Agent Technology for eCommerce

1
Agent Technology for e-Commerce

• Chapter 4 Shopping Agents
• Maria Fasli
• http//cswww.essex.ac.uk/staff/mfasli/ATe-Commerce
  .htm

2
Consumer Buying Behaviour Model

• Consumer Buying Behaviour (CBB) theory provides a
  model that describes the actions and decisions
  involved in buying and selling goods and services
• Most CBB models involve six stages
• Need recognition
• Product brokering
• Merchant brokering
• Negotiation
• Purchase and delivery
• Service and evaluation
• Agent technology can be potentially used in every
  stage

3
Online shopping The problem

• Consumers attitudes towards online shopping have
  changed
• To search for a product, a consumer can
• Visit specific vendors sites that she is aware
  of
• Use standard search engines and keyword retrieval
  to identify potential vendors and products
• In each site visited the consumer can search for
  a product, its price, specification and other
  attributes

4

• This approach has several shortcomings
• There may be hundreds of vendors selling the same
  or similar products checking vendors requires
  time
• Returned results through standard search
  technology may be biased
• If more than one products are required there may
  be no single site that caters for all
• When visiting a new vendor, the consumer needs to
  get acquainted with new interfaces
  time-consuming and also hinders impulse shopping

5

• Vendors may allow users to sign up to receive
  alerts
• Completing lengthy forms may be required which
  may also require the user to provide personal
  information the users privacy is weakened
• Such services are impersonal

6
Using shopping agents

• Users have more choice, but there are too many
  choices information overload
• Shopping agents or shopbots can enhance the
  users shopping experience by
• Helping them decide what to buy
• Finding specifications and reviews for products
• Comparing products, vendors and services
  according to user-defined criteria
• Finding the best value products and services
• Monitoring online shops for product availability,
  special offers and discounts and sending alerts

7
Potential benefits

• For the individual user
• Time savings
• More vendors can be queried and better deals can
  be uncovered
• User can have access to smaller vendors
• Help them make educated decisions
• Psychological burden-shifting

8

• For the marketplace
• Shopping agents and reputation systems can help
  tackle fraud
• Increased competition
• Market efficiency
• Smaller vendors can be visible
• Shopping agents can be used not only on retail
  markets, but also on business-to-business (B2B)
  markets

9
Working for the user

• To be truly useful and work for the user they
  have to
• Be impartial i.e. provide unbiased information to
  the user
• Be autonomous, proactively seek to help the user
  for instance by checking for products etc.
• Preserve privacy when required, the users
  identity may have to be concealed to preserve her
  privacy
• Offer personalized services to the user
• Make comparisons based on multiple attributes

10
How shopping agents work
11

• Similarly to meta-search engines
  screen-scraping
• They parse HTML pages and look for specific
  information
• They rely on regularities in the layout of web
  pages
• Navigation regularity
• Uniformity regularity
• Vertical separation regularity

12
Limitations and issues

• Current techniques for extracting information
  rely on syntax
• Although the information required is stored in
  machine-processable and well-structured format,
  agent developers have no access to this
  information
• Heuristics are ad-hoc, difficult and
  time-consuming to develop and prone to errors
• The resulting systems are cumbersome and vendor
  specific
• New vendors cannot be discovered and queried at
  runtime
• Only able to retrieve limited information and
  comparisons are usually made on price alone
  vendors vendors do not like that, other
  attributes may be important (guarantee, service
  etc.)
• The information retrieved may be inaccurate

13

• Shopping agents make commissions in three ways
• For each hit made to the vendors site
• For sales that result from clickthrough purchases
• For a favourable placement on the shopping
  agents recommended lists
• Recommendation offered may therefore be biased
• There may be discrepancies between reported and
  listed prices due to commissions
• Such shopping agents may create the false
  impression that the best deal has been found

14

• From the vendors perspective
• Although shopping agents improve their
  visibility, they also put their products next to
  those of competitors
• To be competitive a vendor may have to reduce its
  profit margins

15
Shopping agents and Web services

• Web services can be used as gateways to the
  vendors web sites