Web Personalization Techniques for Global Audiences

1
Web Personalization Techniques for Global
Audiences

• From B2C to B2B

Henry Chang Manager, B2B Services E-Marketplace
and Supply Chain IBM T.J. Watson Research
Center Hychang_at_us.ibm.com
2
Definition and Overview
3
Personalization
Definition Use knowledge about a customer to
merchandise, present, modify and deliver products
and services most appropriate to that individual
at that time.

• Real Life Example
• Giving a personal gift
• Giving comforting words to people in suffering

• It must have
• Knowledge about the customer
• Knowledge about supply choices
• Business rules about what to offer
• Dynamic web delivery mechanism

4
A Personalized Shop Page
5
Personalization Illustrated
Analyze who The User is
Analyze what We have
User Profiling
Content Management
Match making
Marketing Control
Deliver Personalized Service
Satisfied Customer
Dynamic Delivery
6
Before Personalization

• Functional web sites
• Customer see the same site
• No knowledge of a customer
• No behavior change based on who the customer IS
  or what the customer did

7
After Personalizations

• User Scenarios
• Targeted marketing campaign
• Up-Selling / Cross-Sell / Co-selling
• Advanced cataloging Packages, Bundles, Kits,
  Sets
• Promotions Featured Product Page
• E-Coupons Personalized in-store / on-site
  coupons
• Advertising On-site banner ads

• Shopping Personalization Gift Registry, Gift
  Certificates, Gift Recommendations
• Member Book Persona l /Order information
• Broadcast e-mail / Fax Timely targeted
  notification
• Personalized view myYahoo
• Buying preferences
• Customer support
• Content Indexing Support a variety of datastore
  formats and multiple content sources Efficient
  indexing for searching

• Personalized pricing dynamic pricing based on
  business and customer factors
• Triggered value-added services Notification of
  price change threshold, currency conversion,
  insurance recommendation

8
Requirements and Business Drivers -- B2C and B2B
9
The On-line Relationship
Business to Consumer Personalization

• Relationship Challenges
• Customer Attraction
• Customer Retention
• Increasing Customer tie-in
• for revenue or influence
• growth

Informational
Transactional

• Home Page
• Programs
• Contact Info
• News

• Shopping Flow
• Catalogue Select
• Registration
• Form Feedback

Web Sites
10
Benefits of Personalization

• The information gathering, analysis, and dynamic
  content delivery possible with the
  personalization server make it economical for
  businesses to manage the entire customer
  lifecycle for individual relationships a million
  (or more) users at a time.
• Awareness and Acquisition, customer is just
  browsing or may be looking for some type of
  content or product of interest to them. If they
  find something of interest to them, they will
  most likely return.
• Segmentation and Targeting, customer may have
  provided some information via a registration for
  or by making a purchase. Using this information,
  the business manager can make decisions about the
  customers behavior.
• Promotion and Reinforcement, the business
  manager has set up some rules about what type of
  content or product information to show the
  customer with similar interests or behavior
  patterns. Dynamically presented content or
  promotion information can help move the site
  visitor toward purchase or repeat purchase.
• Transaction and Fulfillment, Customer has made a
  commitment to the business, through a product
  purchase or site membership -- that individual
  purchase needs to be fulfilled based on
  customers profile and transaction information.
• After-sale Support, by tracking, storing, and
  analyzing information about the customer in order
  to provide the service that will keep the
  customer come back to the site.
• Cross-sell and Up-sell, with knowledge of the
  customers past purchase history and expressed
  needs, the business manager can deliver
  appropriate information to the customer and build
  loyalty.

11
Moving Commerce Online
Manufacturer
Buyer
Distributor
RawMaterials
B
S
B
S
MP
MP
Network-based Commerce Services
12
Trading Process Integration
Business to Business Personalization
Aggregation
Collaboration
Transactional
Informational

• Aggregation of Sellers
• Aggregation of Buyers
• Business Flow Integration

• Solutions
• Programs
• Contact Info
• News

• Request for Quote
• Catalogue Select
• Bid
• Auction

Market Places
Web Sites
13
Business Process from B2B to B2C
B-to-B
B-to-C
Warehouse and Inventory Management
Customer Relationship Management
ProductDevelopment
Supplier Management
Transport
Category Management
Store Management
ProductDevelopmentManagement
E-ProcurementSoftware
Supply ChainManagement
CategoryManagement

