I wonder if such result might be a result of some methodological error (not representative sample, some metric calculations biases, click frauds, etc). Otherwise this is rather sad article :(