PERFORMANCE METRICS

1
PERFORMANCE METRICS

• IAN HUNG BOBBIN LO
• YAN MA LOUIS SZETO

2
PERFORMANCE METRICS DEFINITION
The standard measurement for describing
performance
3
MEASURING ELECTRONIC SHOPPING EFFECTIVENESS
4
MOTIVATION

• Designing an online store with effective customer
  interfaces has a critical influence on traffic
  and sales

5
ELECTRONIC SHOPPING

• Electronic marketplace (EM) becoming a major part
  of commerce in the coming decade
• In 1995, cyberspace sales of 1B in US
• Total US retail industry of 1 trillion
• 2000 projection of 7B-117B

6
ELECTRONIC SHOPPING cont

• Changes how business is conducted
• Business strategy
• Technical infrastructure
• Government policies
• EM demographics
• How technology is used

7
ELECTRONIC SHOPPING cont

• Differences in EM and traditional store
• Service desk ? help button
• Layout ? pull-down menus, product indices,
  search engines
• Very important User Interface (UI)
• Should not assume customers know-it-all
• UI affects productivity in general

8
TRAFFIC AND SALE FACTORS

• 6 factors affecting traffic and sales
• Merchandise
• Service
• Promotion
• Convenience
• Checkout
• Store navigation

9
1/6. MERCHANDISE

• Customers prefer wide range of products
• 5 of stores had 500 items
• 62 of stores had lt 50 items
• Range of products causes 17 variance in traffic,
  but no affect on sales
• Physical properties cannot be sampled

10
1/6. MERCHANDISE cont
11
1/6. MERCHANDISE cont
12
1/6. MERCHANDISE cont

• lt 8 screen area contained images
• Large images Slow
• Small images Poor quality

13
1/6. MERCHANDISE cont
14
2/6. SERVICE

• Provide company profile history
• Quick responsiveness to inquiries
• Anytime, anywhere, any language
• Useful FAQs
• Credit, debit, and payment policies

15
2/6. SERVICE cont
16
2/6. SERVICE cont
17
2/6. SERVICE cont
18
2/6. SERVICE cont

• lt 33 provide company profile history
• 80 had lt 10 lines of such information
• 95 did not have links to related products
• 25 had help for product selection
• lt 9 had FAQ section
• 47 did not have interactive email
• 80 of stores had 1 complaints

19
2/6. SERVICE cont
20
2/6. SERVICE cont
21
2/6. SERVICE cont

• FAQ section increases traffic
• Feedback help section increases sales
• Delays in answering complaints a PR disaster

22
3/6. PROMOTION

• 1 hr promotion ? 4? in variance of sales and
  1.4? in variance of traffic
• Promotion sales, advertising, appetizer
  features
• 6 of stores offer Whats New section
• 76 did not offer incentives
• Related links an important part of promotion

23
3/6. PROMOTION cont

• Advertising
• Banners top/ bottom of page
• Spotlight the logo prior to entrance
• Features first thing you see
• Positions some ads are never read

24
3/6. PROMOTION cont
25
4/6. CONVENIENCE

• Layout, organization, ease of use
• Help function should assist convenience
• 12 had help sections
• Status indicators
• Short text, informative headlines
  (multi-layered), use of white-space, colors

26
5/6. CHECKOUT

• Long checkout ? lose customers
• Repetitive entry of information
• Ideal universal checkout interface
• Shopping cart scheme
• Remove items at will
• w/o moving everything at a time
• Shipping date, out-of-stock info

27
5/6. CHECKOUT cont
28
6/6. STORE NAVIGATION

• Contribution to ease of navigation
• Product search
• Sitemaps
• Products indices
• Site organization and design,
• Links to related products

29
6/6. STORE NAVIGATION cont
30
6/6. STORE NAVIGATION cont

• 4 of stores have site index
• 6 had product search engines
• 22 offered browsing buttons
• Product list ? sales by 61, ? traffic 7
• More description, pictures, and less click to
  purchase increase sales
• Link to master index (site map)
• Groupings, cross-store searches

31
WEB INFORMATION SYSTEMS

• Search engines are important, especially for
  large sites
• Navigation path of user unknown
• Every page must have navigation tools

32
MARKETING MERCHANDISING METRICS
33
MOTIVATION

• Internet has become the channel for sales and
  customer service
• Max return on investment
• Understand the effectiveness of their sites
• Make appropriate action
• Two main perspectives for analyzing the
  effectiveness of the sites
• Marketing
• Merchandising

34
MARKETING

• Definition
• Defined as the activities used to acquire
  customers to an online store and retain them
• Techniques
• Use of banner ads, email campaigns

35
MARKETING METRICS

• Traditionally
• Clickthrough rate the percentage of viewers who
  click on a banner ad
• Conversion rate the percentage of visitors who
  purchase from the store
• Recently
• Ad banner return on investment (ROI)

36
MERCHANDISING

• Definition
• Activities involved in acquiring particular
  products and making them available at the places,
  times, and prices and in the quantity to enable a
  retailer store to reach its goal, it also include
  how and where to display products, and which
  products to advertise and promote
• Responsibilities
• Product assortment and product display, including
  promotions, cross-selling and up-selling

37
MERCHANDISING METRICS

• Traditionally
• Web page hit counts
• Provide a broad indication of visitor interest
• Conversion rate
• Indicates the percentage of visitors who purchase
  form the store
• Useful for evaluating the overall effectiveness
  of the store
• Cannot understand the possible factors within the
  store that may affect the sales performance
• Recently
• Micro-conversion rates
• Provide detailed insight into the success of
  different web merchandising, product assortment,
  and site design strategies

