Redesign of Python site

[Tim Parkin]

>
>>From the most recent stats, the top 20 entry pages:
>
>      Hits            Visits          URL
>1     252455  3.60%   164578  44.76%  /
>2     169489  2.42%   20381   5.54%   /pypi
>3     66001   0.94%   7997    2.18%   /2.3/
>4     47542   0.68%   7664    2.08%   /doc/
>5     25401   0.36%   7049    1.92%   /doc/current/tut/tut.html
>6     39856   0.57%   5825    1.58%   /download/
>7     10903   0.16%   2826    0.77%   /cgi-bin/faqw-mm.py
>8     20848   0.30%   2678    0.73%   /doc/current/lib/lib.html
>9     10952   0.16%   2428    0.66%   /doc/Newbies.html
>10    8372    0.12%   2347    0.64%   /doc/Intros.html
>11    13328   0.19%   1971    0.54%   /doc/current/
>12    7992    0.11%   1616    0.44%   /2.3/highlights.html
>13    2537    0.04%   1490    0.41%   /~guido/
>14    67565   0.96%   1430    0.39%   /doc/current/icons/
>15    4802    0.07%   1324    0.36%   /doc/Comparisons.html
>16    22751   0.32%   1251    0.34%   /topics/learn/
>17    3631    0.05%   1199    0.33%   /doc/FAQ.html
>18    53884   0.77%   1198    0.33%   /cgi-bin/faqw.py
>19    7969    0.11%   1152    0.31%   /doc/current/tut/node7.html
>20    9290    0.13%   1144    0.31%   /doc/current/tut/node6.html

Although essential for working out a new information architecture /
navigational network for a site, stats can be a very misleading way of
working out what are the most used areas of the site. 

Problem #1 : A prominent, badly worded link will get many people
clicking through to it but a large number of these may not find what
they expect. This shows that logs record where people go and not where
people want to end up.

Solution #1: Take into account page visit duration, remove hits where
another unrelated menu item is clicked in a reasonable amount of time,
etc.

Problem #2: There is little record of people who haven't found what they
wanted, obviously an important sector to target. 

Solution #2: Look for people who have left the site without lingering on
any page for a significant amount of time. Trace back there paths and
try to work out why the links they have used 'could' be confusing.

Problem #3: People who know what they are after go straight there and
don't record significant numbers of hits. People who get lost or
confused record a large number of hits. This skews the results away from
your successes to your failures.

Solution #3: Look for short paths within the logs to indicate that
people have found what they want quickly, rank these separately from
other rankings.

There are a lot of other gotchas related to log analysis to bear in
mind. An example of one of these on the current site layout may be the
fact that the search engine sits on another domain so that when someone
searches they leave the site and come back in on a deep link. This
probably accounts for the high number of deep links recorded on the
entry pages, a lot of these results aren't in fact entry pages at all.
This could account for the high number of people leaving from the home
page.

Tim