Web Metrics - Cloudfront.net

Download Report

Transcript Web Metrics - Cloudfront.net

Web Metrics
Sean Fox
Science Education Resource Center
Carleton College
http://serc.carleton.edu
Web Metrics/Web Analytics
• Recording and analyzing direct computercomputer interactions that we hope reflect
website use by real people.
• Incremental improvements on the same old
‘digging through the web logs” we did a
decade ago.
Collecting the Data
• Web server logs
– Easy to get
– Only reflects actual requests to the server
(which often isn’t what we really care about…)
• Javascript ‘Bugs’
– Records page views (window.onLoad events)
– Avoids spiders
– Easier to coordinate with 3rd party services
The easy numbers beg to be
misinterpreted
• Caching.
– Do you know what your headers say?
•
•
•
•
Session Reconstruction and Proxies
Cookies and Javascript turned off ??
It’s really not that bad.
Irrelevant (to us) Use?
– If a visitor figures out they are in the wrong site right away and leaves was
it a successful visit?
– Spiders
– Web services…(what’s a user/visitor?)
There’s still value there
• Know what question you’re trying to ask.
• Integrate web metrics into overall evaluation; ask
complementary questions from multiple
directions.
• Look carefully, keeping all sources of error firmly
in mind. Then look again even more carefully.
• Fancy packages won’t give you answers. It will
take thinking and digging (perhaps in addition to a
fancy package).
Examples/Ideas from SERC
• Is anyone
visiting?
• Where are
they coming
from?
• What are
they looking
for?
5
10 Mar
/0
5
3/
17
/0
5
3/
24
/0
5
3/
31
/0
5
3/
3/
3/
05
24
/0
5
2/
17
/0
2/
10 Feb
/0
5
2/
05
3/
2/
05
7/
1/
2
05
3/
Jan
05
5
/0
0/
1/
2
1/
1
1/
6
Visitors per Day
Visitors to the Tsunami Page
7000
6000
5000
4000
3000
2000
1000
0
Try the same thing for Hurricanes
900
800
Visitors per Day
700
600
500
400
300
200
100
0
8/29/05
9/29/05
Katrina, Rita, and Wilma?
10/29/05
Looking in detail at how users
move through a site
157 visitors entered the site via the
Role Playing Global Warming example page:
/introgeo/roleplaying/examples/globwarm.html
19% continued
to a second page.
2% visited
top page of
Role Playing:
10% visited
various other
pages on the site.
/introgeo/roleplaying 7% visited
a Role Playing
example on a
relate topic:
gwdebate.html
9% visited 0-2
additional pag es
and exited the site.
10% of 157 original visitors
explored the s ite extensively. Each visited
pages outside t he Role Playing section
and saw 5 or more pages on the site.
http://dlib.org/dlib/may05/fox/05fox.html.
The typical paths of visitors moving from
topical example into pedagogic content.
(Data from October and November 2004.)
Questions?
http://serc.carleton.edu