Nowcasting revenues with Google search volumes

Abstract

The usefulness of Google search volumes has been studied extensively in various economic areas, such as predicting product sales, stock markets, economic indicators, and tourism arrivals. The previous literature on predicting sales has been mostly focused on the prediction of specific products like certain movies or cars. A broader view of predicting revenues of whole companies with searches of their brand names has not been taken. Therefore, this thesis investigates the relationship between Google brand search volumes and revenues.

A sample of 18 public Finnish companies was selected for this study and their quarterly revenues and brand search volumes were collected. A baseline regression model without a search variable was compared to an explanatory model including the search variable. The coefficient of the search variable in the second model was analyzed.

The analysis showed a significantly improved model fit in terms of R-squared when including the Google brand search variable in the model. The coefficient of the search variable was positive and statistically significant.

The results indicate that including Google brand search volumes in a revenue nowcasting model can improve the model fit. The results also suggest a positive relationship between searching for a brand and the company’s revenue. However, the results do not come without limitations. The most important limitations were the selection bias when selecting the companies and the use of a simple regression model.