• Product design
• Product development
• Vendor capacity
• Quality
• Sample management
• New product introduction

• Invoice processing
• Payment
• RFI/RFP
• Catalog development
• Vendor certification
• Item management
• Product availability
• Production planning
• Technical specifications
• Second- and third-tier suppliers

• CPFR
• VMI
• Order management
• Product tracking
• Capacity management
• consolidation
• Replenishment
• Reverse logistics

• CPFR
• Goods-in scheduling
• New product introduction
• Planning for seasons and events
• Allocation
• KPIs
• Ranging

14
E-Commerce exchanges are reshaping the B2B
marketplace
B2C (Business to Consumer) B2B (Business to
Business)
B2B Market Opportunity 1.4 trillion by 2003
Industry-Centric Exchanges
Market Makers

• Vertical
• VerticalNet
• Chemdex/Ventro
• Horizontal
• MRO.com
• PurchasePro.com

• Aerospace (Boeing)
• Automotive (GM/Ford)
• Agricultural (Cargill)
• Chemical (DuPont)
• Foods (GMA/PWC)
• Retailing (Sears/Carrefour)
• Utilities (I5 major/PWC)

Technology Platform/Engine
Source Forrester Research, Merrill Lynch
Internet Research
15
Personalization Techniques
16
Personalization for B2C
Analyze who The User is
Analyze what We have
1. User Profiling
2. Content Management
Match making
3. Marketing control
Deliver Personalized Service
Satisfied Customer
4. Dynamic Delivery
17
A Reference Personalization Architecture (Tivoli
TISM)
18
The Match Making Engine
19
(No Transcript)
20
Customer Profile

• From multiple sources
• User Survey of preferences
• Business registration to establish entitlements
• On-line behavior and analysis of past activities
• Active location wireless-access location-based
  talk pad (museum)
• Ldap server to store the common profiles
• E-person Standards
• Allow each functional components to have its own
  additional profile parameters

21
(No Transcript)
22
(No Transcript)
23
Match Making Engine
24
Market Analysis of MatchMaking Technology
25
Collaborative Filtering

• While collaborative filtering is an adequate and
  relatively painless way to personalize product
  recommendations, offers, and ads, it doesn't work
  for all products or all retailers.
• What is needed in an e-commerce application is a
  real-time recommendation engine that is capable
  of taking inputs attributes about the buyer,
  attributes about the products available, and a
  set of business rules and then formulating a
  recommendation.
• The techniques and algorithms used to come up
  with the recommendation can include collaborative
  filters, neural nets, or other AI techniques.
  These can be used singularly or in combination,
  or you can drive the recommendation purely with a
  set of business rules (e.g., if inventory gt
  30 day supply, offer 10 discount).
• The buyer attributes can be explicit (e.g. they
  fill out a survey) or implicit (e.g. analyze
  their click stream, look at purchase history,
  compare their profile to others). Likewise, the
  product attributes can be explicit (e.g.
  captured in the product database) or implicit
  (e.g. via shopping basket analysis of
  buyers with similar profiles).

26
Leading Real-time Recommendation Engines

• First company with U.S. patents on its technology
  so they can claim to have pioneered collaborative
  filtering.
• Offer the ability to return a single set of
  recommendations from multiple users ratings, and
  the creative parsing of databases to tailor
  returns more accurately
• Filtering engine delivers confidence-level
  ratings for each user.
• The product relies on CGI scripts and C
  libraries, and it works with any ODBC-enabled
  database.
• Webmasters can user sample HTML templates to help
  get a filtering-based site up and running
  quickly. Also lets Webmasters handle filtering
  data by genre easily.
• Integrated with Art Technology Group and
  Broadvision

Andromedia - LikeMinds Preference Server
Net.Perceptions - GroupLens

• Plug-in for Marimbas Castanet so that pushed
  content can benefit from collaborative filtering
  smarts.
• Lets Web sites ascertain user preferences
  implicitly--that is, based on actions a user
  makes on the site.
• Lets sites incorporate a variety of feedback from
  users. Users can rate explicitly by giving
  thumbs-up/thumbs-down (or in-between) ratings to
  items.
• Appealing and distinctive features for Webmasters
  such as the ability to program in C or Perl or
  in either. Webmasters will have to contend with
  only a few APIs and functions.
• Version 2.0 includes support for ODBC databases,
  more accurate filtering and multiple
  recommendation algorithms.
• Integrated with Broadvision and IBM Net.Commerce