38
MERCHANDISING ANALYSIS

• 1. Product assortment
• Deals with whether the products in an online
  store appeal to the visitors
• Merchant can adjust brands, quality, selection,
  inventory or price to optimize sales

39
MERCHANDISING ANALYSIS

• 2. Merchandising cues
• Different ways web merchants present their
  products to motivate purchase in online stores
• Cross-sells
• Refers the visitor to a web page marketing an
  item complementary in function to the item
  marketed on the current web page
• Up-sells
• Refers visitor to a web page presenting a similar
  but more upscale item
• Recommendations
• Highlights product pages that are likely to be of
  interest to the shopper based on knowledge of the
  shopper and the behavior of a larger population
• Promotions
• Refers a visitor to a product page for informing,
  persuading and/or reminding the shoppers about a
  product and/or other aspects of the site

40
MERCHANDISING ANALYSIS

• 3. Shopping metaphor
• Different ways that shoppers use to find products
  of interest
• 4. Web design features
• Different ways how hyperlinks group together

41
MERCHANDISING ANALYSIS
42
MARKETING VS MERCHANDISING

• Web Marketing
• Uses banner ads and referral sites
• Clickthrough rates, ad banner ROI
• Controlled in external sites

• Web Merchandising
• Hyperlinks and image links within the store
• Micro-conversion rates
• Controlled internally

43
PROBLEM

• Tracking and measuring
• Classifying each hyperlink by its merchandising
  purposes
• Tracking and measuring traffic on the hyperlinks
  and analyzing their effectiveness
• Attributing the profit of the hyperlinks to their
  merchandising cue type, shopping metaphor, and
  design features

44
MICRO-CONVERSION RATE
Product Impression
Look-to-Click rate
Clickthrough
Look-to-Buy rate
Click-to-Basket rate
Basket Placement
Basket-to-Buy rate
Purchase
45
MICRO-CONVERSION RATE
46
MICRO-CONVERSION RATE
47
MICRO-CONVERSION RATE

• Micro-conversion rate can be calculated for
• individual merchandising cue types
• individual products
• individual shopping metaphor types
• individual design features
• individual banner ads
• Mini-Conclusion
• All the individual hyperlinks pointing to product
  pages in various forms and purposes can be
  analyzed

48
PARALLEL COORDINATES STARFIELD VISUALIZATION
49
DATA REQUIREMENTS FOR THE ANALYSIS OF WEB
MERCHANDISING

• 1. To visualize the effectiveness
• Traffic data, from web server logs
• Sales data, from the database of associated
  commerce server
• 2. To visualize a complete set of
  micro-conversion
• Product impression data
• From content meta-data tags or
• Dynamically parser the content of a web page

50
DATA REQUIREMENTS FOR THE ANALYSIS OF WEB
MERCHANDISING cont

• 3. To examine merchandising purpose
• Implementation time data
• Static hyperlink labels
• Tagged hyperlinks and dynamically parsed
• Overall
• Data preparation is easier for dynamic online
  stores
• Dynamic URLs can be easily identified

51
CLICKSTREAM DATA

• Displays the progression of sessions in terms of
  micro-conversions among shopping steps
• To understand where the store loses customers
• To compare shopping behavior
• To understand the effectiveness of different
  merchandising tactics

52
PARALLEL COORDINATE

• Displays multivariate data sets to identify the
  relationship among the variates
• Used to visualize clickstream
• The sequential steps of look, click and buy
  represented by a series of parallel axes

53
PARALLEL COORDINATE VISUALIZATION
54
STARFIELD VISUALIZATION

• Displays product-oriented information helpful to
  understand the product assortment aspects of the
  store

55
STARFIELD VISUALIZATION cont
56
STARFIELD VISUALIZATION cont
57
STARFIELD VISUALIZATION cont
58
VISUALIZATION SYSTEM IN PRACTICE
59
VISUALIZATION SYSTEM

• Background
• Interactive visualization system implemented as
  part of the E-Commerce Intelligence project at
  IBM T.J. Watson Research Center
• Purpose
• Help users explore and interpret web usage data
  to maximize merchandising effectiveness
• Features
• Visualization of parallel coordinates, starfield
  graphs, color-coding, filtering, zooming, data
  sampling, dynamic querying, and summary data

60
VISUALIZATION SYSTEM cont
61
VISUALIZATION WEAKNESSES

• Does not identify the exact merchandising
  attribute leading to a sale or exit
• Does not track the session time at each sequence
  of the clickstream
• Assumes users traverse the site in a linear
  fashion

62
VISUALIZATION IMPROVEMENTS

• Enhance shopping step data
• Each session is currently identified by a unique
  timestamp corresponding to a data point on the
  parallel axis
• Does not indicate volume or time spent on each
  step
• Extend User Actions
• Apply visualization technique to different web
  paradigms (Online Auction)
• Require a different set of sequential steps with
  parallel coordinates (Click-to-Bid)

63
VISUALIZATION IMPROVEMENTS

• Explore other visualization methods
• Mosaic graphs
• Richer session categorizing variables
• Besides referrers, host names, etc.., include
  shopping metaphors, merchandising cues, design
  features, and customer profile information
• Challenging to quantify/classify variables
• Relationships among category variables may be
  studied for their impact on store performance
• Validation
• More empirical studies over longer time range
  needed
• Provide optimal set of visualizations to
  understand merchandising effectiveness with
  minimal effort

64
OTHER PERFORMANCE METRICS

• Web Server Log Analysis
• Lack integration with knowledge of site layout
• Generally does not relate Web usage data with
  their meaning in commerce
• Datamining
• Supplement visualization results
• Predictive Modeling
• Derived rewards stored in online customer profile
  for personalized product promotions
• Collaborative Filtering
• Dynamic merchandising mechanism for unregistered
  users (Amazon.com)

65
QUESTIONS