In-memory information processing with Hana

If current buzz words are to be believed, the corporate world seems to be increasingly characterized by "disruptive innovations": Existing products or services are more or less unexpectedly displaced by new technologies and their providers.

On closer inspection, however, companies that focus on continuous innovation prove comparatively more successful over a longer period of observation.

With the help of appropriate technologies, the rules of the game can be changed: German companies have long recognized the importance of speed and flexibility.

In times of "just in time" and "batch size 1", short-term changes represent a challenge for operational business process management. Managing unpredictable situations is increasingly becoming a key competence for customer satisfaction.

For the supporting IT systems, this means that flexibility is increasingly required. In the strategic area of operational corporate planning, one often falls back on empirical values from the company's internal past.

With Big Data, there is an almost infinite amount of external data. Linking these together and validating existing "gut feelings" through predictive analytics increases the quality of planning and the possibility of permanent learning.

Two concrete prototypical solutions for practice will demonstrate the innovative potentials for improving corporate decisions and planning that arise from Hana.

Hana scenarios

ERP on Hana: The focus is on supporting interventions during process execution. The objective is the possibility for rapid interaction with immediately available data.

For example, in many industries such as discrete manufacturing, an MRP run can currently only be performed overnight with a complex BOM structure.

Accordingly, an MRP controller can only unsatisfactorily answer the question of whether adjustments are required in the ordering process or in the production program due to changes in customer orders. In contrast, near-real-time, KPI-based reports would be required for monitoring and dynamic intervention.

BW on Hana: The central task of operational planning is to implement strategic targets. This is a system of sub-plans, typically with a focus on short-term measures.

For example, sales planning affects the existing production and service program. In the SAP area, it can be assumed that related activities will continue to be covered by BW systems in the application companies in the future.

An increase in quality can potentially be achieved through a comprehensive analysis of empirical and extensive data sets. Information external to the company is gaining considerable importance in this context - such as geodata and weather data.

In general, tools from descriptive and inductive statistics as well as from data mining can be used for predictive analytics. SAP offers the PAL ("Predictive Analysis Library") for this purpose.

The "Glass Ball Report" prototype developed by Pikon and Pforzheim UAS for Hana ERP looks at the integration of sales and scheduling.

Crystal Ball Report

Like looking into a crystal ball, it's about making a prediction. One looks at the sales order schedule lines due for delivery in the near future along with their contributions to sales.

In order to be able to assess whether deliveries can be made on time and thus invoiced, the coverage of requirements for the material in question must be analyzed on several levels, taking into account all possible MRP elements.

In the best case, there is already an inventory for the material to be delivered, otherwise it is examined for all relevant assemblies and raw materials whether production orders (or only planned orders) or purchase orders (with/without shipping notification or requirement) already exist.

With this overview, Sales can quickly identify where there are bottlenecks and thus deadline threats, and solve these with MRP. The existing SAP standard transaction "Multi-level order report" (MD4C) can only be used for individual sales order items and also does not show sales values.

The original glass ball report was developed for a "conventional" ERP system and ran so slowly that it was only usable as an overnight batch job for a complete analysis of all materials.

During Hana development, it became apparent that the unmodified report did not show any significant performance improvement. This would require the following steps:

To minimize the number of database accesses, the current "depth search" would have to be replaced by a "breadth search". "Code push-down by moving logic from the application server to the database server by creating appropriate views and stored procedures on the Hana database.

In the case of the Glaskugel report, this would have meant a complete redevelopment of the program, a fate that many complex in-house developments are likely to share.

A particular difficulty here is the use of SAP function modules, which themselves are not yet Hana-optimized. Here, we have to wait for Simple Logistics under S/4.

There is potential for improvement at the interface between sales and scheduling. In order to be able to leverage them through Hana, considerable effort will usually be required for the optimization or new development of customer-specific reports.

The second example includes a sales planning and forecasting use case where Hana actively supports a user by providing suggestions for plan creation.

BW on Hana and PAL

Here, the sale of bicycles is to be planned on a detailed level (material/customer). The system provides a proposal that is not only based on historical sales figures, but also takes into account external data (e.g. historical weather data, quality of the bicycle path network).

The advantage of the PAL used is that it can automatically determine statistical dependencies. Within an integrated system landscape, historical sales figures from an SAP ERP system as well as external weather data and geoinformation from a database are loaded into a BW on Hana via an ETL process.

Together, they form the data basis for the planning process. Users can start the forecasting process via planning functions in the BusinessObjects Analysis for Office Workbooks.

These planning functions are implemented as customer-specific planning function types according to the code push-down principle and are thus executed at database level.

Within the planning functions, PAL methods can now be called up with Hana and their results integrated into the planning process in real time.

Two PAL methods are used in the prototype to generate the forecast: firstly, a time series algorithm (exponential smoothing) based on historical data from the last seven fiscal years; secondly, the solution includes a sales forecast functionality based on weather scenarios for upcoming months using a regression analysis.

Also due to the consistent Hana optimization of standard functionalities by SAP, data is provided at the required level of detail in real time and can be used directly for further analyses and planning.

These can lead to changes in sales planning (closed loop). The sales planner can, for example, access monthly or annual values at customer, material group and item level.

BusinessObjects Analysis for Office acts as a front-end tool with different workbooks for forecasting, quantity planning, and revenue planning.

In revenue planning, current price information is read at runtime via Smart Data Access from ERP on Hana without storing it within BW.

During the initial prototyping, it became apparent that in-house developed Abap programs do not automatically achieve a performance gain during an ERP conversion to Hana.

This requires adjustments (e.g., broad search instead of deep search) or a systematic approach to Hana implementation. In this case, nothing stands in the way of a "Non-Disruptive Success" or a "Perpetual Innovation".