Welcome back to our well-known use case series. It is designed to offer a closer look at business value acceleration – driven by the design principles of SAP S/4HANA. The technology-induced implications of the digital economy are huge, though the technology itself is only a catalyst to achieve a fundamentally different business outcome.
The selected use cases, related to SAP S/4HANA Release 1709 are designed to emphasize and visualize the value behind this technological shift, that can be activated by you.
In this blog, we want to focus on one of our major innovations for our 1709 release, which includes predictive analytics. In the last blog, we considered the business value driven by the contract consumption capabilities. Today, we will consider this functionality from a technical perspective.
From “system of records” to “system of intelligence”
The purpose of SAP S/4HANA is to move the core from a “system of records” to a “system of intelligence”. Embedded analytic capabilities are the foundation to do so. 1709 goes beyond, as we add predictive analytic capabilities. This allows to leverage machine learning models which in return will help you drive better decisions.
The 1709 use case we look at is forecasting contract consumption. Based on the underlying predictive analytics algorithm in SAP S/4HANA, we can estimate a purchase contract with a supplier against a full contract consumption. To calculate the point of time of full consumption, the system interpolates historical data stored in the system.
The first step to implement the predictive functionality is to train the predictive model by the analytics specialist within your company. After the predictive model has been trained, the pre-build job, that calculate the contract consumption figures, runs automatically with the launch or refresh of the FIORI app. With each calculation, the predictive consumption data will be displayed in the trend view of the “Quantity Contract Consumption” app.
Train your predictive model
To prevent complex and cumbersome setting processes, SAP S/4HANA 1709 is shipped with a preconfigured predictive scenario using SAP BusinessObjects Predictive Analytics Integrator. This allows to integrate and deploy predictive models in business applications and enable customers for purchase contract prediction. However, to be able to adopt predictive models to individual customers, the model is first going to be trained on specifics and historical data.
The following steps are required on customer side. Let me introduce the app creation lifecycle for standard embedded analytics apps. All blue steps are required for the standard embedded analytics app, those in orange are additional steps to train the prediction engine.
For the predictive apps, there is one additional step on SAP side, the predictive scenario, which contains the predictive model. Based on required data, you can define the input of the predictive model. The same applies for the output of the model, which can be predefined as well.
The predictive model is trained on the customer´s specific data and by the customer, which allows more accurate forecasts. It considers special behavioral patterns of the customer`s data.
Automated Predictive Library (APL)
To make the predictive model most accurate, the Automated Predictive Library (APL) in the SAP HANA database plays a fundamental role. The SAP HANA APL is an Application Function Library (AFL) that offers the possibility to leverage data mining capabilities of the SAP Predictive Analytics automated analytics engine on customer´s datasets, stored in the SAP HANA database.
Using the SAP HANA APL allows to create different types of models to answer business questions, such as classification/regression models, clustering models, time series analysis models and recommendation models. As the APL optimizes machine learning models systematically, there is no need to perform this tasks manually. With the application, we ship the predictive model container which is then trained on actual customer data. From a security perspective, customer data always stays in the own SAP S/4HANA system, no need for a replication to other systems for training purposes.
The step to create and train the predictive model is carried out by an analytics specialist. Most common, power users for analytical applications from a business department. For them, we ship a standard SAP Fiori role “SAP_BR_ANALYTICS_SPECIALIST”. In addition, the Business Catalog SAP_BW_BC_UMM_PC.
Now you might think that the task of the analytics specialist is complex, but this is not.
To train the predictive analytics model, it starts with a simple one-click access. The evaluation of the predictive analytics model is also fairly simple. The predictive analytics model is rated using stars in a scale from 0 to 5. models that receive a 5-Star ratings are considered a good fit. Based on this rating, it can be decided whether the predictive model shall be activated. In case of activation, the predictive analytics app will produce new data. This data can again be leveraged to re-train the predictive analytics model over time. Re-iterate the training after some time with new data and insights make the predictive analytics model more powerful with each cycle.
Taking into consideration the value levers, we mentioned in the business view of this use case, this use case is driving the value lever “Speed up signal to action” and “Raise process intelligence” in the category of increase effectiveness. This is based on machine learning algorithm, the full contract consumption of a purchase contract is predicted, before it happens.
Watch out for our upcoming use case blogs in the weeks ahead and check out the collection of these use cases, where you can find all links in one place!
For more information on SAP S/4HANA, check out the following links:
SAP S/4HANA release info: www.sap.com/s4hana