Should I buy my new iPhone now? Predictive Event Forecasting for Zero-Inflated Consumer Goods Prices

Kurzfassung/Abstract

Price Comparison Sites enjoy great popularity because they enable customers to make better – more informed, less costly – buying decisions. We use a large dataset of daily minimum prices for 238 smartphones from Price Comparison Sites to develop a methodological approach to model the characteristics of consumer electronic goods prices and advice customers on their purchase time decision. The introduced method consists of two stages, the first one generating a multi-step price drop probability using a customized ARMA-GARCH model combined with an ensemble bootstrap. The second step constitutes a threshold for the estimated event probabilities that is optimized to achieve the highest possible savings for the customer. The derived recommendation advices users either to wait with the purchase or to buy the product of interest immediately to avoid losses due to increasing prices. In the evaluation, we compare our developed approach to two benchmark strategies: “always buy” and “always delay”, where the latter is expected to be the dominating strategy as prices show strong deterioration. The evaluation of 150,654 automated decisions shows that the proposed decision support system outperforms both benchmarks and delivers statistically as well as economically significant results. We reach savings of up to 26.67 € per recommendation and show that our fully automated decisions generate a significant added value and save money for the clients.