Category: statistics

This is the one you have been waiting for… The heads-to-head comparison of Google Analytics with some competitors, 10³ Bees and StatCounter. All three web analytic products, are free at the level I am using them. All of them require their own JavaScript code to be installed in the pages and are totally dependent on that code for measuring any results. If a browser either does not understand JavaScript, or has it disabled, it will not be accurately counted. For this purpose, I have separated these products from the true ‘log analysis’ products I covered yesterday, Urchin & Analog. However, based on the statistics at W3Schools on global browser usage, they find about 94% of users have JavaScript enabled. As their methodology is not described, and they have a disclaimer as to reliablity & local variation, this must be taken into account. Urchin, and other products, has the ability to combine the techniques of log analysis & JavaScript browser sensing. I wish I had access to that level of functionality in my Media Temple account. The Urchin Help pages at Google have a good description on the benefit of using a combined technique.

Well, enough stalling! Here’s the numbers!

Google Analytics

	Visitors	New Visitors (%)	Visits	Pageviews	Avg PV/Visit
Month Total	198		231	463
Average per Day	6.6	80%	7.7	15.4	2.04

10³ Bees

10³ Bees doesn’t give too much detail, but it tries to capture search-terms (as does Goolge Analytics and StatCounter) providing a means of adding them to a ‘To Do’ list. Sound familiar? Think HitTail. Everybody is trying to do it, but I still like HitTail better as it tries to do more analysis on the words by making specific suggestions. More on that in another post! I wish I hadn’t missed their promo deal that ended April 30th, I was planning on taking advantage of it, but got distracted working on the template. Oh well! Hope the rest of you weren’t sleeping!

	Visits
Total	336
Average	11.2

StatCounter

Finally, there is StatCounter. StatCounter gives a lot of detail, like Google Analytics, in a very easy to navigate interface. It’s free if you don’t mind losing the detailed level of analysis that a large log file gets you. Up to 100 log entries is free, and then you can purchase 1,000 or 10,000 or more log entries to allow you to get more detailed analysis. Summary data is always available. StatCounter also confirmed the 94% JavaScript enabled browser figure from above for my site. Though I would really like to know how they figure that when they only get a log entry when it’s enabled… that’s a good trick that I will have to ask them. If it could see the log entries from the web server’s point of view, then it could compare one method vs. the other.

	Page Loads	Uniques	First Time	Returning
Total	355	247	203	44
Average	11.8	8.2	6.8	1.5

Note, StatCounter’s totals are low since I didn’t start them until April 7. I’ll give a full update the end of May. At least we can compare the daily averages.

Analysis

While I would think I could make the following equivalences for terms used, the numbers still don’t line up!

• Google Pageviews to StatCounter Page Loads (15.4 vs. 11.8 average)
• Google Visitors to StatCounter Uniques, and possibly 10³ Bee’s Visits (6.6 vs 8.2 vs 11.2)
• Google New Visitor percent to StatCounter (First Time / Uniques)

However, comparing the numbers really doesn’t support a close correlation here! Only possibly for the new visitor percentage, Google’s 80% vs StatCounter’s 83%. Close. As this is a small amount of data, and there was missing data for the beginning of the month for StatCounter, I will withhold any deep opinion on accuracy just yet. However, the variation certainly makes one think…

I added FeedBurner Site Stats yesterday, and paid for an account with HitsLink (I had used there 30-day free analysis a few years ago, and still had an account setup.) I’ll see how they stack up against a continued comparison of the other contenders for May. As I am remembering, I do like HitsLink. Now, I’m interested in seeing who is more accurate. I will also compare the overall page views to the lo-based analysis in an upcoming summary.

Well, it’s been one month since I started working seriously on developing this site, so it’s time for a little recap of the numbers.

Urchin

First, since I’m running my site at Media Temple, the account includes Urchin for web analytics. (Note, Google purchased Urchin some time ago, and at last report was planning on releasing an upgrade of the product. But, I find that the interface of Google Analytics and Urchin are extremely similar. You can draw your own conclusions.) However, I don’t look at it much as it is limited to being account-wide. I host three different domain names under the same account, and it does not give an easy way to separate out the data. So, these numbers are inclusive of the other domains–which are minimal traffic really, but they do influence the numbers.

