Oliver Schaer, Nikolaos Kourentzes and Robert Fildes, ISF2018, 18th June 2018
With shorter product life-cycles and increased competition, generating pre-launch forecasts is a vital task for companies. Forecasting the success of a new product is challenging as one need to estimate the market potential. In practice, this is typically done by expert judgment. However, there is substantial evidence that experts are biased when forecasting new products.
Alternative approaches which rely on large surveys or conjoint analysis can be expensive and provide limited data points before launch. Moreover, throughout the pre-launch phase, consumer preference for short life-cycle products change significantly. A potential solution for obtaining more timely data points is to use online user-generated information.
Pre-release buzz reflects the aggregate anticipation of consumers towards a new product. Various studies report improved forecast accuracy when incorporating pre-release information from sources such as search engines, blogs as well as discussions taking part in forums. However, a majority of them only investigate the forecasting potential for the first initial weeks of sales, which from an operational point of view might not suffice, given that traditional time series models a require reasonable size of sales history.
In this research, we provide insights on whether search traffic information from Google Trends is useful in improving the estimation of the market potential before product launch. Search traffic is particularly interesting as it is regarded as a proxy for consumer interest in a product, but it also captures partially marketing expenditures. We (i) develop an approach to augment analogy based information from previous generations with pre-launch search traffic; (ii) compare the forecast performance against forecast from traditional analogy based methods and (iii) investigate the leading properties of pre-launch buzz.