Понравилась презентация – покажи это...
All Data Is Not Created Equal Brian Monahan, VP of Marketing @Brionic
2retail site largest nd Source: comscore multi-platform data including mobile ending July 2015
is a giant personalization machine
We have learned that data can be misleading Demographics Online Interests & Behavior Already Purchases Your Brand or Category ? Geographic Lapsed User of Your Brand or Category Complimentary Items Outside Your Category
The Power of Commercial Data ROC curve True Positive Rate True Positive Rate ROC curve False Positive Rate False Positive Rate
Historical purchase data is significantly more predictive for future purchases than any other demo/behavior data ~1000 # of features needed without prior purchase data 70% Accuracy (demo, social, geo, other behavior, etc.) ~10 # of features needed with prior purchase data at same scale 90% Accuracy (historical purchase)
How It Works Differences in key online and in-store purchase behaviors of shoppers exposed to media (test) and nonexposed (control) are measured and analyzed Proprietary first-party Walmart/Sam’s Club data matched to U.S. households Impressions are served across Walmart-owned media channels, tracking those exposed Matched data is anonymized and synced to online cookies Households are then used to segment Walmart.com shoppers for targeting and measurement
WMX Control Method Test users Attribute 1 Control users generated from algo Attribute 2
Media Auto Optimization WMX data allows Walmart to optimize media elements in real-time based on actual in-store sales, improving media over time. On average, auto-optimized tactics see 1.5x – 2.5x higher lifts (vs. non-optimized line items) Source: WMX Optimization Report for Pampers 81191
Modern Media Planning is beyond human comprehension … it takes a software solution.