Traffic Tracking Online – who do you trust?

Server logs are the way to go. There's no replacement for a statistics package installed on the server that processes the log files to produce statistics.

Javascript trackers and other on-page stats programs are only as reliable as the technology they use. Google Analytics can only record data when a browser executes the javascript code on the page.

Even when you are directly processing server logs, stats can be misleading. Page views and unique visitors can only be predicted after removing numbers for page bounces and search engine robots.

My newest classifieds site got about 37,000 "hits" last week. 99% of those were GoogleBot downloading all of the 37,000 pages to be indexed.

As for two instances of Google Analytics reporting different numbers, there's got to be an issue with the deployment of the script. It is likely that not all the pages have the tracker code on them.

There's actually a study going on right now comparing some of the most popular stats programs. A May 2007 interim report is available here...
http://www.stonetemple.com/articles/analytics-report-may-2007.shtml

Let me preface this comment by saying I've always been one to educate dealers/IM's that stats are extremely important and one of the biggest benefits of internet marketing. With that said, in most cases I don't think minor variances between vendors really matter. Of course a case like Alex mentions with 1000 uniques from one vendor and 5000 from another WOULD matter, but I'm assuming that's an exaggeration used to illustrate his point. For the most part stats give you a general idea of traffic, trends, site paths, and referrers. There is a focus these days on numbers, and that focus is warranted to an extent. I think having exact numbers would be nice but not necessary. As far as impacting overall data, I'm sure if we could see the activity of visitors who are not counted they would be a representative sample of total visitors, meaning they wouldn't change any trends you may identify.

Now, when you're talking about things like banner ads where you sometimes pay per click, that changes things because differences in statistical reporting means a difference in money spent.

Just my $.02 here, and I'm anxious to see the results/implications of the meta analysis done by stonetemple.com.

I think there are 3 big factors at play:

1) Laziness vs. Complexity = Complexity Wins. People screw up all the time setting up software, and unfortunately tracking codes don't throw big red error messages! No CEL here.... Everyone's guilty of this at one point or another. Unless you have a very small site, there's probably multiple templates, Ajax or Flash content becomes a black hole, broken links, non-canonical urls, etc. etc. these can all skew things. It is hard work to really keep an up-to-date, complete accounting of everything on your site(s).

Corey (first poster above) is right - you gotta have a few methods going on together to audit and keep things reasonably close, but while, yeah, you need to do server stats - his example shows why you also need to do JS stats. It's great to know you're about to have 30k+ pages show up on google (woohoo!!!) but if your boss thinks you just got 30K eyeballs, well, that's bad news for you next week.

2) Terminology , or "lies, damn lies, and statistics". I still run into clients who are counting hits and not pageviews, not counting unique visitors, etc.

The thing is, even those stats are pretty meaningless unless you look at the whole funnel. For instance, I like to show my SEM clients how I do searches.

I enter a queries, right click to open into tabs on 5-10-15 results, whether organic or PPC, and only then start looking at results. I'm my own worst nightmare (but Google loves me!) - all those sites are going to see me hitting in and then spending a minute, 2 minutes, 15 minutes, who knows how long, on the homepage... half the time I'll find what I'm looking for half way through the stack and close out the rest of the tabs.

Honestly, for me that's become a HUGE bad habit. I'll have upwards of 70 tabs open sometimes. I'm a bad example, but that sort of behavior and other things like it has to be skewing #s.

3) Incentives - This, I think is the biggie. Anyone who gets analytics and tracking, and is pushing it, selling it, or buying it has a vested interest, and it's easy to fall into the trap of reading what you want to read. This may or may not cross over to unethical behavior, but it is difficult for people to stay unbiased when there is big money on the line.

I think the whole clickfraud issue got a bit overblown from this very aspect - even these 3rd parties have a vested interest in finding fraud, whether it's real or not. If they don't find any, how are they going to justify continuing operations?



Unfortunately I don't see a really good way around this without major client-side tracking (read as: horrible invasions of privacy) getting a firm foothold. If someone could develop a truly privacy-protecting, double-blind tracking plugin for firefox/IE, etc and license the info on particular sites to trusted stakeholders, then that'd be fantastic... but I don't see that anywhere in the pipeline.

Personally, and in my business, I think the best answer for the near term is regular open communication between client and vendor. We 're pushing hard on paid SEM for dealers, and we're running into a lot of misinformation, bad websites, and bad internal tracking.

From our POV, the only statistic that matters in the end is the bottom line, and we feel we just need to go beyond what might normally be expected to make sure that this traffic is turning into leads for sales staff, and at the end of the day, that staff knows which leads we provided, and that we want to help them as much as possible. I know that sounds sickeningly warm and fuzzy, but it's really just watching out for ourselves. Yeah, someday we might screw up and our numbers will be wrong, or a competitor could come in strong and try to make us look bad, but no matter what happens with that, it's tough to argue with regular visits, checkups and phonecalls.

The stats varying across different tracking programs is very normal. You will see different numbers between server log analysis tools. But at the end all things can be explained.

Why does GA and your banner ad varies... it might be that GA is not implemented on all pages... or it might be that your banner ad is not implemented on all pages.

Why does JavaScript tracking varies from a log file... missing GA code on all your pages... You are actually counting in PDF, Text, or Jpeg files in your log file analysis tool.

Why do log file analysis tools differ... it might be that one tool has a better record for Search Engine robots and one does not or you do not have the same configurations on both i.e filters.

I personally track everything using four different tools... AwStats and Webtrends for Log Files (both are normally off by 100 visitors) ClickTracks or Hitbox and GA for JavaScript versions. They all differ but not by much.

If you are interested in more Analytics and conversion stuff check out LunaMetric http://www.lunametrics.com/blog/ who knows we might run into each other at the next EMetrics Summit.

I think the main problem in our industry concerning analytics is the interpretation and implementation of change based on these results. Unless you are authorized to represent your dealership in their online marketing and positioning, taking this information to the powers that be can prove far more difficult than getting accutely accurate analytics.

IMO