Personalization

1
Personalization Interactivity
2
Interactivity A Multidimensional Construct
3
Response to Interactivity Depends on Motivation
for Visiting the Site

• Why do people visit Web sites?

• To search
• To find particular information while expending
  minimal time and energy
• Purposive, task-specific behavior
• Pre-purchase deliberation
• Think more about products

• To browse
• To be delighted and entertained
• Recreational/ experiential behavior
• Not efficiency-oriented
• Think more about Web site execution

4
Features Likely to Increase Online Purchasing
5
Sticky Sites
Sensory Breadth
Vividness
Sensory Depth
Telepresence Feeling present in the created
environment
Speed
Range
Interactivity
Mapping
6
Personalization Systems
Customer Preferences in Product Space Customer Preferences in Product Space
Key Product Attributes Uniform Highly differentiated
Qualitative, Complex (e.g., experience goods) Endorse Customer benefit choice simplification Firm needs satisfaction database Collaborative filtering Customer benefit educated word-of-mouth Firm needs extensive user data and clustering
Quantitative, Few (e.g., search goods) Rule-based Customer benefit interaction management Firm needs user models and observable triggers CASE Customer benefit show individuals utility function Firm needs extensive product database and user cooperation
7
Rule-based system
8
Collaborative Filtering

• Based upon other users overall ratings of items
  in the database, as well as the target users own
  evaluation.
• Identifies other people who are nearest
  neighbors (most similar) to the target user.
  Those nearest neighbors most similar to the
  target user are weighted most heavily.
• Prediction of what the target user might like is
  based upon what the weighted nearest neighbors
  liked.

9
A Closer Look at Collaborative Filtering
10
Criticisms of Collaborative Filtering

• Inefficient for marketers since you can not
  control what is recommended
• Black Box approach - if it works, we dont know
  why.
• Does it work?
• Industry models (LikeMinds, Firefly,
  NetPerceptions) are proprietary algorithms and we
  dont know how well they work.
• Selection of similarity index (who is closest to
  the target user?) and clustering algorithm (how
  do we identify the nearest neighbors?) is often
  atheoretical, adhoc, and a guess.

11
Product categories?

• Marketing issue is that you dont have any say on
  what product the software will recommend.
• Taste is important.
• Product must exist - not concept testing or new
  products.
• Consumer doesnt have sufficient first-hand
  experience.
• The fewer desirable items, the more effective.

12
Profile x Timing

• Increasing Levels of Knowledge
• Valid name database
• Valid name
• User ID
• Cookies
• Log Analysis
• Session online behavior
• Average visitor statistics
• Anonymous

13
Branded Response
14
Yesmail.com

• What solution would you favor for the recruitment
  of new members? The network solution? Build a
  proprietary list? Both?
• What does yesmail bring to its members? To its
  clients? Would you recommend that yesmail adopt
  a different pricing policy?
• What is the future of agents such as yesmail.com?
  On the basis of this, what would you recommend
  that yesmail do?

15
Onsale

• When Kaplan created the Onsale format, he called
  it a new form of retailing that exploited the
  unique characteristics of the net. In what ways
  is Onsales model different from traditional
  retail models?
• Is Onsale performing well? What are the
  benchmarks for performance in this market?
• Evaluate the AtCost model. What are the benefits
  and risks? How would you propose they overcome
  the risks?