A tool called Ad Price Prediction - released by online ad revenue optimization service PubMatic - could help publishers predict the cost of an ad unit in real time.

The predictions (pdf) are based on real time API pricing and machined algorithms. The tool has apparently already created a 70% increase on ad inventory for ten major publishers who participated in a beta trial of the unit, PubMatic said, using data from its older platform.

Comparing PubMatic to Google's technology used to predict clicks, company co-founder and CEO Rajeev Goel said that PubMatic technology can optimize every single ad impression automatically, which is a significant improvement over manual optimization that relies simply on best practices.

For publishers, PubMatic could be useful in determining the price difference for ads based on behavioral targeting and ads based on contextual targeting - and then serve the higher paying ad to the consumer.

To further improve the algorithm, PubMatic processes more than 100,000 data transactions for 6,000 publishing customers and uses the data collected through 'machine learning' as feedback.
According to PubMatic's AdPrice Index, released in January, 4Q08 reflected a sizable decline in online display ad pricing, compared to the same period in '07 across all verticals, regardless of website size.