emetrics : aol and netezza-powered web analytics
DESCRIPTION
How does one of the world's most highly-trafficked websites, takes a best-of-breed approach to data analytics to fully understand and engage online participants. AOL utilizes clickstream data to create person-centric views of individual visitors and leverages behavioral targeting and predictive analysis techniques to optimize every message across every web interaction for individual visitors, increasing customer loyalty. Senthil describes how AOL links web data to actionable systems (email marketing, site personalization, etc.) to maximize the value of website optimization initiatives and how they mine social network behaviors and interactions to customize offers for increased conversion rates.TRANSCRIPT
Transforming Our Business with
a Best-of-Breed Approach
AOL and Netezza-Powered Web Analytics
Senthil Kumar Mohan
Projection
Projection is a voyage that collects data over time to predict and target behavioral patterns by combining web analytics data with back-office systems to achieve actionable solutions.
( )∑Forecasting
Optimization
Projection = fPredictive Modeling
Web Analytics?
“Web Analytics is the measurement, collection, analysis and reporting of Internet data for the purposes of understanding and optimizing Web usage. “
Advertising
Page View Visitor
New view of Web Analytics“ Web Analytics is an instrument for measuring both Quantitative and Qualitative metrics across the organization. ”
Suppliers
Partners
Sales
Advertising
HRM
Accounting
Budgeting
Audit & Control
CustomersChannels
Procurement
Logistics
Distribution
CRM
FINANCE
OPERATIONS
ERP / SCM
WebAnalytics
Metrics
Suppliers
Partners
Sales
Advertising
HRM
Accounting
Budgeting
Audit & Control
CustomersChannels
Procurement
Logistics
Distribution
CRM
FINANCE
OPERATIONS
ERP / SCM
WebAnalytics
Data Explosion
30M customers x 10 web actions x 2 geo x 2 time period x 12 months = 14,400,000,000 combinations
50 customers x 10 web actions x 2 geo x 2 time period x 12 months = 24,000 combinations
KnowledgeKnowledge
DataData
Optimization
Predictive Modeling
Forecasting
Ad Hoc Reports
Lost ValueLost Value
Value Delivered By Traditional Web Analytics Tools
Value Delivered By Traditional Web Analytics Tools
Drill DownReports
Traditional Web Analytics
KnowledgeKnowledge
DataData
Optimization
Predictive Modeling
Forecasting
Ad Hoc Reports
Lost ValueLost Value
Value Delivered By Traditional Web Analytics Tools
Value Delivered By Traditional Web Analytics Tools
Drill DownReports
Revenue
Cost
Loss
Fixed Cost
Profit
Alternate View
Data Transformation
Industry Approach
BI Analytical / Reporting Tools
Data Sets Data Sets Data Sets Data Sets
Data Stores
Data Marts
Optimized Approach
Data Sets Data Sets Data Sets Data Sets
BI Analytical / Reporting Tools
Large Massively Parallel Processing RDBMS
Data Transformation
Sampling Vs Full Dataset
Sampling fails on the following scenarios:
Change in Business Logic
Multi Domain Integration
Segment Analysis
Behavioral Targeting
Forecasting
Predictive Analysis
Database – Server – Storage: Consolidated
100x THE PERFORMANCE
½ THE COST
UNMATCHEDSIMPLICITY
The Netezza TwinFin™ Appliance
High-performance databaseengine streaming joins,aggregations, sorts, etc.
