Most of us will nowadays search for products online before buying. We read product reviews, share our thoughts and interact directly with companies by liking and commenting on their social network posts. This digital footprint offers a new data-rich opportunity for analytics, freely available to organisations. Since the Cambridge Analytica case, there is a thin line between targeting individuals, and using this data in an aggregated form as an insight measure (read our blog post). In the context of business demand forecasting one prevalent question is whether such data can improve our forecasts. In the increasingly disruptive business environment sudden shifts in demand occur more often and challenge traditional forecasting method. The oft-made claim is that user-generated online information, such as searches and social network shares, can substantially improve forecast accuracy.
A recent study by the Centre for Marketing Analytics and Forecasting investigated this literature and found that most research suffered from experimental weaknesses, including lack of adequate benchmarks or rigorous evaluation. In two case studies our conclusions are in stark contrast to the literature, with results that question the usefulness of these inputs within an operational forecasting setting. To avoid over-claiming the results of our own analysis, we stress that we focused specifically on operational forecasts and further work should be done to evaluate online information for other uses. Nonetheless, our results show important thorough forecast evaluation is and how effective are established forecasting techniques. On the more positive side, this also implies that the debate on increasing consumers’ online privacy does not necessarily lead to a lost opportunity for improving companies’ forecasting abilities. Maybe consultants and (some) academics are claiming more than they can deliver!
Download the paperBack to News