Seemingly contradictory data - can anyone explain this?
There are two (Web Trends) tables that one would expect to contain consistent, if not similar data for search enigne referrals. The tables in question are "Top Referring Sites" and "Top Search Engines" We would expect that if a "Top Search Engine" is in the overall "Top Referring Sites" that each table would show the search engine referring the same number of visitors.
One might expect the "Top Referring Sites" to show a higher number, if the "Top Search Engines" table counts keyword referrals, when a search is performed but not referrals from directory categories, for example in YAHOO! In any case, I would expect the tables to differ in the same direction for a given engine..
When looking at statistics for Google in these tables, we discovered the following inconsistencies:
In January, "Top Referring Sites" showed Google with 7760 referrals.
In January, "Top Search Engines" showed Google with 12170 referrals, what would seem to be a large discrepancy, if it is assumed that these two numbers are measuring the same thing. (I don't think they are)
In February, "Top Referring Sites" showed Google with 10360 referrals
In February "Top Search Engines" showed Google with 2920 referrals, again a large discrepancy, if we assume these two numbers are measuring the same thing. (The discrepancies are in opposite directions from month to month??????)
The site ONLY has a SERP presence, no advertising at all, and directory.google.com is not showing up anywhere in the top 100 "Referring Sites"
I noticed with YAHOO! in the "Top Search Engines" table, the number is higher than in "Top referring sites" presumably because "Top Referring Sites" breaks YAHOO! down into search.yahoo.com - dir.yahoo.com and - google.yahoo.com
I get this monthly report from someone else's web site, they set it up and run it, I don't know how it is configured, but I'm at a total loss to explain this quirky Google data. Anyone have any ideas?
I guess another question is what exactly do those two tables really measure?
Just a guess...since I don't use Webtrends...
Could Referalls indicate users coming from the search engine, while "Top Search Engines" indicates the number of times spidered by said engine? Does Webtrends have a seperate report for spiders?
thanks for the reply..I'mnot real familiar with Web Trends, I'm not sure if spiders are thrown into those tables or not. "Top Search Engines" is followed by a keyphrase detail table and the referrals seem to match up with the plain "Top Search Engines" table.
I suppose if "Top referring Sites" contained spider visits and the "Top Search Engines" didn't, that could explain it.
Is a spider generally assigned a referrer, or a "No referred" as would be someone arriving at the site by just typing in the URL?
I don't use Webtrends much, but as I recall part of the problem with odd "search engine" reports is the definition table that tells WT which referrers are SEs. (Ditto for spiders - you may see an entry in your "browser" list that should be a spider.) Google is often problematic, as depending on your exact syntax, "google.yahoo.com" might show up as a Google referral or a Yahoo referral. And, of course, it might or might not show up as a search engine referral depending on whether it matches a pattern in the SE definition table.
Also the cache feature of google causes problems. each graphic file is a referral in the search engine table. Might this be that since the vistor is singular that the referring site only gets clicked once while the referring search engine gets clicked multiple times?
Skibum, I have found loads of inconsistencies in using web trends as our log analyser - love the way in some of the tables where the % dont add up to a 100% when totalled!
Regards to SE referrals and top referring sites discrepancy, I have seen this as well. Have you updated your Keyword.Inc file?
pete - We were just looking at that the other day, and communicating with someone at WebTrends tech support as well. The keyword.ini file I'm convinced, plays a part, but it sounds like WebTrends has a million things that can be adjusted. It's really going to require some documentation and going under the hood to learn how to get WT to report in a way that is meaningful and reliable.
Could it be that "top referring sites" lists www.google.com, www2.google.com, directory.google.com, and groups.google.com separately while "top search engines" lumps them?
I got another wierd one - I have less impressions then search engine referrals! someone please explain that one!
In order to understand the output of any WT report you have to have access to the configuration: The profile, filtering and reporting etc.
The keyword.ini file can group all trafic from a SE in one tabe - even if it comes from different domains or subdomains.
The definition for each SE look like this
The name is an identifier. The ID's is the domains to track and the KeywordIndicator is the parameter that hold the searched words or phrases.
If you want to seperate the traffic from different parts of an engine then you could do it like this:
Another thing you have to remember is that WT did use to have serious problems with referrers in older versions (I think it was previous to 4.5) so You should use the lates version of the software!
> I got another wierd one - I have less impressions then search engine referrals! someone please explain that one!
I think that has to do with the document definition. If you have documents not listed here that visitors see comming from SE's then you would get this result :)
How bout when WT reports Google Cache hits?
Something like this:
cache:gv8pgfx-24u:www.doctorx.com/serv/invitrofert.htm ivf treatment
Does this mean that someone clicked first on the cached page Google puts in the listing, and then clicked through to the site, with the search phrase being "ivf treatment"?
The cache referrers sometime show up as top referrers, but just not sure if that trailing phrase is the search term used...........
That would mean 60 people found the site on a phrase like "male factor infertility motility morphology subfertile"...which seems like one heck of a search phrase, and more traffic than seemingly more common search phrases.
>Google Cache hits
That is a page request from Google's cache where the user searched for "ivf treatment". (They searched and hit the cache link on the SERP.) Google caches the page, but not the images or external calls, so all of those would be pulled from your server each with the full referrer of: cache:gv8pgfx-24u:www.doctorx.com/serv/invitrofert.htm ivf treatment. So if you have 20 images, the referrer would show 20 times for 1 pageview.
So does that mean in WebTrends, the referring sites, do not just count pages. If Google or any other SE refers someone to an image, as it sorta sounds like the Google cache does, then it is counted as a referral, regardless of whether it is a file for a page, or for some part of a page?
Is there any way to tweak WebTrends to only count specific file types in the referral log, or at least exclude extensions lik gif, jpg, png, and whatever else is not likely to really be a prequest for a web page?
Yes, you can set up a filter. If you go to the edit menu of a profile and make an Exclude filter based on file types, then you can select all image types (or whatever file types you want to filter out). I usually do that as standard since I only want to track pagerequests.