THE SALES FORECAST PROCESS IN COMPANIES: ORGANIZATIONAL AND TECHNOLOGICAL ASPECTS

The development of increasingly sophisticated forecasting techniques, in parallel with the rapid development of computers and other information and data manipulation technologies, have led several companies to become increasingly interested in the sales forecasting process. This growing interest is mainly based on the dissemination and use of personal microcomputers that are increasingly powerful and endowed with various resources. Currently, it is possible for every manager to implement sales forecast models in electronic spreadsheets (eg EXCEL*, LOTUS*, etc.) as a subsidy to their planning and control activities, whether in the strategic, tactical or operational field. It is clear that the perfect understanding of the various quantitative forecasting techniques allows managers to effectively use forecasted values ​​(or cold numbers, a term frequently used in several Brazilian companies) as a starting point for incorporating their judgment and sensitivity regarding various issues. such as, for example, competitors' actions, promotions, etc., and for discussion with other company departments on issues such as capacity planning and scheduling of machine downtimes for maintenance, definition of service levels, product availability, etc.

It is noticed that the role of purely and simply intuitive forecasting, practically the only tool available to managers before the spread of microcomputers, is diminishing. The human mind, despite having unique characteristics regarding complexity and power for storing and associating information, is subject to biases and emotions, being generally optimistic and underestimating future uncertainty, especially with regard to sales forecasting. Currently, the most efficient and accurate managers in sales forecasting are those capable of composing an adequate mix between the result provided by quantitative techniques, their market sensitivity and the restrictions imposed by the various departments of the company.

Virtually all companies, whether small, medium or large; state, national, private or multinational, need to plan their resources for production, distribution and purchase of inputs or services vis a vis uncertain future conditions. In addition, the need to forecast sales is not only common to almost all types of companies, but also to the various functional departments, which need sales forecasts as a fundamental element of their decision-making process.

In this sense, as the planning needs of the different departments of the company are different, the management of the sales forecasting process involves, above all, the management of the people who make the sales forecasts in the companies. This people management generally encompasses aspects related to the organization, procedures, motivations, recognition and reward of the personnel involved in the preparation of forecasts and their integration with the other departments of the company.

The organizational aspect refers to the specific roles and responsibilities of the sales forecaster. Some questions that must be answered by the company:

• Who is responsible for sales forecasting?
• How is the accuracy (error) of forecasts measured and how is performance evaluated?
• How does sales forecasting integrate with salespeople's recognition and reward mechanisms? Is there any relationship between recognition and reward mechanisms and accuracy in forecasting? The procedural aspect refers to understanding how sales forecasting techniques and their decision support systems influence the rest of the company. Some questions present are:
• How do marketing and sales departments perceive the impact of sales forecasting on production and logistics activities? And vice versa?
• Are those responsible for forecasts aware of the different existing techniques and do they know which one is most appropriate for the company?
• Do forecasters know the full potential of the decision support system purchased or developed internally by the company?

It is extremely important that these questions be answered when designing the forecast management system. If these issues are ignored, the forecasting process cannot be controlled much less improved. For example, if marketing, sales, production, and logistics departments develop independent sales forecasting methodologies, there will be no integration into the decision-making process, much less anyone who is accountable for the accuracy of the method.

Another extremely important element for managing the sales forecasting process is the diffusion of new information technologies. For example, in large retailers in the US and Europe, the automation of POS, associated with the use of bar codes on products, has allowed the adoption of more sophisticated and efficient sales forecasting systems that seek to take advantage of sales information collected in real time. .

For example, a large European retail chain with French capital developed a daily sales forecast model for its products. As it was intended to make daily forecasts for thousands of products, the developed model is not very complex in order to allow obtaining results in an acceptable time. This model is applicable to all stores in the retail chain, with only the value of the parameters varying from store to store. In forecasting sales of a product for a given day, the model takes into account the day of the week, the day of the month, the proximity of a holiday and festive dates. Among the main conclusions obtained, we highlight:

• Different days of the week imply different sales values: on Saturday the volume of sales is greater than on other days of the week.
• sales volume is higher at the end and beginning of the month. After the 25th of each month, the day from which most people start receiving their salary, sales are higher than the average, remaining significantly high until the 8th of the following month.
• Any day of the week usually has higher or lower sales if it is close to or coincides with a holiday.
• Products from the same family (eg hygiene and cleaning) have very close seasonal factors.
• Stores located in different regions: periphery, interior, coast, etc., present completely different sales patterns.

The third and last element relevant to the management of the sales forecasting process is the choice of the appropriate sales forecasting technique(s) for the company's reality. In general, a forecasting technique consists of the mathematical or statistical calculation used to convert historical data and parameters into future quantities. Forecasting techniques generally fall into two main types:

• Qualitative techniques. These techniques depend exclusively on the expertise of the forecaster(s), being generally more expensive and laborious than quantitative forecasting methods. They are ideal for situations where there are no historical series available and/or human judgment is required, being developed through opinion polls, panels and meetings of specialists.
• Quantitative techniques. These techniques are divided into two main subgroups: time series and causal models. Time series techniques use historical sales data as a basis for determining patterns that may repeat themselves in the future. Examples of time series techniques are Moving Averages, Exponential Smoothing, and Classical Decomposition. Causal models seek to relate sales (dependent variable) with other factors such as GDP, inflation, time, population, etc. (independent variables). Examples of causal models would be linear regression and non-linear regression techniques.

In this way, the execution of effective sales forecasts requires a procedure that integrates three main components: forecasting techniques, new information technologies (decision support systems) and people management, as shown in the following figure.

A successful company that managed to articulate people management with the adoption of new information technologies and the choice of appropriate sales forecasting techniques was North American Pepsi-Cola. This company acquired a sales forecasting system, EssBase, which integrates and consolidates the daily revenue of commercial offices in 35 countries, where 25 different brands are handled at the level of 15 SKUs, totaling US$15,5 billion/year. This system, designed to operate in a client-server environment, allows its users, individually or in groups, to analyze the sales data of all its distributors from anywhere in the world, from different levels of aggregation, in the following dimensions: time, geographic coverage, types of brands and packaging, customer groups, etc. One of the main achievements of this system was the unification of techniques and decision support systems used by the various commercial offices in sales forecasting, unifying the company's expectations regarding future sales.

Finally, we realized that the sales forecast is an important input for the planning not only of companies from different sectors of the economy, but also of practically all their departments. The question that must be put on the agenda is not “should companies forecast sales?”, but rather “how should companies forecast sales at the lowest possible cost?”.

The following figure illustrates two important points to emphasize regarding the two main components of the total cost of sales forecasting:

TOTAL COST OF FORECAST SALES = COST OF PERFORMING THE PROCEDURE + COST OF FORECAST ERRORS

We realize that the cost of using the manager's sensitivity as a sales forecast is often very low, however, the cost incurred with forecasting errors (eg, buying excess capacity) more than outweighs these savings. On the other hand, the use of very sophisticated models, whose understanding is restricted to specialists, is not advisable: their operating costs are high, not being compensated even if the precision of the forecasts is acceptable.

Companies, therefore, should adopt forecasting procedures according to their forecasting needs with regard to:

• the forecast horizon (short, medium or long term),
• type of product (class A, B or C; new or existing),
• type of decision to be made (the department that will use the forecast).

Once the real forecasting needs are understood, the company must choose the method that is closest to the ideal operating region, as shown in the figure below, that is, the one that presents the best cost/accuracy trade-off.