Is it possible to automate the sales forecasting?

19 Июня 2014

Introduction.

The success of the company is ensured by the loyaltyof its customers. Reducing the number of cases of goods trading company ensures the loyalty. Effective forecasting, you prepare the source data for the Procurement Division.

They planning to timely delivery and ensure the continued availability of necessary quantity of goods in warehouses. The challenge of forecasting it is tedious androutine work. It requires the attention of a large number of experts, product groups, which are perfectly aware of the demand for goods, especially its inputand output range. If a large cadre of experts for the company is unprofitable, and experts are unable to cope with the volume of tasks, which naturally leads to errors and result in significant losses to the company. Then come to the help of  IT solutions.

Task.

Develop a solution that provides forecasting tools for product groups. The characteristics of the implementation:

  • processing of large volumes of information about sales and availability of goods in stock;

  • mathematical sales forecast - forecasting as an Assistant;

  • storing the history of sales forecasts;

  • visualization of the adequacy of predictions of the past periods.

Solution.

The analysis of existing methods of forecasting sales of goods identified the following shortcomings:

  • the methods mostcommonly implemented using Excel spreadsheets that cannot handle large amounts of source data (a large assortment of goods or a significant amount of sales history);

  • existing methods using classical statistical extrapolation, which, in most cases, is not suitable to describe sales forecasts for future periods.

For an expert in forecasting commodity group to get information:

  • about product a vailability in the investigated period;

  • the nature of and demand for commodity procurement;

  • cases of special orders;

  • estimates of mathematical trend and forecasts;

  • comparative characteristics of their projections for the current and prior periods, fortimely response to changes in demand for the item.

To store has proposed data on projections in documents (Figure 1). This will keep a history of past forecasts and refer to them as needed. Details of the document were: period sales analysis, predictions, sales script (seasonality), average forecast and planned underlying sales growth.

Figure 1– «Sales forecast».

Period sales analysis is sufficient to calculate such indicators as: linear trend coefficients (equation of the form x = bt + a), the average number of days of the sale of goods, the availability of goods in stock. The script is required to modify the demand in different months, the forecast period (for example, seasonality, which may be different for different groups of products), as well as to highlight the true demand of the commodity sales data. The document shows information on the deferred demand forecast at the beginning of the period (the «Deferred demand» column in Figure 1) and planned deliveries of goods (column «Container» in Figure 1).

The order of work with documents the next:

  • filling of products with characteristics in common script sales (seasonal);

  • determination of the forecasting period;

  • processing of data on the availability and sales of goods, determination of the true demand for the goods (column «The average sale» in Figure 1), calculation of the coefficients of a linear trend (column «b=» in Figure 1), population forecast of last period (column «Forecast of the average» in Figure1);

  • next expert corrects average forecast (column «Forecast new» in Figure 1) and algorithm extrapolates this new prediction for the sel ected number of months prediction, taking into account the calculated parameters of a linear trend with a sales script products;

  • after which a visual assessment of the adequacy of the automatic way the data on projections (Figure 2);

Figure 2 - assessment of the adequacy of data on projections.

Presentedon Figure 2 a chart to the expert to analyze last period, and make sure thatthe algorithm would have enough information to calculate for measures in automatic mode. Also on the chart «sales / prediction» is held a comparative evaluation of sales forecasts and actual sales. This allows you to respond to rapidly changing demand for a product. The third chart on Figure 2 shows deviations expert’s opinions, if any have (chart of blue color) from fully automatically collected data about the forecast (chart of yellow color).

Conclusions.

The Submitted solution allows for forecasting a large range of goods with a high volume of sales history. Unlike existing method, extrapolation is not the classic statistical method, but ascript on the sales trend, what increases the accuracy of forecasts and it is proved in the following article. The expert, without having to resort to complicated manipulations with data, can change the forecast, resulting automatic way, and algorithm, if necessary, extrapolates forecasts again, but taking into account the opinion of the expert. Advanced Visual part solutions allow you to reduce the number of errors, as automatic algorithm, so expert too.

Thank you for your interest. Dmitry Kalchenko. I await your comments (registration is required).

Translation fr om Russian Tatyana Kalchenko.

Короткая ссылка на новость: http://task-idea-solution.org/~okBTM