delays and minimize the costs associated with new drug discovery by optimizing the process for genome analysis Improve and speed decision making for hospitals which conduct cancer detection based on DNA sequence matching Technical Implementation Leveraged the combination of SAP HANA, R, and Hadoop to store, pre-process, compute, and analyze huge amounts of data Provide access to breadth of predictive analytics libraries Benefits For pharmaceutical companies, provide required new drugs on time and aid identification of driver mutation for new drug targets Able to provide a one stop service including genomic data analysis of cancer patients to support personalized patient therapeutics Our solution is to incorporate SAP HANA along with Hadoop and R to create a single real-time big data platform. With this we have found a way to shorten the genome analysis time from several days down to only 20 minutes. Yukihisa Kato, CTO and Director of MITSUI KNOWLEDGE INDUSTRY

27 Cisco Systems, Inc. High-Tech Deliver Predictive Insight in Near Real-Time 5 seconds Seasonality analysis with any selected filters including country, segments and sales levels Business Challenges Provide Cisco sales with real-time insights to make better decisions Better data delivery decisions to improve the business - for optimization and to drive topline growth Lack of understanding of underlying sales performance drivers and the seasonality of buying patterns 2 seconds Cluster analysis across multiple quarters for a combination of sales level and quarter to determine sales achievement level The HANA platform at Cisco has been used to deliver near real-time insights to our execs, and the integration with R will allow us to combine the predictive algorithms in R with this near-real-time data from HANA. The net impact is that we will be able to take the capability which takes weeks and months to put together, and deliver just-in-time as the business is changing. Piyush Bhargava, Distinguished Engineer IT, Cisco Systems Technical Challenges Unable to transform the results of statistical analysis into actionable insights Benefits Better respond to customers and deliver tangible business value Reduce time to transform information into insights and improvement in the quality of decision-making on those insights Ultimately drive higher profitability and growth

28 Bigpoint Gaming Industry - Predictive Game Player Behavior Analysis 5,000 events per second loaded onto SAP HANA (not possible before) 10-30% increase in revenue per year Interactive data analysis leading to improved design thinking and game planning Business Challenges Increase conversion rates from free paying player Increase the average revenue per paying player Decrease churn keep paying players playing longer Technical Challenges Leverage real-time data processing in SAP HANA and classification algorithms with R integration for SAP HANA to deliver personalized context-relevant offers to players Analyze vast amounts of historical and transactional data to forecast player behavior patterns Benefits Real-time insights Per player profitability analysis and increased understanding of player behavior Increase data volume and processing capabilities to communicate personalized messages to players At Bigpoint in the Battlestar Galactica online game, we have more than 5,000 events in the game per second which we have to load in SAP HANA environment and to work on it to create an individualized game environment to create offers for them. In this co-innovation project with SAP HANA, using Real Time Offer Management Bigpoint, we hope to increase revenue by 10-30%. Claus Wagner, Senior Vice President SAP Technology, Bigpoint

30 SAP Predictive Analysis Integration with SAP HANA Leverages SAP HANA s native predictive capabilities Pushes predictive processing into SAP HANA Intuitively design complex predictive models Visualize, discover, and share hidden insights

31 Demo Design a Predictive model with SAP Predictive Analytics

32 Smart Meter Analytics Identify energy consumption pattern that can be used for Customer Segmentation Smart meter data volume is huge Using SAP HANA + PAL K- Means algorithm Clustering of smart meter data applying the k-means clustering to >20 million x 96-dimensional vectors High performance clustering computation Challenge Solution Results

33 Predictive Analysis in HPA Customer Analytics

34 Customer Analytics Account Intelligence Target User: Sales Manager, Sales Rep with real-time and complete view of all Customer activities Comprises Sales, Financials, CRM, and external sources like customer satisfaction, social media data Audience Discovery and Targeting Target User: Marketing, Sales Manager rapidly and easily segment large customer populations Manage Target Groups for insight to action Trigger initiatives like customized offers for each segment and channel Customer Value Intelligence Target Users: Marketing Manager, Product Manager Deep insight into revenue and profitability of customers, products and channels.

