©2015 Apigee Corp. All Rights Reserved. Preserving signal in customer journeys Joy Thomas, Apigee Jagdish Chand, Visa.

Slides:



Advertisements
Similar presentations
1 ©2009 MeeMix MeeMix – A personalized Experience.
Advertisements

Experience Guided Shopping & Search Guiders ® Deliver Measurable ROI Through Reports Metric Reports Deliver Unique Customer Decision Insights Guiders offer.
Market Assessment for Small Businesses. Lecture Contents Marketing Mix/ Demand/ Demand Estimation Sampling Plan/ Data Collection and Analysis Market Survey.
A PowerPoint Presentation
1 Evaluation Rong Jin. 2 Evaluation  Evaluation is key to building effective and efficient search engines usually carried out in controlled experiments.
Seeing is Believing i.e. Conversation Manager Anne-Celine Bringer - Account Development Lead.
1 © 2010 SAGA Worldwide, LLC. All Rights Reserved.
Recommender systems Ram Akella February 23, 2011 Lecture 6b, i290 & 280I University of California at Berkeley Silicon Valley Center/SC.
Chapter 14 The Second Component: The Database.
Recommender systems Ram Akella November 26 th 2008.
The United States Postal Service processed over 150 billion pieces of mail in 2013—far too much for efficient human sorting. But as recently.
12 -1 Lecture 12 User Modeling Topics –Basics –Example User Model –Construction of User Models –Updating of User Models –Applications.
Overview of Web Data Mining and Applications Part I
WEB EDITORS MEETING Welcome. GOOGLE ANALYTICS Google Analytics provides powerful digital analytics for anyone with a web presence, large or small. It's.
Measuring and Monitoring Social Media Presence Measuring and Monitoring Social Media Presence Rim Dakelbab.
The Business Intelligence Stack. Today Consolidation SAP buys Business Objects Oracle acquires Hyperion IBM acquires Cognos, SPSS Microsoft acquires ProClarity.
Cohort Modeling for Enhanced Personalized Search Jinyun YanWei ChuRyen White Rutgers University Microsoft BingMicrosoft Research.
Query Log Analysis Naama Kraus Slides are based on the papers: Andrei Broder, A taxonomy of web search Ricardo Baeza-Yates, Graphs from Search Engine Queries.
Computational Advertising Duygu Gunaydin Lu Li Shuanglong Wei.
LA.com Exclusive Guides Sponsorships. Los Angeles Newspaper Group 8 newspaper web sites in Southern California Impactousa.com 12 Million + Page Views/month.
Linkedin. What is Linkedin? Linkedin was established in May 2003 Operates the worlds largest professional network on the internet Linkedin’s mission is.
Big data analytics with R and Hadoop Chapter 5 Learning Data Analytics with R and Hadoop 데이터마이닝연구실 김지연.
WEB ANALYTICS Prof Sunil Wattal. Business questions How are people finding your website? What pages are the customers most interested in? Is your website.
Walter Hop Web-shop Order Prediction Using Machine Learning Master’s Thesis Computational Economics.
Google Confidential and Proprietary 1 Google Russia Business Opportunities and Ads Update.
3 rd Party Data & Audience Targeting © All rights reserved. 3 rd Party Data – Collected both online and offline by 3 rd party data companies such.
Precision Going back to constant prop, in what cases would we lose precision?
AdWords Instructor: Dawn Rauscher. Quality Score in Action 0a2PVhPQhttp:// 0a2PVhPQ.
Application of SAS®! Enterprise Miner™ in Credit Risk Analytics
FALL 2012 DSCI5240 Graduate Presentation By Xxxxxxx.
242/102/49 0/51/59 181/172/166 Primary colors 248/152/29 PMS 172 PMS 137 PMS 546 PMS /206/ /227/ /129/123 Secondary colors 114/181/204.
Confidential 2008 Roadmap. confidential 2008 Solution Roadmap Main Themes The ChallengeOur Approach Actionable Analytics Non effective data analysis with.
USING HADOOP & HBASE TO BUILD CONTENT RELEVANCE & PERSONALIZATION Tools to build your big data application Ameya Kanitkar.
From Devices to People: Attribution of Search Activity in Multi-User Settings Ryen White, Ahmed Hassan, Adish Singla, Eric Horvitz Microsoft Research,
