Automatic Data Collection: Server Logs

1
Automatic Data CollectionServer Logs
2
As with all methods, have to ask

• What are the goals for your system?
• What constitutes success, or good quality
  service?
• How can you conceptualize and operationalize
  quality?
• What information can you get using this method?
• How will this info help you evaluate performance?

3
Sources of data about visits and visitors

• Provided by users
• Registration, and whatever demographics and
  preferences are asked about
• Captured by system
• Server log files
• Cookies

4
Benefits of monitoring data

• Can yield lots of data for relatively low
  investment
• Unobtrusive outcroppings
• Numbers communicate well
• Numbers are useful for comparisons
• hits are up 20 over this time last year

5
However what do the data mean?
6
One example of simple stats

• Compare January (with DWR photos) to March (DWR
  photos removed)
• http//elib.cs.berkeley.edu/webstats/Mar2002.html

7
(No Transcript)
8
(No Transcript)
9
Common measures

• According to Forrester Research, many companies
  still use hits as the primary measurement of
  website success, followed by page views and
  session length.

10
Hit

• The retrieval of any item, like a page or a
  graphic, from a Web server. For example, when a
  visitor calls up a Web page with four graphics,
  that's five hits, one for the page and four for
  the graphics. For this reason, hits often aren't
  a good indication of Web traffic. See page view.
• http//www.webopedia.com/TERM/h/hit.html

11
Measuring success

• Companies sometimes make the mistake of buying
  elaborate software packages that analyze data a
  million ways, and then neglect to look at the
  most basic, day-to-day measurements of how a site
  is doing in its primary function.
• For an e-commerce site, those basic measurements
  are conversion ratethat is, the ratio of buyers
  to visitorsand average order size.
• For sites that make money via advertising
  banners the number of ad banners viewed
• other sites can measure traffic from return
  visitors versus traffic from new visitors.
• Remember one of the most basic elements of
  delivering a good customer experience making
  sure that pages load quickly, even when the site
  is barraged with traffic.
• http//www.cio.com/archive/051500_parade.html

12
(No Transcript)
13
Server logs contents

• Time
• IP Address
• Server
• Action
• Object
• Result code and size
• Browser version and platform
  
• Referring URL

14
Server log contents

• Time IP Address Server
  Action Object Result code and size
  Browser / version and platform Referring URL
• 015017 216.126.148.89 - ICICWEB1 GET
  /images/pdq.gif - 200 793 290 Mozilla/4.0(compati
  bleMSIE4.01Windows98) http//128.231.164.
  190/pdq.html
• 015018 216.126.148.89 - ICICWEB1 GET
  /images/banner1.gif - 200 4067 294
  Mozilla/4.0(compatibleMSIE4.01Windows98)
  http//128.231.164.190/pdq.html
• 015018 216.126.148.89 - ICICWEB1 GET
  /images/news.gif - 200 1054 291
  Mozilla/4.0(compatibleMSIE4.01Windows98)
  http//128.231.164.190/pdq.html

15
What you can get from server logs

• http//www.reportmagic.org/sample/index.html
• http//www.mach5.com/products/analyzer/index.php

16
Some issues in using log data

• Differentiating users from machines or proxies
• Cookies and registration
• Relating IP addresses, user locations, user
  characteristics
• identifying sessions
• Cookies assumptions about nature of sessions
• Measuring hits
• cached pages?
• Interpreting results relative to your goals

17
One source recommends

• Who is visiting your site
• unique visitor identification so you know whether
  a visitor is returning to your site.
• The path visitors take through your pages --
  visitor trails
• knowing each page a visitor viewed and the order,
  you can identify trends in how visitors
  navigation through your pages.
• what element (link, icon) a visitor clicked on
  each page to go to the next page.
• How much time visitors spend on each page
• They say A pattern of lengthy viewing time on a
  page might lead you to deduce the page is very
  interesting or very confusing.
• ButHow do you know what (else) the user is doing?

18
Recommendations, cont.

• Where visitors are leaving your site
• The last page a visitor viewed before leaving
  your site might be a logical place to end the
  visit, or it might be a place where the visitor
  bailed out.
• The success of users experiences at your site
• Purchases transacted, downloads completed, and
  information viewed are concrete indicators of
  tasks accomplished.
• From Tec-Ed, Inc., "Assessing Web Site
  Usability from Server Log Files" on Tec-Ed., Inc.
  Web site http//www.teced.com/c_and_p.htmlWU

19
Another example promises statistics about

• Web server activity
• number of visitors, the number of unique IPs,
  bandwidth used, number of hits they received,
  broken down by Time Increment, Day of the Week,
  and Hour of the Day
• Type of data visitors access on your site
• Web pages viewed, files downloaded, directories
  accessed, images accessed during a time period.
  Broken down by Page Views, Browsing Sequences,
  Downloaded Files, Accessed Directories, Accessed
  Images.
• Referrer information
• Referring Domains and Referring URLs. (Referrers
  are sites with links to your site. )

20
Promises, cont.

• Search engine performance
• the search engines which referred visitors to the
  site, the phrases and keywords visitors searched
  for broken down by Top Search Engines, Keywords,
  and Each Search Engine.
• Visitors' geographic region
• Displays a Most Active Countries graph and a
  table showing which Countries your visitors come
  from.
• Browsers and platforms visitors used
• Errors visitors encountered at the site

21
Promises, cont.

• Advanced visitor filters
• Visitors who accessed specific pages or files.
• Visitors who came from specific referring URLs.
• Day of Week (Example see what happened on a
  specific day) Hour of Day.
  
• Visitors whose first visit is a specific page.
• Visitors' countries or regions.
• Visitors who make purchases on your web site see
  information on visitors who actually buy
  something from your web site.
• Source http//www.123loganalyzer.com/features.htm

22
cookies

• Simulate continuous connection, session
• Identify user
• Store info about user, preferences, past
  activity
• http//www.netscape.com/newsref/std/cookie_spec.ht
  ml

23
Cookies

• the server nytimes.com wishes to set a cookie
  that will be sent to any server in the domain
  nytimes.com
• The name and value of the cookie are nytime-s
• The cookie will persist until Tues April 8
  142504 2003

24
Set-Cookie NAMEVALUE expiresDATEpathPATH
domainDOMAIN_NAME secure

• NAMEVALUE a sequence of characters. The only
  required attribute.
• expiresDATE valid life time of that cookie.
  Once reached, cookie no longer stored or given
  out.
• domainDOMAIN_NAME When searching the cookie
  list for valid cookies, domain attributes of the
  cookie are compared with domain name of host from
  which URL will be fetched. Default is the host
  name of the server which generated the cookie
  response.
• pathPATH the subset of URLs in a domain for
  which the cookie is valid. If not specified, is
  assumed to be the same as the document described
  by the header which contains the cookie.
• Secure Cookie will only be transmitted if the
  communications channel with the host is a secure
  one.

25
Other methods

• Analyses of queries on site search engines
• Emails
• Customer queries and requests for more
  information
• Customer complaints
• Suggestion boxes

26
Analyses

• Frequencies
• Cross tabulations
• Page visited by IP address
• Correlations
• Beware of assumptions about causality
• Graphics
• Exponential distributions