Improving Planning Accuracy in the Company with Predictive Planning
An efficient planning process is essential in businesses to allocate and utilize resources effectively. Often, this highly complex process, which requires a multitude of data, information, and stakeholders, is still carried out using Excel, making it very time-consuming and prone to errors. Since manual planning often relies on the intuition and experience of individual employees, it’s challenging to systematically improve planning accuracy.
Feature-based Forecasting (FBF) is a predictive data analysis method aimed at improving planning accuracy and reducing the time required by developing a model based on an algorithm.
What is Predictive Planning?
In general, different planning procedures can be applied:
- Classical Approach
- Driver-Based Approach
- Data-Driven Approach (Predictive Planning)
In the classical approach, the initial situation is a classical planning set-up in which manual planning is based, for example, on an initialization of historical values. This procedure is still carried out in most companies, but it is time-consuming, error-prone and has a low planning accuracy. The driver-based approach includes the consideration of internal and external factors influencing planning. Expert knowledge is divided into different dimensions, such as competition, innovation, human resources, culture and strategy. In this approach, value driver trees make it possible to plan the essential factors and to map interdependencies within drivers. However, as these driver models are based on human experience, crucial factors can also be overlooked or misattributed here. This is where the data-based approach comes into action. This can then be used to transfer the identified factors into a machine learning algorithm to automatically generate data-protected predictions for the use case. Here, predictive analytics is applied to planning, therefore the name Predictive Planning. Feature-based forecasting is a method of predictive planning or predictive analytics.
But how does feature-based forecasting work?
To be able to use data-driven alerts sensibly, a high-performance data integration and real-time data transmission is necessary. Live data from one or more source systems is uploaded in real time to target systems such as Qlik Sense. The customer can then view this live data directly in a customized dashboard. Through the additional setup of alerts, a push notification, for example in Teams, is sent to the responsible person when key performance indicators (KPIs) reach or fall below predefined thresholds. This person can then initiate further measures if necessary.
This method is suitable for many different systems and scenarios, from production to sales. Feature-based forecasting uses mathematical models and algorithms to achieve reliable results, making it a category 3, data-driven approach. This approach is a distinct process that needs to be customized to the specific use case. Because Feature-based Forecasting requires not only a deep understanding of data but also a solid foundation of knowledge about the specific business model. In general, the FBF process can be divided into the following 5 steps:
- Business Understanding
- Data Understanding
- Data Preparation
- Modelling
- Evaluation & Deployment
A specific use case takes us into the world of sports betting: the experts at H&Z.digital were tasked with a project by a sports betting provider to increase the planning accuracy of revenue forecasts for football bets in Germany. The technical foundation for this project was the planning and reporting tool “Board”. As a technical solution for this project, an artificial neural network was built, which can take in, evaluate, and transmit information. Thanks to Feature-based Forecasting, the consultants were able to significantly increase the planning accuracy for the sports betting provider, making revenues and resources much more predictable and reducing risks significantly.
If you want to hear more details about the exact process, the advantages and disadvantages, as well as the use case described above, you can read our full article in the DF&C magazine or watch our on-demand webinar on the topic.
If you still have questions or suggestions about Feature-based Forcasting or H&Z.digital in general, please contact us by email at any time: info@hz.digital.
