Sales Seasonality and Its Impact on Forecasting
Sales seasonality is one of the key characteristics of products in forecasting. Incorrect accounting for seasonality leads to a decrease in forecasting accuracy, resulting in several problems:
- Decrease in the manufacturer’s service level at the beginning of the high sales season.
- Increase in inventory at the end of the season.
- Decrease in the efficiency of sales chains due to inaccurate resource planning.
- Inability to plan the supply of necessary raw materials for product manufacturing in advance.
- Overloaded logistics capacities during peak seasonal loads.
Types of Seasonality
Temporal Seasonality There are many products that experience predictable cyclical changes in sales throughout the year, week, or even time of day. This seasonality is clearly defined by time and can be forecasted based on historical data.
Annual Seasonality Among products with annual seasonality, we can highlight: grapes, citrus fruits, ice cream, and bottled water. The curve of annual seasonality is smooth, unlike products with holiday seasonality.
Annual Seasonality Among products with annual seasonality, we can highlight: grapes, citrus fruits, ice cream, and bottled water. The curve of annual seasonality is smooth, unlike products with holiday seasonality.
Figure 1. Example of annual seasonality for ice cream
Weekly Seasonality For example, alcohol has pronounced weekly seasonality. This group of products sees a sharp increase in demand on Fridays.
Figure 2. Example of forecasting beer sales with weekly seasonality and pre-New Year spike
Daily Seasonality Almost all products have this characteristic. Daily seasonality is related to the overall number of customers visiting the store. It is known that peak traffic in stores occurs in the evening when people return from work, so it is natural that all products have sales peaks at this time. However, for calculating daily sales, daily seasonality is excessive.
Holiday Seasonality There are also many products where increased demand is determined by holidays that occur annually. For these products, it is necessary to separately collect sales statistics and build forecasts. Holiday seasonality is characterized by short spikes, unlike annual seasonality. These spikes are tracked by marking holidays such as Christmas and others. To forecast demand for products considering annual and holiday seasonality, a sales history of at least two years is required. This history allows capturing recurring sales events and forecasting them based on previous values. Algorithms, using the marking of seasons and holidays, collect sales statistics from historical data and extrapolate it considering the sales trend for the forecasted dates.
Holiday Seasonality There are also many products where increased demand is determined by holidays that occur annually. For these products, it is necessary to separately collect sales statistics and build forecasts. Holiday seasonality is characterized by short spikes, unlike annual seasonality. These spikes are tracked by marking holidays such as Christmas and others. To forecast demand for products considering annual and holiday seasonality, a sales history of at least two years is required. This history allows capturing recurring sales events and forecasting them based on previous values. Algorithms, using the marking of seasons and holidays, collect sales statistics from historical data and extrapolate it considering the sales trend for the forecasted dates.
Conclusion
Sometimes it seems that making a forecast considering weather and holidays is not so difficult without special services. And for one store and 10 items, it is indeed simple. But if we are talking about 100 stores and 15,000 items with different seasonality and demand characteristics, making a forecast manually is very labor-intensive, and its accuracy is likely to be insufficient. If we talk about a manufacturer supplying 500 counterparties, each of which has its own promotions for our products, then a forecasting service is simply necessary.