Demand forecasting for new product launches: The Arduino case

The industrial market in which Arduino S.r.l. operates is characterized by a great degree of uncertainty and competitiveness. The main objective of any business is to satisfy customers’ demand, which requires developing competitive products both in terms of quality and offered price. In order to achieve this goal, firms operate in a network made up of different stakeholders: among them, the most important are final customers, suppliers of raw materials and semifinished products, distribution and logistics companies, and service companies. The relationships among the stakeholders, together with the flows of information and materials between them, make up a structure known as the Supply Chain. Such structure links all the businesses sharing the same goal, which is satisfying the demand of the final customer. At every step of the Supply Chain, the product acquires greater economic value to the customer. The efficiency of the Supply Chain is ensured by the collaboration of all its actors and the alignment of their incentives. A factor which undermines the definition of a clear set of incentives and which can jeopardize the profitability of a business is the variability of demand. To assess the factors which affect the variability of demand, it is needed to analyze the demand generation process. Typical influencing factors are the heterogeneity of customers, seasonality of products, trends in the purchasing behavior of customers, or the heterogeneity and differentiation of the product range offered by the business. Efficient businesses try to forecast the demand for their products as accurately as possible. This is because the availability of accurate demand forecasts massively impacts a firm’s profitability, as it allows the company to maintain a high service level by fulfilling all demand, without incurring extremely high inventory costs which derive from excessive production. Also, an accurate demand forecast will allow the business to improve their procurement process by ordering the right quantities, thus overcoming the obstacles constituted by the uncertainty of the availability of raw materials, and the high lead times for their arrival. The aim of this Thesis is defining a model that is suitable to forecast the demand of new products launched on the market. As no historical time series of demand is available, the problem faced is in the middle ground between a diffusion model and a demand forecasting model. Therefore, the output of this Thesis should be a model that is able to estimate the behavior of the demand curve of the product and to assess the order of magnitude of overall demand in a set time period following the launch of the product on the market. Then, after the product has started its diffusion product, and when enough historical time series will be available, the forecast will be updated using traditional demand forecasting algorithms already in used by Arduino. In order to assess the reliability and performance of the proposed model, widespread forecasting error indicators such as MAD and RMSE will be used. Moreover, a comparison between the performance of the current process used by the firm, which relies on qualitative estimations made by our Marketing team and Supply Chain team, and the new proposed method will be carried out. The performance of both methods will be evaluated both in terms of impact on the inventory and of offered service level.