SAP Predictive Analysis Overview & demo SAPSA 2014 yann chagourin y.chagourin@accenture.com
Content Predictive analytics: what is it? SAP Predictive Analysis: solution overview Demo: customer churn prediction in telecom Use-case Copyright 2014 Accenture. All rights reserved. 2
Predictive analytics: what are we looking at? What are my typical shopping baskets? Which pieces of my equipments will fail, when, and where? What are the characteristics of the customers who are churning? Which of my claims are likely to be frauds? What kind of similar groups of customers do I have? Are there predictable patterns in the variations of my customers demand? Copyright 2014 Accenture. All rights reserved. 3
SAP Predictive Analysis: algorithms out of the box association clustering classification outliers classification forecasting Copyright 2014 Accenture. All rights reserved. 4
Elements of project methodology: CRiSP-DM model Copyright 2014 Accenture. All rights reserved. 5
SAP Predictive Analysis: solution overview Rich client, integrated with Lumira Data sources: Excel, Hana, free-hand SQL Algorithms run locally or on Hana Integrated with R Roadmap: merge with KXEN products Copyright 2014 Accenture. All rights reserved. 6
SAP Predictive Analysis: architecture PA + Hana standalone PA PA client model PA client model run PAL Data Data Data Hana run Copyright 2014 Accenture. All rights reserved. 7
SAP Predictive Analysis: demo Scenario for the demo: prediction of customer churn in telecom area Based on a simplified data set, from actual data from The Center for Customer Relationship Management at Duke University, used in a Duke/Teradata tournament in 2003. PA version used: 1.19.0, from SAP s trial offer Copyright 2014 Accenture. All rights reserved. 8
SAP Predictive Analysis: demo A powerful and relatively easy-to-use tool Strong vizualization capabilities Not a magic bullet, data knowledge will be important: initiate patterns recognition understand & validate the proposed models integrate with business processes Copyright 2014 Accenture. All rights reserved. 9
Use-case: Warranty POC Scenario VELUX, a global company within building materials, have just migrated their entire BW solution to HANA. They were now looking how to leverage the new environment and to harness the power that HANA brings, including the SAP Predictive Analysis. In order to gain momentum at the start of the Predictive Analysis project and to validate the use case and benefits, a Proof of concept (PoC) was initiated leveraging the Accenture Innovation Center for SAP and the Accenture Warranty Analytics team. Challenge Solution Results and Benefits VELUX gives a long warranty on their products. Today they forecast manually on how much they will need reserve for warranty cost going forward. By improving the warranty forecast the client hope to achieve the following: With a more accurate forecast, free up funds and resource to be used elsewhere Get a deeper understanding of which defects drives the warranty claims and identify how these can be reduced Move away from a manual and individual-dependent solution By analyzing the data with Predictive Analysis, identify trends in defects and mitigate those before they have a larger impact. By analyzing production defects by plant, get a holistic view of production quality across the company and identify areas that need attentionty across VELUX and identify areas that needs attention The PoC demonstrated the capabilities of the SAP Predictive Analysis tool and the Warranty Analytics expertise from IDC analytics team The PoC was conducted using VELUX data exported to the Innovation Center environment into an SAP HANA data model. The Analytics team in the IDC conducted the analysis using SAP Predictive Analysis on top of HANA This was all completed within a six week period with a team of Client and Accenture with technical support from SAP resources, across Denmark, India and the United States The solution tested hypotheses of the data, including: What products, product groups and components contribute most to defects? What components are likely to fail together? What is forecasted defect frequency by product? What are the outliers by product, age, factory, etc? Delivered an SAP Predictive Analysis HANA Proof of Concept using client data within the Accenture Innovation Center for SAP lab environment Completed concept to delivery within a rapid deployment timeframe of 6 weeks Demonstrated SAP Predictive Analytics capabilities and benefits of tool, and put forth a vision for a future roadmap of SAP PA for the client Client Warranty Specialist to leverage capabilities demonstrated in the POC to more effectively forecast warranty costs and defect trends before, and to mitigate financial impact Accelerated future delivery by leveraging POC solution to rapidly launch Predictive Analysis implementation Copyright 2014 Accenture. All rights reserved. 10