Software

Portrait Software Announces the Release of Uplift Optimizer 5.3.

Did you market to a customer that eventually purchased - OR - did your customer actually buy as a result of your marketing campaign? This is the question that as plagued marketers for years. Portrait Software, a leader in next-generation marketing solutions, has announced the release of its latest software update for solving this specific marketing challenge.

Uplift Optimizer is the industry's only packaged solution available today that can predict the change in customer behavior due to a marketing action. The solution packages 10 years of expertise into an easy to use application which can allow you to make an immediate impact to an organization's marketing.

How does it work? Moving beyond simple behavior and lift models (likelihood to buy, likelihood to churn, risk of default), Uplift Optimizer is able to predict how each customer's behavior changed specifically as a result of the marketing treatment through the incorporation of the behavior for both targeted and control customers. Among other results, marketers can now clearly predict which customers will act positively upon the marketing (the "persuadables"), which are indifferent (the "sure things" and "lost causes"), and which may in fact act negatively (the "sleeping dogs").

This uplift score allows Portrait's customers both to make 60-70% reductions in mailing volumes without reducing overall return, and to detect and correct negative campaign effects within their marketing. (Negative effects can be significant: in extreme cases a campaign can be "net negative" - i.e. worse than doing nothing - but still contain a sub-segment of customers that was positively impacted.) This approach has resulted in hundreds of thousands of dollars additional return from individual campaigns.

The latest release, Uplift Optimizer 2.5, extends the core uplift modeling algorithm of the system and delivers a guided best-practice methodology designed to work even more effectively in real-world applications:

• More flexibility:
  • Case weighting to handle treatment bias
  • Binary, continuous and two-stage uplift models
  • Bagging to average multiple uplift models across different data samples
• Enhanced robustness:
  • Variance-based pruning
  • K-way cross-validation
  • Stratification of training/validation splits
• Increased productivity:
  • Automated feature creation and variable reduction based on estimated goodness-of-fit (Qini)
  • Multiple runs for parameter tuning
  • Diagnostic reports including uplift & score heat-map profiles showing the actual and predicted lift in each segment, and a treatment stability report highlighting segments where the test/train split is potentially not balanced
• Improved integration:
  • SAS integration including SAS data input & output, SAS scoring code generation and SAS macro support for driving uplift runs from within a SAS script
  • Full integration with Portrait's Customer Analytics workbench, the Quadstone system, delivering high-performance, visual, interactive data exploration, mining and customer ETL capabilities

Contact Marlene Jaworksi at (617) 457 5218 for more information, or request our white papers.