SQL CompilerQuery PlanOptimizeAdmin
Processor &streaming DB logic
Slice of User DataSwap and Mirror partitionsHigh speed data streaming
Hosts
Snippet Blades™
(S-Blades™)
Disk Enclosures
Case Study : Netezza Appliance
Key benefits of a Netezza-powered approach to web analytics
100x THEPERFORMAN
CE
½ THECOST
UNMATCHEDSIMPLICITY
Best-of-breed analytics tools deliver competitive advantage
Orders/Leads
Age Group
Actionable Insights & Metrics Look-alike Modeling
Customer Research
Satisfaction
A/B Testing
Textual Analysis
Income Group
Segmentation Logistics
Problem Resolution
Conversion Rates
Revenue: How/Why
DemographicUser Data
Targeting disparate data silos
Gender
Behavioral Targeting
Behavioral Targeting is about reaching people, NOT pages
Technology AOL has developed the gold standard in Behavioral Targeting
Scale AOL Media sites and advertising network reach 9 out of 10 consumers online
Integrity AOL operates with full respect for consumer privacy; we allow users to opt-out and we do not collect any personally identifiable information
Optimization Results-based performance using the Internet’s best optimization technology
Quality We combine best-in-class data providers with clean, well-lit inventory
Behavioral Modeling
SERVE RELEVANTADVERTISING
The segment population is eligible for relevant targeting across the AOL Media sites
USERS ARE SLOTTED INTO RESPECTIVE AOL MEDIA BEHAVIORAL CATEGORIES
Specific content consumption on AOL Media sites, and the recency and frequency of that consumption, is mined and modeled to
define and create AOL Media segments. Users are slotted into their respective segment(s) based on observed visitation behaviors and
these are stored in their cookie file
CONSUMERS BROWSE AOL MEDIA SITES ONLINE
Users on average spend more than 3 hours and 45
minutes longer on AOL than Yahoo! and Facebook*
User A recently visited
AOL Living
AOL Living
AOL Health
Moviefone
AOL Technology
parentdish
…delivered to“AOL Living audience”
*Nielsen Online, June 2010
Behavioral Targeting Suite
Audience BehaviorsTarget one of 150 behavioral segments (e.g. Auto Intender, Apparel Shopper)
Custom Audience BehaviorsTarget a custom segment of users who have displayed relevant, discrete behaviors
Audience ExtensionTarget a specific set of AOL Media users outside of AOL Media content at a lower CPM
Advertiser LeadBackRe-target users who visit your website
Creative LeadBackRe-target users who have clicked or seen an ad banner
Sponsorship LeadBackRetarget users who have been to specific sponsorship area of AOL
Reverse LeadBackTarget anyone but your users
SearchBackTarget users who have made a category-related search on AOL Search or AOL properties
BEHAVIORAL TARGETINGBehavioral Targeting is an
opportunity to reach users based
on displayed behaviors online:
content visited, search queries,
ads clicked and/or viewed,
actions taken on a website.
OUR SUITE INCLUDES THE FOLLOWING SOLUTIONS
Demographics Targeting
High
High
High
Recency of visits to YourWebSite.com
Inco
me
Age
Age: 50-65Income: LowRecency: Low
Age: 18-24Income: HighRecency: Low
Age: 25-35Income: HighRecency: High
Age: 36-49Income: LowRecency: High
Geographic/Daypart Targeting
User: Target users based on the DMA look-up, or destination query data
NON PII BASED TARGETING Target users in specific geographic locations: country, state, city, DMA, zip code (AOL Media), 3rd level zip code (Advertising.com network).
DESTINATION BASED (only on AOL Media)
Reach users on results pages based on the destination queries submitted on MapQuest or AOL Travel Guides.
GEO CONTENTSITE TARGETING Target content that is geographically or locally-centered.
TIME OF DAY Target users by time of day regardless of time zone.
DAY OF WEEK Target users by day of the week.
Site: Target users while they consume geo-focused content
Target users by time of day – or day of week – based on user time zones orEastern Standard Time ad server configuration.
Content targeting
How it Works:Content targeting is the by-product of Behavioral and Geographic targeting. By identifying the customers behavior, location and time we can deliver the relevant content.
1 2 3
Identify the Behavioral segment
Identify the DMA / Geo segment.
We deliver relevant content.
Demo, Geo and Behavioral data is used to create look-alike models. This model is an anonymous method of identifying users and matching them with relevant content.For example, people from specific DMA code typically like product X and tend to drive Y. Then whenever a visitor shares those characteristics, we can make more intelligent decisions about which content and ads to present.
Case Study : Look-alike modeling
Projection
KnowledgeKnowledge
DataData
Optimization
Predictive Modeling
Forecasting
Ad Hoc Reports
Drill DownReports
Revenue
Cost
Loss
Fixed Cost
Profit
Thank youhttp://www.linkedin.com/in/senthilkmohan