27
Personalization for B2B

• Vendors have not explicitly addressed the needs
• Extending the B2C personalization mechanism to
• Active profile of the community
• Business processes
• Intelligent value-added brokers

28
B2B Auction/PO Business Scenario
Seller Advisor Broker
A
Buyers
Sellers
Excess Inventory Auction
X
Buyer Advisor Broker
Notify Buyer
A
Generate PO
X
eMP
29
Discussion (1)

• Capturing business data and business rules of the
  trading community
• Explicit transactions
• Implicit trend analysis
• Adding third-party data
• Driving campaigns based on Geo
• Cross sell/Up sell
• Community building
• Substitute-selling replacement products for
  out-of-stock items

30
Discussion (2)

• Ability to syndicate subsets of master catalog
  for target buyers
• Ability for each member to control branding ---
  not lost in the sea of supplier lists
• Ability to customize presentation and customer
  experience
• Personal broker to intervene in the business
  process (say, automatic notifying of the
  replenishment time)

31
Personalization for Global Audiences
32
The Complexity of Global Audience Profiling

• Delivering personalized information based on
• Customer persona
• Procurement, CEO, Employee, QA
• Enterprise Size
• Small, Medium, Large
• Geo
• Germany, British, Spanish, Switzland
• Language
• French, German,

• And more
• Subject interests
• (multiple choices)
• Survey and defaults
• Industry
• Buying pattern
• Type of goods
• Volume

33
Match making and Content

• Multi-attribute match making, the complexity is
  similar
• Content DB is made easier by Unicode
• NLS input, Unicode storage, NLS output
• Supported by DB2, Notes, Oracle

34
A Technical Problem

• Java Server Page is the preferred method to
  create dynamic web pages
• Separating logic and HTML
• Embedded Java beans
• Bean logic handles globalization nicely with
  database queries
• Static text in JSP is hard to be handled

35
Solution Options

• Creating separate JSPs
• Converting static text into Java bean calls
• Use an additional filter to do text substitution
• Maintain translations in an XML file
• Use ltNLStextgt lt/NLStextgt
• Adds another pass to the JSP delivery

36
High Level Architecture of Option 3
37
Word and Sentence Mismatch
Current jsp ltNLSTEXT ID449gt Welcome to your
IBM Web site -- please lt/NLSTEXTgtltA
href"/..."gt ltNLSTEXT ID450gtpersonalizelt/NLSTEXTgt
lt/Agt ltNLSTEXT ID451gt your Web site now.
lt/NLSTEXTgt Current xml ltNLSTEXT
ID"449"gt Welcome to your IBM Web site --
please lt/NLSTEXTgt ltNLSTEXT ID"450"gt personal
ize lt/NLSTEXTgt ltNLSTEXT ID"451"gt your Web
site now. lt/NLSTEXTgt
38
Examples
39
B2C A Portal for ISP Access (Tivoli Internet
subscription manager)

• The screen is divided into
• Navigation bar where a users subscription set is
  mapped
• The ISP customization to push content to the user
• The user selected interests area

40
Personalized Brand Info
Personalized Interests Info
Personalized Nav bar according to the
subscription and where user has has visited
41
(No Transcript)
42
(No Transcript)
43
Conclusion
44
--

• Reflection
• Vendors and E-commerce evolutions
• Future

45
Reflection

• One-to-one full personalization was very SLOW in
  performance
• Quick and fast skin-deep personalization
• How are you, Mr. Holloway?
• My Yahoo
• Deep personalization
• Intensive computation has not been popular
• Chess playing as the ultimate example
• Collaborative filtering does not work everywhere
• Personalization has become a commodity
• Must have but not additional advantage
• Solution vendors are gaining

46
Vendor Grouping
Trading Network
Vendors who enables B2B activity
i2-Ariba-IBM mySAP-CommerceOne-GE
47
Personalization Evolution
48
Future The Human Factor

• Inside out transparencies vs. control and
  manipulation
• Culture-based personalization to provide
  additional dimension to e-commerce
• Example Western linear learning vs. Eastern
  spiral learning