What are the inventory forecasting methods?
Forecasting sales ensures that companies have sufficient cash flow to conduct their business. Forecasting sales also implies implementing inventory management strategies in order to meet market demand more efficiently. Indeed, the main challenge of inventory forecasting is to avoid situations of over-stocking, under-stocking and stock-outs, which are the causes of important prejudices suffered by establishments evolving in the retail or large-scale distribution. However, only a procedure conducted in compliance with the fundamentals of inventory forecasting guarantees satisfactory results. This is provided that a company chooses the most appropriate forecasting method in its context.
Average-based inventory forecasting methods
Two types of forecasts can be used here:
- the moving average method;
- the weighted moving average method.
The moving average method
This is the oldest method of estimating stocks. Calculations are based on the match between future and recent consumer demand. Future forecasts are therefore based on the average of past demand over a given period, which is considered to be the most relevant. The fundamental condition for the forecast to be realized is therefore to have enough historical data to exploit.
The advantage of the moving average method is its simplicity of application. It is especially suitable for companies that have a certain stability in customer demand for their products. The disadvantage, however, is the inability of an organization to extrapolate trends and thus deal with the uncertainties that may arise.
The weighted moving average method
The weighted moving average method follows the same principle as the moving average method. The only difference is that here, a greater weight is given to the most recent requests from consumers in the period considered. Under these conditions, the calculation of the average of past requests will be sensitive to the weight of recent requests.
This forecasting method has the advantage of modulating the weight given to the various requests and of being more precise with respect to the quantity of stocks to be ordered. However, like the moving average method, it does not forecast trends or seasonal variations.
Inventory forecasting methods based on smoothing
There are 4 methods based on smoothing:
- the single smoothing method;
- the double smoothing method;
- Holt's method;
- the Holt-Winters method.
The simple smoothing method
The principle of simple smoothing is easy to integrate: the forecast of future stocks is established by combining the last actual demand with the last forecast; the latter being itself the result of the previous demands and forecasts. A smoothing weight is assigned to the last demand to further refine the predictions.
With this method, the volume of historical data to be exploited is less than with the moving average or weighted moving average method; this is clearly an advantage. However, it is not suitable when it comes to taking into account a trend, a cycle or a seasonality.
The double smoothing method
Here, the company is asked to repeat the smoothing a second time, finally treating the first smoothing as if it were a new request. This is the most suitable method when the raw data show one or more trends. Specifically, two different weights are used to update the components in each period, in addition to the intermediate forecast.
The purpose of double smoothing is not only to smooth the level of the data, i.e. to eliminate random variations, but also to smooth the trend and thus eliminate the effect of the trend on the smoothed values. These values can be obtained either with optimal weights or with weights defined by the company.
Holt's method, also known as the "trend smoothing method", is based on the double smoothing technique. It is a sort of improved version of it. This sophistication in the technique is important not only for assessing the level of demand, but also for taking into account trend effects as in the case of double smoothing.
In principle, this method is accompanied by an even more detailed knowledge of the sector of activity in which the company instigating the inventory forecast operates. In calculating this forecast, managers must take into account the level of data and the slope. This is a weighted average between two constant estimates from the last observation or that have been chosen.
The Holt-Winters Method
The Holt-Winters method is still referred to as the "triple smoothing method":
- the first smoothing refers to the forecast of the demand level ;
- the second smoothing is related to the forecast of the slope;
- the third smoothing is for seasonality.
In fact, this forecasting technique builds on the calculations of the previous method by including seasonality effects in the time series. In these cases, the forecast is simply multiplied by the seasonality indices. If the periodicity is monthly, the index for each month is calculated using the demand for the months of the previous years.
Inventory forecasting methods based on quantity/interval separation
When demand is stable and erratic, inventory forecasting methods based on averaging and smoothing are very effective. However, they soon show their limits when demand becomes sporadic. In this case, the methods based on the quantity/interval separation should be used. There are two of them:
- Croston's method;
- the SBA method.
Croston's inventory forecasting model is for orders of a part or slow moving item to replenish inventory. Thus, it is designed for historical data with multiple zeros. This technique combines two types of forecasting through smoothing. These are:
- the forecast of the quantity requested by considering only the non-zero amounts in history;
- the forecast of the size of the interval between two non-zero requests.
These two smoothings are performed independently so that they can be used individually in the case of separate forecasts or jointly in the case of a single forecast. It should also be noted that the forecasts are non-trending and non-seasonal.
The SBA method
This method of forecasting inventories, which we call the "Synthetos and Boylan approximation", makes it possible to correct the bias created by the Croston method. Indeed, experience has shown that the independence between the history of non-zero demand and the history of the intervals between two non-zero demands is not as rigid as one might think. A correlation between these two measures should not be overlooked.
This is why this method is more than interesting. It proposes the same algorithm as Croston's method, with the difference that an SBA forecast replaces the last conditional phase by a formula in which "alpha" represents the correction coefficient determined by the evaluation of the bias.
In the end, the estimation of a company's product inventory and therefore of the quantities to be ordered is based on assumptions. Since the future is not certain when it comes to forecasting sales, it is all the more difficult to be precise in your calculations when you don't have the right benchmarks. This is why an effective forecasting method for inventory replenishment is useful and must be carried out without delay.
We recommend these other pages: