Demand Forecasting

Demand forecasting is a complete business process that makes a preliminary assessment and planning of the demand for the company's goods, materials and services, which helps to maintain maximum profitability. Without quality demand forecasting that takes into account customers' needs, preferences and intentions, seasonal factors, weather events and important social events, companies risk either missing sales opportunities and missing profits or, if demand is overestimated, being left with significant amounts of surplus product. In a number of categories, especially in the food segment, incorrect demand forecasting leads to direct losses.

In this paper, we address the issue of demand forecasting from statistical data processing and mathematical model training.

Why Demand Forecasting is Important for Business

Today's market is dynamic and businesses have to adjust. Planning for the calendar year after a number of socio-economic stresses has already moved into the category of strategic tasks and is no longer so effective in terms of capital turnover. Companies have to do more immediate analytics (3, 6, 9 months) and plan in less time with less labor and for shorter time frames, including the next day or two. This requires accurate statistical calculations and their competent analysis. The base of demand forecasting is analytics of market dynamics, its causes and effects, trends of its change. Companies focus on demand forecasting to solve subsequent tasks in the supply chain: determining purchase volumes, planning sales, planning and allocating production, planning machinery and equipment repairs, planning and generating purchase orders, shipments, etc. Demand forecasting is also important to inform processes such as financial planning and risk management in an organization.

Key Principles of Demand Forecasting

To construct the forecast as effectively as possible, it is necessary to known them. A few basic principles to be sure to keep in mind when making demand forecasts:

1. It’s more effective to give a forecast for several goods than for a single one;
   
2. The shorter the planning horizon, the more accurate the forecast will be (due to changes in the market).
   

Forecasting is divided into several types depending on the timeframe for which it is made:

• Prompt (up to 30 days);
  
• Conjuncture (Quarter-Half Year);
  
• Short-term (1-2 years);
  
• Medium-term (2-5 years);
  
• Long-term (5-10 years);
  
• Perspective (more than 10 years).
  

Qualitative and Quantitative Methods of Demand Forecasting

A perfectly crafted forecast consists of quantitative and qualitative steps. This requires statistical data from multiple sources across the supply chain.

Qualitative data is obtained from external sources. It can be news, social media, competitor analyses.
Quantitative data is gathered from internal sales statistics, search/web analytics. Advanced analytics incorporates modern collection and analysis tools (artificial intelligence, databases) for fast data processing.

Expert and Statistical Methods of Demand Forecasting

Conventionally, we can divide the methods of demand forecasting into two groups: expert and statistical.

Expert Method. The forecast is made as a person's subjective assessments. No one can accurately predict market dynamics, so the risk of this method lies in the human factor.

Several types of this method are practiced:

• Commission — Experts peer analyze and make a forecast;
  
• Interviewing — Talking with an expert on complex aspects of the forecast;
  
• Questionnaire — Interviewing an expert or several by questionnaire;
  
• Delphi Method — A compilation of other methods into a comprehensive survey of experts.
  

Statistics. This method uses calculations and analysis of demand statistics, identifies future changes based thereon. Include:

• Simple Moving Average (SMA). Demand is calculated based on sales over a certain period (days, weeks, months). Stably selling goods are analyzed by it, because only actual sales are taken into account without factors of dynamics. Because of this, an unsteady-selling product can make large margins of error.
  
• Schreibfeder weighted average. Here, too, actual sales are used, but the result is ranked by the importance of product characteristics - these are called weights. The system of weights is chosen according to the needs of the firm and on its basis they make their own (seasonality, age of sales)
  
• Exponential Smoothing (ES). The demand calculation is based on the number of past sales and the forecast of the past period.
  
• Autoregressive Method. Selecting several periods for which you want to count sales. For each period, a set of coefficients is developed to calculate the long period taking into account shifting factors (seasonality, number of working days)
  

Forecasting Tools

When using a statistical method to forecast market demand, the main tool is software because it’s very difficult, time-consuming and expensive to collect statistics manually. The software should be user-friendly, with clear algorithms. In addition to the software, the tool is the methodology of the forecasting system.

Application of Demand Forecasting in Supply Chains

Demand forecasting in supply chains became more popular after the pandemic when customer needs and expectations began to change rapidly, accelerating the dynamics of changing demand. Most companies optimize and integrate supply chain management practices into the business.

How Demand Forecasting Improves Supply Chain Planning

Without competent demand forecasting, taking into account all customer needs and the market situation, companies risk freezing funds in surplus stocks of raw materials or finished products, incurring losses when working with perishable goods, losing revenue and profits due to missed sales opportunities, losing customer loyalty due to delayed deliveries or regular unavailability of necessary goods. Thanks to the forecasting system, you’ll be able to optimize procurement volumes and eliminate operational and financial risks for your business.

Proven Integrated Planning Tools by GoodsForecast

GoodsForecast Integrated Planning Platform is a modern business planning platform for manufacturing and trading companies using advanced artificial intelligence. Its included GoodsForecast.Demand Planning module is designed to support demand forecasting and sales planning as part of the Sales and Operations Planning (S&OP) process. The module includes customized forecasting models and enables interactive construction of sales plans with the participation of experts.

How our Tools Help with Demand Forecasting

GoodsForecast solutions support all planning processes in the enterprise. They can be flexibly customized to meet the needs and specifics of the enterprise at minimal cost.

The system allows to consider the mutual influence of all links of the end-to-end supply chain and provides a unified information space, thus contributing to the improvement of business reputation and financial performance.

The combination of methods, high quality mathematical models and high speed data processing makes it possible to provide the most accurate forecasts.

Comparative testing indicates that employing GoodsForecast models can improve the forecast quality, minimize the likelihood of errors and their percentage in predicting demand.

What’s the Advantage of Our Tools?

Our company took the 2nd place in the world rating of forecasting accuracy (M5 Forecasting Competition). This was helped by the math algorithms we implemented in the GoodsForecast platform. It’s not only effective but also convenient:

• A unified platform that integrates all of the necessary tasks to manage a company's supply chain;
  
• Flexible customizations;
  
• Utilizing the latest AI-technology combined with expert opinions.
  

The platform integrates all planning processes:

• sales planning,
  
• production planning,
  
• supply chain planning,
  
• procurement of raw materials and supplies,
  
• logistics,
  
• customer service, inventory management.
  

Project experience has shown that by using the GoodsForecast toolkit, the following outcomes can be achieved:

• Accuracy increased by 10-30%;
  
• The level of service increased up to 10%;
  
• Inventory levels decreased by 30-35% on average;
  
• Production costs reduced by 5-20%.
  

Stories of Customers Using GoodsForecast's Tools

Our firm created an automated stock control system for associated products at Gazpromneft fuel station chain in 2021. By 2022, this solution aided in automating roughly 300,000 hours of manual labor — which is 78% of the time that was formerly consumed by a business process. There was a 33% decrease in product shortages, a 12% reduction in stock levels, and staff were reassigned to other important duties.

Conclusion

Demand forecasting tools must be able to make forecasts with different horizons - from a couple of days to 1.5-2 years, consider various factors within the company and in the market, quickly process large data sets and support the modernization of the company's processes. GoodsForecast solutions have been proven in over one hundred projects. These highly developed mathematical algorithms are integrated into modern AI. We invest more than 20% of our working capital in AI research and refinements to make our tool the most effective for your business.