July 12th, 2011

New version of u-Predict launched. Improved accuracy, greater efficiency & faster than ever!

u-Predict, IPONWEB`s world leading probability modeling and user response prediction technology has just gotten a new upgrade.

Probability modeling is essential for algorithmic decisioning and programmatic buying for yield optimization, advertiser metric optimization (eCPM, CTR, CPA), impression valuation and user response modeling – all essential in Real Time Bidding.

IPONWEB’s latest version of u-Predict, codenamed Delta, features better quality performance, faster learning and more efficient probability extraction than previous versions.

Under significant testing prior to release, Delta was able to achieve significantly better quality performance for the same basic cost. It is now also able to learn faster and provide better prediction modeling for new campaigns, further reducing the optimization learning curve.

Another significant improvement has been made in Delta`s ability to extract probability trees more efficiently, significantly improving decision processing speed.

Also included in this release, were a number of features that enable Delta models to become more stable against data corruptions, as well as improvements in the use of time sensitive data and `priors`. Priors were first introduced to u-Predict back in 2010, essentially they help u-Predict model key metrics like CPA or post click conversion that normally require large amounts of data to determine probable response accurately. Delta features enhanced time phasing of the u-Predict decision tree models, so that the most recent historical data is better incorporated and prioritized into any prediction model, creating a faster, more accurate, rolling learning model to offset prior assumptions.

If you would like to learn more about u-Predict Delta and its capabilities, please contact us for more information.