35 Predictive Use Cases inside Customer Analytics Account Intelligence ABC Classification Product Recommendations White Space Analysis (Customer Stratification) Sales Pipeline Forecasting Probability to churn Audience Discovery and Targeting ABC Classification (parameterized) RFM Analysis Cluster Analysis ( ) ( ) ( ) Decision Tree Analysis Customer Profiling (similar behavior for campaigns) Customer Value Intelligence Customer Lifetime Value Product Affinity: Propensity score (e.g. probability that a customer will buy a certain product model based ( ) prediction) Cross selling (Next best product e.g. customer profiling with clustering and decision tree; association analysis) Covered in development system; ( ) under development

36 Performance and Insight Optimization Services Enabling complex use cases with SAP HANA Examples of innovative business solutions advanced by SAP HANA: Tax Compliance and discovery engine for public sector Industry-specific optimization Algorithms and processes Affinity insight and forecasting for retail Industry-specific predictive modeling Proprietary models for retail, banking, utilities, manufacturing, and beyond High performance compliance engine for chemical companies Back end Sophisticated aggregation and cleansing algorithms High volume industry specific data relationship algorithms SAP HANA

37 Affinity Insight for Retail Challenge Tailored to the roles of retail category managers and merchandising managers Both of whom are concerned with product relationships and their associated financial performance in achieving retail corporate targets. Solution Using SAP HANA and the PAL Apriori algorithm with modification for maximal performance for the target usage Results See affinity relationships easily on a broad scale Understand product relationships that contributed to affinity relationships Conduct simple what-if queries against a demand model See store-level performance relative to the product dimension.

39 The Bottom Line Companies can no longer focus solely on delivering the best product or service To succeed, they must: Uncover hidden customer / employee / vendor / partner trends and insights Anticipate behavior and take proactive action Empower your team with intelligent next steps to exceed customer expectations Create new offers to increase market share and profitability Develop and execute a customer-centric strategy Target the right offers to the right customers through the best channels at the most opportune time

40 Thank You Thank You! Contact information: Alexis Fouquier, BI & Data Warehousing, SAP

41 Appendix

42 SAP HANA adoption model A platform that scales Localized SAP HANA as a data mart Fast time to value Bring data into SAP HANA for a high value scenario Flexible and extremely fast realtime analysis Non disruptive adoption Pervasive Analytics SAP HANA as a platform for all your analytics Multiple scenarios Consolidation database that scales with multiple nodes Unstructured data analysis (e.g. Text analytics) Predictive analytics Analytics for mobile users Globalized SAP HANA as a platform for analytics and applications Genomic DNA analysis in real-time to transform cancer patient care Increased speed, accuracy and visibility for drug discovery Real-time Big data (R+Hadoop+HANA) 408,000 faster than traditional disk-based system

43 SAP HANA Hadoop Support in SAP Data Services 4.1 Extend the Big Data Ecosystem to SAP HANA Files SAP HANA, Sybase IQ, other Target Systems Databases SAP Data Services SAP Data Services Web & Others SAP HANA can integrate with Hadoop via SAP Data Services 4.1 Ability to load and read large volumes of data into and out of SAP HANA from/to Hadoop via SAP Data Services 4.1 Reading from and loading into Hadoop Hive Support (table-like access to Hadoop data) Direct HSFS file support through PIG scripts

44 SAP HANA text search capability Native full text search Exploit unstructured content in SAP HANA Leverage one infrastructure for analytical and text search workloads in both OLAP and OLTP use cases Reduced duplication, latency, and operational overhead Graphical modeling Easy to use search definition Built into existing SAP HANA Modeling tools UI UI Components UI Components UI toolkit Rapid development of search enabled applications through reusable UI building SAP HANA Metadata Suggestion Search Metadata Information Access Services blocks Search Fuzzy Ranking Snipets Modeler Search Search Engine Model HANA Studio Tables Column Store Linguistic Processing Text Processor

45 SAP Visual Intelligence for SAP HANA Enable business users to visualize & manipulate data quickly 1. Fast Get answers at high speed 2. Engaging Empower people to interrogate data 3. On Massive Data Volumes Add greater information context Real time on detailed data on the fly Fewer layers, simpler landscape, faster ROI Intuitive discovery and analysis experience Self service visualizations and analytics Ask any question Insight from structured and unstructured data On operational data for agility across your value chain

46 R Integration for SAP HANA Business Benefits Advanced Analytics Users can leverage the range of statistical functions from R and the power of SAP HANA for new advanced analytics use cases Performance Advantage Achieve better business results through faster predictive analytic cycles on the SAP HANA platform and via the R integration for SAP HANA Ease of Use Analysts already using R will be able to use functionality without any additional training IT Benefits Total Cost of Ownership Continue to leverage existing investments in SAP HANA and R Flexibility Flexible and configurable integration scenarios provide options whether to use the predictive capabilities of SAP HANA and/or R Enterprise Fit Coexists within an organization s IT infrastructure R provides an incredibly comprehensive library of predictive analysis algorithms SAP provides an efficient integration between SAP HANA and R Customers are empowered to apply this range of analysis to SAP HANA data