Introduction of application function library big data, big challenges, and big opportunities
4 / 33 © 2016 sap se or an sap affiliate company. all rights reserved. figure 1: moving decision making from sense and respond to predict and act edw and etl bi agile bi predictive analytics raw data cleaned data standard reports ad hoc reports and olap self- service bi agile visualization generic predictive analysis predictive modeling optimization what happened? why did it happen? what will happen? what is the best that could happen? userengagement maturity of analytics capabilities predictiveinsight legend edw = enterprise data warehouse etl = extract, transact, load bi = business intelligence olap = online analytical processing
5 / 33 © 2016 sap se or an sap affiliate company. all rights reserved. the result of predictive analytics is not a bi arti- fact like a query or a report, but a mathematical representation (a "model") of the relationships found between different elements in the data. you can then use this model to perform what-if analysis to maximize or minimize a value. or you can apply the model against entirely new data to predict a future outcome such as "a customer’s probability to purchase" or "projected revenue for the next four quarters." making predictions in real time is even more valu- able: rather than waiting until the end of the day or end of the week to run an analysis, you can make a decision as soon as data becomes avail- able. this can mean millions of dollars or, in some industries, could be the difference between life and death. however, most organizations have many hetero- geneous technologies supporting legacy environ- ments, leading to inefficient data management and significant operational overhead. in addition, many organizational silos are common in larger organizations where multiple teams are responsi- ble for the data infrastructure. this is illustrated in figure 2. there is an ever-increasing set of technologies designed to make each of these steps faster. however, both the amount of data being gener- ated and the speed at which this data needs to be analyzed is growing exponentially so that traditional methods of analysis will soon no longer be viable. figure 2: data analysis – multiple silos versus in-memory computing separated transactions + analysis + acceleration processes • redundant data in and across applications • inherent data latency transact analyze cache accelerate etl etl versus one in-memory atomic copy of data for transactions + analysis • elimination of unnecessary complexity and latency • acceleration through simplification sap hana® legend etl = extract, transact, load
sap hana as a predictive analytics platform 12 / 33 © 2016 sap se or an sap affiliate company. all rights reserved. today, a single data platform can support the different predictive needs of business analysts, data scientists, and even application developers in an organization. the predictive engines of sap hana are implemented at the application function library (afl) layer, which enables almost any other process to use or embed these predictive capabilities with minimal effort