Copyright © 2009 Pearson Education, Inc. Slide 6-1 Chapter 6 E-commerce Marketing Concepts.
© 2009 Eyeblaster. All rights reserved Gen Citro – Sales Engineer EMEA Channel Connect for Search EB Orange 246/137/51 EB Green 52/70/13 EB Gray 161/161/161.
Anindya Ghose Sha Yang Stern School of Business New York University An Empirical Analysis of Sponsored Search Performance in Search Engine Advertising.
Page 1 CSISS Center for Spatial Information Science and Systems Design and Implementation of CWIC Metrics Weiguo Han, Liping Di, Yuanzheng Shao, Lingjun.
Google Confidential and Proprietary 1 Google University Google Analytics and Website Optimiser Dyana Najdi, Customer Analytics Manager, EMEA Lee Hunter,
Analytics Overview What’s Out There? What You Are Already Aware Of What You Might Not Be Aware Of New Features! What You Aren’t Using in Google.
Bayesian networks Classification, segmentation, time series prediction and more. Website: Twitter:
To Blog or Not to Blog: Characterizing and Predicting Retention in Community Blogs Imrul Kayes 1, Xiang Zuo 1, Da Wang 2, Jacob Chakareski 3 1 University.
Querying Business Processes Under Models of Uncertainty Daniel Deutch, Tova Milo Tel-Aviv University ERP HR System eComm CRM Logistics Customer Bank Supplier.
Moving Beyond Standard BMV Reports Using Data Repository Session 373 Presented by: Ian Proffer.
Operated by Public Health England Making the most of Weblogs Web analytics in brief How to use ‘Google Analytics’ What can we obtain from raw data.
© 2009 All Rights Reserved Jody Underwood Chief Scientist
Page 1 CSISS Center for Spatial Information Science and Systems CWIC Metrics: Current and Future Weiguo Han, Liping Di, Yuanzheng Shao, Lingjun Kang Center.
CONFIDENTIAL1 Hidden Decision Trees to Design Predictive Scores – Application to Fraud Detection Vincent Granville, Ph.D. AnalyticBridge October 27, 2009.
242/102/49 0/51/59 181/172/166 Primary colors 248/152/29 PMS 172 PMS 137 PMS 546 PMS /206/ /227/ /129/123 Secondary colors 114/181/204.
Presented to: Space 150 Dan Murphy Triton Digital.
Copyright © 2001, SAS Institute Inc. All rights reserved. Data Mining Methods: Applications, Problems and Opportunities in the Public Sector John Stultz,
Who are we? Palo Alto based startup focused on e-commerce conversion solutions Founded in 2011 with expertise in SaaS, analytics & e-commerce marketing.
Learning Bayesian Networks For Managing Inventory Of Display Advertisements Max Chickering Mad Scientist Live Labs Microsoft Corporation Max Chickering.
DeepBET Reverse-Engineering the Behavioral Targeting mechanisms of Ad Networks via Deep Learning Sotirios Chatzis Cyprus University of Technology.
Introduction Web analysis includes the study of users’ behavior on the web Traffic analysis – Usage analysis Behavior at particular website or across.
Feature Generation and Selection in SRL Alexandrin Popescul & Lyle H. Ungar Presented By Stef Schoenmackers.
Fraud Detection with Machine Learning: A Case Study from Sift Science
We help businesses achieve online success! © All rights reserved. 8-digital.com - Proprietary and Confidential.
Advertising Overview. Types of paid ads SEARCH Bid on keywords on various search engines DISPLAY Pop-up Banner Mobile Social Video NATIVE Promoted (social)
BASICS analytics What’s SEO? The time machine has two settings:
ACTi Retail Big Data Solutions
Recommender Systems & Collaborative Filtering
Copyright © 2011 Pearson Education, Inc. Publishing as Prentice Hall
50 ABOUT KOBIT HOW TO USE SERVICE COSTS Member Registration
Leveraging HubSpot Enterprise To Grow Your Agency Services VARs Should Provide #VARFormable
Online Tool Screen shots
G-CORE: A Core for Future Graph Query Languages
50 ABOUT KOBIT HOW TO USE SERVICE COSTS Member Registration
Technical Capabilities
Ryen White, Ahmed Hassan, Adish Singla, Eric Horvitz
Presentation transcript:

