OK, so what is WebTrends actually *good* for?

Serious question about the metrics you trust in it

[DanieleB]

First post, though I've used this forum as a resource for well over a year now. Please be easy on a n00b if I break protocol somehow. ;)

I've had growing frustration with the lack of proper path tracking in WT Log Analyzer. We finally nagged our sysad into dropping a cookie in the file for us, which he did willingly and well -- only to have it completely ignored by WT in the next log analysis. I'm sure that will come as no surprise to regulars here; searching old threads reveals a consensus opinion that path tracking is by and large a red herring, and we will continue to get single-file "paths" without a disproportionate amount of effort.

OK, fine. I concede defeat. But this now calls into question the validity of the other metrics we're tracking. If we can't get decent path tracking, it follows that anything we're getting from Top Entry/Exit counts and Single Access pages (just for example) is also seriously suspect. I suppose, as someone else remarked, that's why they call it Web"Trends" ... :P

That being the case, I come here to ask those of you veterens who continue to use WT -- even with your tongues firmly planted in your cheeks -- what metrics you *do* consider useful from their log analysis. I presume, for example, that the "most/least accessed pages" is still somewhat trustworthy, at least as a thumbnail sketch. Referrers at least provides a trend overview, though the phrases that come out of there sometimes are beyond understanding ... What others that I should be looking at with a less jaded eye?

[cgrantski]

Did you tell the WT program about the new cookie? Then it won't ignore it. Conceding defeat isn't necessary.

[DanieleB]

Yep, and tried it with several different configurations. None had any effect at all. Just on a lark, here's a sample from our log:

<IP> - - [15/Aug/2005:10:03:49 -0700] "GET /knowhow/template_list/index.html HTTP/1.1" 200 41238 "http://www.example.com/knowhow/kb_contents/organizational/iso9000.html" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1)" SID=4298fdc348df4300caeeb61

SID is, of course, the site tracking cookie (session-based). WT completely ignores it. I've tried every possible configuration. <shrugs>

[edited by: tedster at 1:52 am (utc) on Aug. 21, 2005]
[edit reason] use example.com, not real domain name [/edit]

[cgrantski]

You've put SID (no equals sign) into the session tracking setup screen (I assume since you said "every possible configuration" at least one of those included no equals sign). You've also attached that particular session tracking method for the profile you want to analyze. Right? And you've looked in the Top Visitors report to see if the cookie value is there, and it's not? Not even toward the end of the list? (Often the first part of the Top Visitors list is crowded with 'bots that don't accept cookies.)

If all three of these are true, then I'm wondering about the lack of quotation marks around the cookie. All the Apache logs I opened just now (several different clients) have "" around the cookie field.

[DanieleB]

Yes to all of the above, though the bit about quotation marks is potentially interesting. We poked around for quite a while trying to decide on format, and everything we found was ambiguous on the use of quotes. I'll ask him to give it a try.

The initial question was a serious question, though. It was asked in some frustration, sure, but I genuinely do want to know what you (and others, if they're out there) consider as trustworthy or solidly-reliable metrics from WT.

[cgrantski]

Okay, serious question then. I'd be the wrong person to answer it because I see it as such a complicated topic; i'd fill up pages. Besides, I'm not really sure what you mean by path analysis being all wrong, etc. I put a lot of effort into getting reports to be as accurate as humanly possible. On a complex site this can mean days of research and tuning reports and filters and other things. It's much harder to do that spadework than to set up the reports. Much of what I do involves going through pieces of logs line by line, sessionizing and tracing. This includes logs generated by javascript tagging, before anybody starts a commentary about logs being no good. So I think I have a decentidea of whether WebTrends (and others) really do report properly on what's in the logs. I also do a lot of QA-type browsing of sites followed by checking my click list against what gets into the logs, so I think I have an okay idea of whether logs record properly what was done on the site. I also bother the vendors whenever possible to find out the exact logic and arithmetic behind their stats.

Somebody once gave me a consulting contract to write a paper on the main reasons for inaccurate web traffic reporting, but I got too busy to do it.

My personal bottom line is that there is some degree of error and uncertainty in WebTrends & competitors, and if we understand how much error there is and why, then we can benefit hugely from using the flawed information (compared to using no information at all). I'd also go out on a limb and say that 90% of all reporting setups probably are wrong in some way, but in ways that can be corrected. My third definite statement is that there's no end to all the things that can go wrong.

The current version of WT is so powerful in terms of what it can filter or rearrange or whatever that I feel just fine about using 7.5 provided the homework gets done. I wouldn't use it or ANY system as a substitute for an accounting system, though.

[DanieleB]

Well, we tried it with the quotes. Did another run today, and I'm still getting those stupid one-page paths. I'd accept that there really *are* several one-page paths -- I know for a fact there are -- but when I set the number of elements to 100 and I still get the same 23 paths each time ... well, it's just not working.

At least it did make the cookies show up in the top visitor logs, so that's progress anyway.

We desperately need analysis to be as out-of-the-box as possible. I realize it's never going to be truly off the shelf, but we're a small organization, and to say we're understaffed would be an understatement of British proportions. I wear too many hats to be spending days reviewing log files by hand, and so does our COO-type who does the sysad work for us.

For general things like over all page popularity, we can at least get a reasonable trend from WT, but knowing that the paths aren't tracking properly, I find it hard to believe the Enter and Exit page counts, referrers, etc. I begin to wonder if *anything* other than the general page traffic counts will even be in the ballpark. Which makes me wonder why I'm spending hours and hours of computer time running the reports, and GB of disk space storing them.

But now I'm ranting. So I'll stop and just say thanks for the help.

[cgrantski]

I sticky'd you --- maybe we can get it resolved off line.