Using AI to Optimize Demand Planning with Microsoft Dynamics 365 Supply Chain Management

Microsoft Dynamics 365 Supply Chain Management (D365) Demand Planning is utilizing artificial intelligence to help users make educated decisions in demand planning. This new feature seamlessly integrates with D365 and other data sources to implement machine-learning forecast models, detect outlier data, and utilize best-fit algorithms. At the same time, providing users with automated advanced analytics and intelligent feedback.

Demand planning is configured with state-of-the-art forecasting models. The demand planning app provides several AI-powered forecasting models, including:

1. ARIMA – A powerful and flexible model that can capture complex patterns in time series data. It is widely used for forecasting in various domains like finance, economics, and operations management.

2. ETS – ETS is a widely used forecasting technique, especially for short-term forecasting, and is often compared to ARIMA models in terms of performance. It is known for its flexibility, ease of use, and ability to handle a variety of time series patterns.

3. PROPHET – Prophet is a forecasting library developed by Facebook’s Core Data Science team. This model is particularly useful for forecasting time series with complex patterns, such as those found in e-commerce, finance, and social media data. Its flexibility, ease of use, and ability to handle a variety of time series characteristics make it a popular choice for many forecasting applications.

These models use machine learning to analyze historical data and generate forecasts. The variety of these models allows businesses to find a model that best fits their use case.

The new application also provides users with outlier detection models. Outlier detection algorithms improve forecast accuracy, identify anomalies, and reduce noise in time series data to provide consistent and reliable forecasts. Demand Planning uses statistical techniques like Seasonal Trend Decomposition (STL) and Interquartile Range (IQR) to automatically detect and remove outliers from the input data before running the forecasting models.

If businesses are unsure which forecast model is best for them, D365 Demand Planning provides a best-fit algorithm. The algorithm runs all the available forecasting models using actual data to find the model with the most accurate output. This ensures that every business use case uses the most accurate forecast model.

Demand Planning forecast outputs can be quickly analyzed with CoPilot AI. CoPilot can provide contextual insights and recommendations directly within the demand planning application’s user interface, such as shift detections, trend analyses, anomaly identifications, and forecast accuracy assessments. The application will display key metrics and analysis in a dedicated “insights” section, allowing planners to quickly understand the forecast’s state without generating reports manually.

The D365 Demand Planning application’s capabilities continue to grow. New enhancements to the product include further leveraging AI and machine learning, such as incorporating external signals like weather or economic data into the forecasting models and developing a “planning agent” that can automatically detect issues in the forecast and suggest solutions.

Written by: Owen Mitchell

Next Steps: