This document provides an overview of the big data and predictive analytics ecosystem by mapping out the key components. It shows how data flows from various sources through data management, processing, analytics and applications. It outlines different types of analytics and examples of applications across various vertical industries. The document also includes examples of real-time bidding and different players in the digital advertising ecosystem.
TechConnectr's Big Data Connection. Digital Marketing KPIs, Targeting, Analytics, & Optimization
1. Deep Dive
Marketing Big Data and Predictive
Analytics
Bob Samuels
TechConnectr.com
TechConnectr@gmail.com
@techconnectr
Graphic Source: Gleanster - An Intro to Big Data for Marketers
3. BI Platform / Reporting
OSS
Visualizations
Unstructured / Search
Indexing / Metadata
Search
NLP
Hadoop Analytics
Hadoop Dev Platforms / Automation
HDFS
Predictive Analytics
“Big Data” EcoSystemAPPLICATIONSTOOLSDATAMANAGEMENT
STRUCTURED UNSTRUCTURED
Transactional
DB
OSS
High Performance
Analytical DB
NewSQL
Enhancement
Distributed
NoSQL
Graph Document
Key Value /
Column
Enterprise
Apps
Internet
Apps
Social Media Web Content Mobile Devices Camera / DVR Sensors / RFID Logfiles
Hadoop
aaS
HDFS Alternatives
DBaaS
HANA
GraphDB
Filesystem
EMR
Text / Sentiment Analysis
Data as a Service
Data
Warehouses
vFabricL
Drill
Vertical Market Applications
Impala
Messaging Optimization Data Integration / CEP
OSS
IMDG
Redshift
Based on Source: Perella Weinberg Partners
AI
5. Bob Samuels
The TechConnectr – www.techconnectr.com
Cell: 408-206-5858
Strategic * Marketing Analytics * Client & Partner Door Opening * Demand Generation & Nurturing * Financial ROI Optimization
Real-Time-Bidding eMail Recommendation Engine Search Demand Side Platforms CRM Loyalty Programs
Display Web Analytics Games Customer Experience Mobile / Location SEO Video
Targeting / Personalization Community / Social Marketing Automation Yield Optimization Re-Targeting
Data Management Platform Sharing Tools Integrated Marketing Management Feedback / Surveys
Corporate Structured Data
Structured / Unstructured
Content Management
Data as a Service
Web Content / Search
Social Media
Images / Video
Mobile / Location
Sensors / RFID / Satellite
Machine / Log Files
Customer Personalization
Digital Mktg / eCommerce
Healthcare / Bioscience
Insurance / Risk Mgmt
Investment Management
Telecom / Utilities
IT & Operations
Manufacturing / Logistics
Oil & Gas Exploration
Government & Defense
Business Intelligence
Dashboards / KPIs
Data Discovery
Descriptive Analytics
Statistical Packages
Predictive Analytics
Machine Learning
Prescriptive Analytics
Decision Management
Graphs / Visualization
Hardware & Infrastructure
Natural Language Processing
ETL / ELT
Data Integration
Data Governance
Marshalling
MapReduce
Databases
Hadoop / In-Memory
Distributed File Systems
Digital Marketing Applications
DATA SOURCES DATA PROCESSING DATA ANALYTICS APPLICATIONS
6. Multi-channel two-way messaging
Website
Mobile site
Mobile app
CRM / ERP
POS
Call Center / IVR
Email
Display
Social
DATA
LAYER
Onsite
Online
Offline
Customer
History &
Profile
Credits to Ensighten for graphics
8. Increasing Value of Data
Data- BI – Predictive - Prescriptive
PrescriptivePredictiveBiz IntelligenceData Mining
9. Another way to look at Analytics Levels
Dash
Boards
Analytics
Prescriptive
Pivots
Predictive
http://practicalanalytics.wordpress.com/2011/05/01/the-vendor-landscape-of-bi-and-analytics/
10.