©2015 Apigee Corp. All Rights Reserved. Preserving signal in customer journeys Joy Thomas, Apigee Jagdish Chand, Visa

©2015 Apigee Corp. All Rights Reserved. Overview Customers journeys and event data Customer Behavior Graph Queries on Behavior Graphs Predictive models on behavior graphs 2

© 2014 Apigee Confidential – All Rights Reserved Customer View: A journey 3 Consumers interact with enterprises through multiple channels at multiple points of time Each of these interactions is an event with a timestamp and the sequence of interactions is important

© 2014 Apigee Confidential – All Rights Reserved A graphical structure can identify common interactions and influences 4 Common Interactions & InfluencesCustomer Journey

Customer behavior graphs vs. social graphs 5 Behavior Graph Sequence of events: –Actions experienced and taken Social Graph Links between people & activities –At a point in time Behavior graph Social graph

©2015 Apigee Corp. All Rights Reserved. Model for User Behavior Users act on nodes in a temporal sequence of events USER PROFILE UserID: U56 Gender: M Geo: San Francisco Interests: Bikes, Fashion USER PROFILE UserID: U57 Gender: F Interests: News, Finance Age: NODE PROFILE Type: Content PageID: P100 Category: Product Review SubCat: Mountain Bike NODE PROFILE Type: Creative ID: Creative95 Category: VideoAd Advertiser: BikePros EVENT Type: PageView UserID: U56 PageID: P100 TimeSpent: 180 seconds Scrolls: 3 EVENT Type: AdView UserID: U56 AdID: Creative95 PlayTime: 30 sec Rewinds: 1

©2015 Apigee Corp. All Rights Reserved. Aggregated Behavior Graph (ABG) Impressions: 1 TimeSpent: 20 Clicks: Impressions: 4 TimeSpent: 10 Clicks: 0 Impressions: 5 TimeSpent: 30 Clicks: 1 Combine Characteristics Represents flow & behavior of all users Automated construction from event streams Information preserving Aggregated representation Permits drill-down Useful for reasoning about customer flows Count unique users at node/edge Aggregate metrics at nodes/edges Measure drop-offs on a path (funnel) Profile traffic at a node or edge Analyze flows for user segments

©2015 Apigee Corp. All Rights Reserved. Examples of queries on Behavior Graph Count the number of users who went from A -> B -> C -> D Find the distribution of (Age, Gender) for the people who took the path P-> Q ->R Of all the females in California who went from C to D, what are the most likely nodes that they are likely to visit next Of the people who bought a computer 3 months ago and received an offer for a discounted printer 1 month ago, what fraction of them have bought printers in the last month All these queries would be painful to express in SQL on a large event table 8

©2015 Apigee Corp. All Rights Reserved. Predictive Analytics on Behavior Graphs Past behavior of consumers is the best predictor of future actions The behavior graph allows one to search for patterns of consumer behavior that are correlated with responses of interest Using the patterns we can build a Bayesian model to predict what users will do next Use the predictive model for recommendations, targeting and churn prediction 9

©2015 Apigee Corp. All Rights Reserved. Comparison with other Machine Learning algorithms Most machine learning algorithms assume that the training data for a learning algorithm is in a form of a large table of examples, with responses in one column, and features in other columns, e.g. Logistic Regression, Random Forest, etc. These algorithms are designed for profile attributes such as age, gender, country, etc. To handle event data, the data scientist typically creates aggregate features out of the event data (e.g. total purchases over the last year, total purchases over the last month, etc.) The behavior graph allows the data scientist to automatically search over a large space of aggregates to use in the predictive model 10

©2015 Apigee Corp. All Rights Reserved. Summary Event data should be treated differently from profile data A graphical data structure designed for event data can efficiently answer queries on event based patterns Event based patterns can be used to build predictive models for targeting, recommendations and churn prediction There is a need for a common query language to express queries for event data 11

Thank you