Our approach for Predictive Controlling / Accounting is based on the standard procedure in the field of data analysis: A model is created on so-called training data (e. g. sales 2011–17) and then checked with the test data (e. g. sales 2018). The algorithm with the optimal model on the test dataset is then selected and validated with an independent dataset (e. g. sales Jan.–Jun. 2019). State-of-the-art algorithms are used to create the model. The KPI for the model evaluation is always selected in individual consultation with the respective customer.
Predictive models by s-peers achieve an accuracy of 95–99 %.
Scenario A: The results are calculated directly in the database (HANA) according to the model and are transferred into the SAP Analytics Cloud (SAC) for further planning. The calculated values can be adjusted manually and/or be simulated graphically in a Value Driver Tree (VDT).
Scenario B: The calculated results are imported into the consolidation solution (in this case SAP SEM-BCS) for the consolidation of rolling extrapolation and planning.
Scenario C: The forecast results are calculated directly in the SAP Analytics Cloud (SAC) and can then be used for other purposes (e. g. in SAP BW or HANA).
Predictive models by s-peers are universal: They can be used independently of existing analytical systems and/or system landscapes.
The automation of the respective process generates enormous efficiency potentials and supports the objectification of the forecast: Planning distortions due to individual “gut instincts” or political behavior are reduced maximally by using this methodology.
However, IT-supported and individually created forecasts can be a very good combination: Human experience and impulses (possibly not yet available in digital form) can be combined advantageously with calculated, highly precise forecast models. All in all, this results in a maximally realistic evaluation of the situation – and is therefore a sound basis for decisions. An example: One of our customers demonstrates the potential of this hybrid concept by combining the manual forecasts of his individual sales representatives successfully with IT-based versions in the “Demand Forecasting” area.
According to our experience, automated processes should always be implemented very sensitively in the organization. Because despite all technological possibilities and statistical procedures: The “human factor” is still highly relevant – the individual person validates the models continuously to detect new cause-effect relationships and also acts as the corrective for all irregular and/or disruptive developments.
Therefore it is all the more important to involve employees at an early stage and to generate trust sustainably: Automation should never appear as a kind of “magic”. For example, the actual process can be designed in such a way that when a corridor of permitted future values (defined in advance) is left, countermeasures are initiated which, in turn, lead to interactions by staff members.
An automated solution for forecasting and planning undoubtedly offers a great potential but also requires a certain amount of caution: An approach of this kind always depends on the quality of the respective database and also reaches its limits quickly in the case of structural breaks.
Predictive projects are now desired with increasing frequency. Simultaneously, however, there are often major deficits with regard to the clear definition and articulation of the associated issues. Many times, this results in a quite extensive but hardly goal-oriented occupation with currently “hot” topics like e. g. IoT, sensor technology or machine learning. Such technologies are undoubtedly promising for the future. However, most projects of this kind finally come to nothing because the basic prerequisites in terms of data (basis, quality, usability) and interfaces are lacking largely or completely.
This brings us back to the confidence in such new measures and approaches for Corporate Management. Thus you should initially drop any further or additional unknowns: Ideally, you start in an area which already has a very good database.
This is one of the reasons for our clear recommendation to establish your first added value in the Accounting area: Here, tried and tested standard systems already exist – increasingly in a single-circuit system. Central data storage in an integrated business suite (e. g. SAP S/4 HANA) offers a standardized basis. In addition, time-consuming reconciliations between the Finance and Controlling areas are no longer necessary because fragmentation is no longer required. For all analyses, an integrated and highly performant database is available in real time.
- Less effort for non-valuable activities
- More precise results and enhanced transparency thanks to increased information density
- Clarification of interrelationships and cross-functional dependencies (across a wide range of business areas) through disclosure of cause-effect chains
- Massive increase of response capacity due to automation – calendar orientation increasingly becomes obsolete
- Objective forecasts which give a fact-based view
- Multiple use and reusability of algorithms provided by s-peers
- Wide variety of evaluation options in assured quality
- Early possibility of performance evaluation via simulation