ПРОГНОЗИРАНЕ НА ПРОДАЖБИТЕ НА БЪРЗООБОРОТНИ ПОТРЕБИТЕЛСКИ СТОКИ НА БАЗАТА НА ДАННИ ОТ ПРОДАЖБЕНИ ТРАНСАКЦИИ С ВЕРОЯТНОСТНИ МОДЕЛИ
USING PROBABILITY MODELS FOR FAST-MOVING CONSUMER GOODS SALES FORECASTING BASED ON TRANSACTIONAL DATA

Summary/Abstract: The interest in modelling the structure of consumer preferences and consumer choice behaviour in order to predict future purchase decisions has a long and rich tradition in marketing. Finding a practical solution to this problem is becoming increasingly critical with the increased availability of customer transactional individuallevel databases. The focus of this paper is on the application of specific class of probability models that are well-suited to meet the rising challenges a marketing manager faces. Our objective is to define, estimate and a priori validate a predictive probability model by aggregating the individual-level purchase history (i.e. purchase frequency and recency). We demonstrate the relatively high performance of NDB and BG/NDB models using transactional data of a cohort of new customers. This supports our thesis statement that probability models are very suitable for identifying and forecasting purchasing behaviour. The availability of precise individual-level predictions concerns any serious attempt to derive customer lifetime value on a systematic basis.

• Details
• Contents