150 likes | 266 Views
San Diego, California. Sept 2012. Edward Montes. Verticalization As Applied To Advertising As An Enterprise System.
E N D
San Diego, California. Sept 2012. Edward Montes Verticalization As Applied To Advertising As An Enterprise System
Digilantis an independent marketing technology company that provides marketers with a platform to support the entire media buy –from planning through execution, measurement and optimization– with ultimate precision and transparency.
Advertising As An Enterprise System An Enterprise System represents a cross-functional, integrated information system, used by organizations to support business processes and provide an underlying platform for data integration
Generic Systems • Systems are complex, tend to be comprehensive, tightly integrated software consisting of business process or logic that is hard-coded into the system • “One Process Fits All” • Options: 1) Reconfigure the system or 2)re-engineer the process to fit the system • One Process Fits All Approach Destroys Competitive Advantage
Vertical Strategy • A strategy aimed at minimizing the gap between practiced processes and system embedded processes • Specialized software for specific verticals where industry-specific processes are embedded into the system from the design stage • Better Tailoring Products To Practices
Today’s Truisms • Too Many Choices • Competition Is Fierce • Black Box Mentality • No Material Difference between Targeting and Valuation • Optimization is the equivalent of winning in a specific attribution model
Simple Methodology One day attribution: • The process takes one day of clicks /conversions and 45 days of preivous impressions/clicks and markseacheventwiththeircorrespondingweightsaccordingtothe attribution models. • The process isthenappliedforeach day with at least 45 previous days of data. 0 0 0 1 0 1 0 1 1 0 Target I I I C I I I C I C Conv Time Day of execution 45 days of imp / click
More Methodology Period with full information about target Time Day of execution 45 days of imp / click 45extra day to check if impression/ click is a zero • The first 45 don’t have all the possible events that could have caused the conversion / click, and the last 45 don’t have the conversions / clicks needed.
Click and Conversion Models: Predictors • A number of target basedpredictorsexistforeachmodel: • Target iscomputed in previous days of theeventsselected in thesampleaggregatingby: • Geo • Url-domainHierarchy (Urlisusedifit has enough data, domainisusedotherwise) • Twodifferentdepths are usedto capture short-term and long-termeffects: • 2 previous days (one predictor for geo and one forurl-domainhierarchy) • 20 previous days (one predictor for geo and one forurl-domainhierarchy) Time 20 days aggregation Day of sample 2 days aggregation
Modelling Results: CPG Click Attribution Target based predictor by URL-DOMAIN hierarchy for target [TARGET] and depth [DEPTH]
Modelling Results: CPG LCLI Attribution 20 Day surfing behavior still strongest predictor but 2 Day GEO importance jumps dramatically
Targeting vs. Valuation • Understanding the most important predictors of performance allows you to target to desired outcome • Understanding most important predictors of price, in conjunction with target, creates efficiency
We believe the future of Advertising Technology is to allow for Customization At Scale: • Your Data • Your Inventory • Your Algorithm • .
For more information, please contact: info@digilant.com Or visit our website www.digilant.com