Keyword portfolio optimization in paid search advertising

This paper uses investment portfolio theory to determine budget allocation in paid online search advertising. The approach focuses on risk-adjusted performance and favors diversified portfolios of unrelated or negatively correlated keywords. An empirical investigation employs averages, variances and covariances for keyword popularities, which are estimated using growth rates for 15 major sectors taken from the Google Trends database. In line with portfolio theory, the results show that the average keyword popularity growth is strongly related to the standard deviation of growth for each keyword in the sample (R² = 74%). Hypothesis testing of differences in Sharpe ratios documents a significantly better performance of the proposed approach compared to that of other strategies currently used by practitioners.