search word bid optimization

Business Problem

Companies are making significant investments in bidding on Google, Yahoo, and Microsoft search keywords.  There are far too many possible keywords and combinations of keywords to manage using manual processes.  Even bid optimization utilizing predictive models is sub-optimal because traditional statistical techniques are unable to digest the vast sums of data available generated through paid search (i.e. millions of test results every day).  The end result is wasted marketing resources.

Solution

Machine learning technology searches through vast quantities of data to find subtle predictors of behavior.  This technology can produce linked predictive models where predictive “quality scores” flow into predictions of number of impressions and predictions of number of clicks, etc.  The bid optimization solution proliferates millions of bids, thereby capturing thousands of opportunities for greater efficiency.

Results

Results were monitored daily and bids fine-tuned as needed. The optimization service reduced the client's cost per conversion in excess of 40%, while keeping the absolute number of conversions virtually the same.