MODELING AND FORECASTING DEMAND FOR A DIGITAL PRODUCT

Abstract

Different approaches to modeling and forecasting the demand for a digital product, namely the paid activities of Facebook, are explored in the article. The company is given to reject its arrivals in the main form of advertisements. The available range of data on the company's activities allowed to build forecast models based on adaptive short-term forecasting methods, namely the Brown method and the adaptive multiplicative Holt-Winters model taking into account the quarterly seasonal factor. These models have the ability to continuously take into account the evolution of the dynamic characteristics of the studied processes, to adapt to these dynamics, giving weight and high information value to the available observations, if they are close to the current time. The models were tested for adequacy using a number of criteria, including the RS-test, the series criterion based on the median of the sample, the Student's t-test and the Darbin-Watson test. The comparative analysis of the obtained results by models allowed to choose a model that gives a fairly accurate result. The analysis also showed that there is a quarterly seasonality and, accordingly, a significant decline at the beginning of the year and income growth in recent quarters. The forecast for the 4th quarter of the next period is built. The forecast is compared with real data and the prospects for the development of digital products in Ukraine are determined, in particular, the spread of the use of digital services and products in many areas.