In forecasting, hierarchical granularity levels refer to the different layers of data aggregation that can be performed. Often there is the need to forecast not only on one layer but also on these different aggregated levels. Common hierarchical granularity levels in such applications can be based on item characteristics or geographical location.
Let’s take an example application of forecasting demand of different products in a supermarket chain. The supermarket can have products in different categories such as vegetables, fruits, drinks, computers, toys, etc. The categories can be further grouped into food, hobbies, work, etc. This is the case where hierarchical granularity levels are based on product characteristics. On the other hand, the sales need to be forecasted for different stores, which can be in different cities and states. These hierarchical granularity levels are based on the geographical location. If you are a demand planner for this supermarket chain, you need to forecast for these different levels. For example, you need to forecast not only the number of laptops sold for a certain model, but also the total number of laptops sold overall. Also, it is common to not only forecast the sales for each individual product, but also the total sales of the product in each city and/or each state.
Forecasting on different hierarchical granularity levels helps in various aspects of businesses such as budgeting and financial planning, resource allocation, supply chain management, strategic decision making, performance monitoring, identifying customer behaviour, risk management and optimisation of operations.
AI & Analytics Engine facilitates hierarchical granularity level forecasting in their demand forecasting templates. The user can define their own hierarchical levels including the grand total level and the Engine automatically generates accurate forecast for all of these levels.
🎓Instances of a lower hierarchy level cannot belong to multiple higher levels. For e.g.: product laptop cannot be in both entertainment and work categories. If you need to categorise them to two categories then use two products such as entertainment laptop and work laptop.
Defining hierarchical granularity levels in AI & Analytics Engine. On the right side panel shows all cases in which forecasts will be made.
Forecasts for two different hierarchical levels on the Engine. Top figure shows the forecasts for family = BEVERAGES and city = Quito and bottom figure shows forecasts for all products in all locations.