Advice for comparing Web traffic numbers

From: Henderson, Jay <jhenderson_at_spss.com>
Date: Thu 22 May 2003 07:26:19 -0500

Jim's comments were spot on. A few comments from the vendor
perspective. We often get asked about comparing the numbers from our
product with numbers that come out of other products. The following are
possible sources of discrepancies that come to light when comparing the
metrics and measurements from two different web analytics products or
services.

o Robot Definition and Filtering - Does the product filter out requests
made by robots? If so, based on what criteria? If one vendor does
something different, or has a different robot list, this would cause a
difference in resulting metrics. Ask them to provide a list of robot
exclusions.

o Resource Definition and Filtering - Different vendors may filter out
different resources for hit counts. For example, does one of the
applications filter out images? If the other vendor does not, or if one
definitions of what images are different, this would cause a difference
as well.

o Page Definition and Filtering - Different vendors may have different
definitions of what resources are considered pages. If the NetGenesis
definitions of what a page view is do not match up, the page view counts
will differ. For example, does the application consider 404 status code
pages to be page views? OR, should a PDF or Flash movie count as a page
view, a hit, or neither?

o Hostname Filtering - Most applications have the ability to filter out
certain hostnames, usually to eliminate internal traffic. Clients
would need to make sure that the way they have configured both products
for hostname filtering is the same as the way they have configured
NetGenesis for hostname filtering.

o Visit algorithm - Specifically, detail the criteria used to define a
new visit and distinguish it from an already occurring visit.

o User algorithm - Specifically, what criteria is being used to define
a unique user? Registered user, Cookie, IP, other?

The above are the points of distinction from which clients should start
to explore. There are lots of "knobs" that one can turn in products, and
unless clients have them turned exactly in accordance with how the other
vendor runs, results will be different. In addition to the
configuration of the two products, there may be certain things one does
(e.g., how new visits are merged with existing visits) that the other
vendor simply cannot do.

Hope this helps give you a few places to start looking.

Cheers,
Jay


--------------------------------------

Jay Henderson
jay_at_spss.com
Senior Product Manager, Web Analytics
SPSS Inc.
http://www.spss.com/netgenesis/
617.665.9320

--------------------------------------




Received on Thu May 22 2003 - 07:26:19 CDT