Would be interesting to do this as a double blind study to evaluate bias in advertising