How to Select the Best Demand Forecasting Model
Aaron Ethridge
Last Update 3 bulan yang lalu
Selecting the appropriate forecasting model is crucial for accurate inventory planning in Planster. Here is how to choose the best forecasting method based on your sales channels, available data, and business needs.
Consider the following guidelines to determine the optimal forecasting model for your business.
For Retail Channels:
Bottom-Up Forecasting: Use this model if you are forecasting for retail stores and have access to store-level data. This method leverages a detailed calculation of store count x shelf facings x velocity. It is ideal for established retail channels where granular information is readily available.
Top-Down Forecasting: Use this model when overall sales targets or a total market size drive inventory requirements. It is especially helpful for new products or promotions where detailed historical data is limited.
For Direct-to-Consumer (DTC) Channels:
Prophet: Select this model when clear seasonal patterns or historical promotional impacts exist.
Linear or Exponential Smoothing: Use these models for products with steady growth or consistent historical performance.
ETS is preferable when occasional irregularities or short-term fluctuations appear in your historical data.
SARIMA: Choose this model when you have robust historical sales data (at least two or more years) to accurately capture complex, long-term patterns.
Practical Recommendations
For new products without historical data, start with a Linear or Top-Down, and then adjust based on early sales performance. You can utilize benchmarking (leveraging sales data from similar existing products) or upload your own data. On this, see the article, Forecasting Demand for New Products.
For products with clear seasonality, Prophet typically provides the most accurate forecasts.
When historical data is extensive and stable, SARIMA and ETS models are highly reliable.
Reviewing Model Performance
Regularly evaluate your chosen model’s accuracy to ensure continued reliability:
Compare forecasted versus actual sales periodically.
Adjust your selected model based on observed forecasting accuracy and current business conditions.
Need More Help?
For personalized advice on selecting and optimizing your forecasting models, reach out to Planster Support - we are here to help!