11. Business Intelligence Analytics / Visualization
Big Data BI & Analytics/Visualization
Solution Providers
Oracle Essbase Laurén
15. Example: Recommend Engine
Targeted eMail & Web Messaging & Timing
• Provide Recommended Action: re: Predictive Analytics and Patterns:
– Marketing spend effectiveness (how much to spend)
– Targeting (who to target; when, with what message; what medium)
– Promotion differentiation (how to differentiate offers)
– Contact strategy (how to contact customers over time)
• Targeting Precision by Groups / Clusters: (examples below)
– New: Predict lead conversion, welcome second offer; high predicted LTV
– Growth: Based on product interest browsing; Cart – Behavioral, Brand, Need Clustering & AOV
– At Risk – high value at risk; disengaged; high returns, complaints
– Lapsed – need-based cluster re: reactivation; Focus: high value-high size of wallet
• Use Collaborative Filtering & Clustering; Propensity Modeling
– Use in different contexts to solve different problems
– Start grouping by product behavior. And build in range.
• i.e. Shoe Retailer – distinguish moms from jocks from execs – clusters
– Can start contextualizing the email or the website for the individual
• Relevant, personalized eMail & web benefits include:
– Increase open and click rates while minimizing unsubscribe rates
– Predict which customers are most likely to engage, reactivate or complain
– Customize email frequency and content by customer segment
– Measure and optimize the ROI of email campaigns for specific customers
– Maximize email revenue and campaign performance
16. Unique Selling Proposition
• Relevance.. Key to e-marketing success
• Help identify which data us useful and which isn’t
• Help identify which algorithms are most useful
• Customer-focused & Marketing-focused analytics
– Better Relations with Customers (Satisfaction; Up-sell; Retention, Targeting to specific actions &
interests; Risk Management)
– Spend Marketing Money Wisely - Customer Acquisition; SEM, SEO, etc)
• Multi-media Sources – ex: are they up for renewal; which emails are they responsive
to; what pages are they looking at on website; any calls / complaints / inquiries;
• Predictive Analytics:
– Detect Changes of Behavior; Sources; Trends – quantity, quality – risks & opportunities
– Group – Buying Pattern; look at DNA; ie based on what they buy.. Old vs athletes, region
• With that, may merchandisestore, email differently
• clusteringmodels for products,brand and behavior.
– Predicting what is going to happen – what is likelihood of coming back to store, buy
– Correlative – if bought this, what is next thing to buy… look at similar person, neighbors
• Support
• UI / Ease of Use – run reports & analyses – answer questions
• Customer Metrics, Advanced Clustering, & Predictive Analytics Models
• Interfaces & APIs – social, web, email, POS, CRM, ERP, ESPs
24. Ad Exchanges & DSPs
Online Ad Exchanges DSPs
Examples: Yahoo! bought Right Media in April, Google
bought DoubleClick in May and Microsoft bought AdECN
in August , all in 2007
Examples: DataXu, Invite Media (acquired by Google in
2010), Turn, Mediamath, Xplusone, AppNexus, Acuity
Ads, (Rocket Fuel)
Enable bid-based ad “trades” between buyers and
sellers on their platforms. In this case, media buyers
have to use a different system to access each exchange.
DSPs allow media buyers to buy from multiple biddable
media sources through a single interface, which gives
buyer access to more liquid inventory.
Buying from multiple exchanges is time consuming and
inefficient from companies.
Manage, optimize, and execute bid-based buys. DSPs
also feature algorithmic optimization capabilities that
dynamically alter bid prices based on performance data.
Ad Exchanges is a layer below DSP.
DSP is a layer on top of AD exchanges. These companies
can access inventory from multiple exchanges with no
need to aggregate inventory through relationships with
publishers.
Typical campaign buys from multiple ad exchange so it is
difficult to achieve unique reach or optimal frequency.
Reach and frequency can be better controlled using one
interface.
Use of DSPs is constantly growing, but is still a small share in Overall Display Media Buying
Source: http://www.shilpagupta4.com/2011/09/09/quick-guide-to-demand-side-platform-dsp/