Urchin Data from MT
Date Range: 04/01/2007 – 04/30/2007
Total Sessions	3,535.00
Total Pageviews	10,451.00
Total Hits	16,003.00
Total Bytes Transferred	125.96 MB
Average Sessions Per Day	117.83
Average Pageviews Per Day	348.37
Average Hits Per Day	533.43
Average Bytes Transferred Per Day	4.20 MB
Average Pageviews Per Session	2.96
Average Hits Per Session	4.53
Average Bytes Per Session	36.49 KB
Average Length of Session	00:06:39

Analog

Next up is the overall statistics from Analog, which is based on downloaded logs from Media Temple. Analog is the original Open Source product of Dr. Steven Turner, now a principle with ClickTracks, another log analysis product. The logs there are not kept forever, as I like to do so I can do long-term analysis, but only for 10days. I thought I had missed downloading some of the logs since I was missing April 6-10, however, in looking at the Urchin stats, I see the same days show zero hits also. So, something must have been going on with the logging on the site.

I’ve used analog for years on other sites, and it is very configurable to get the numbers you want, however you have to develop the right ‘question’ to ask first. I’ve developed custom configuration files to run for different time periods: daily, weekly, and monthly. I went back and ran the daily reports for every day in April, so I can review the details, but for this report I’m just going to quote from the overall summary of the monthly analysis. I’ve configured Analog to exclude the other domains in my logs, as well as hits from my own browser, so it is explicitly for external access to Chasing A Dream.

Analog
	April 1 – 30	Last 7 days
Successful requests	7,653	2,469
Average successful requests per day	256	352
Successful requests for pages	4,822	1,641
Average successful requests for pages per day	161	234
Failed requests	347	96
Redirected requests	670	226
Distinct files requested	396	251
Distinct hosts served	584	265
Corrupt logfile lines	1
Unwanted logfile entries	16,842
Data transferred (megabytes)	81.07	26.45
Average data transferred per day (megabytes)	2.72	3.78

Making a big assumption here, which is not good to do–we are going to equate the terms from these two reports as follows:

• Urchin’s “Total Pageviews” = Analog’s “Successful requests for pages”
• Urchin’s “Total Hits” = Analog’s “Successful requests”
• Urchin’s “Total Bytes Transferred” = Analog’s “Data transferred”

So the head-to-head comparison is as follows:

	Urchin	Analog
Hits	16,003	7,653
Pageviews	10,451	4,822
Bytes	125.96 MB	81.07 MB

As you can see, it’s hard to analyze these head-to-head when the basic numbers are so out-of-line with each other. In my general review of the details of both reports, I felt that the other sites were not contributing significantly to the difference, but was basing it on the variation in the way the programs did their counting. This may not be true. So, I’m going to pause a moment and run the same Analog report, without making any exclusions, so it should be based on the same data as Urchin. BRB…

Here is the comparison with the revised numbers for Analog:

	Urchin	Analog	Difference %
Hits	16,003	15,458	3.5%
Pageviews	10,451	8,734	19.7%
Bytes	125.96 MB	124.09 MB	1.5%

That’s much better. But, note there are still differences! I’ve added the percentage difference based on the Analog numbers so you can see how much variability there is, particularly in the Pageview analysis. While the definition of ‘hits’ and ‘bytes’ are fairly clear cut–though not necessarily the same to everyone as we see–the definition of what exactly constitutes a ‘pageview’ is still up for debate. At least between Urchin and Analog. Analog at least gives us clear definitions of what it’s terms mean, and Dr. Turner explains exactly what you can & cannot get out of web analytics.

I’ve got more details coming on the other products I’ve been comparing, but if you are itching for more, now, I sugget you check out Avinash Kaushik’s Web Analytics Tools Comparison: A Recommendation.

Yesterday was a busy day working on Chasing a Dream in the background. So busy, I didn’t get to write any posts! I’ll try to make up for that a little today. What I was busy doing was adding some additional web statistic tracking applications and merging old accounts into one at Google Analytics.

After running into one of Darren Rowse’s post at ProBlogger, What Blog Metrics Packages Do You Use? on how he tracks his stats, I decided to try some of his suggestions. It also led to other options too.

I wasn’t interested in SiteMeter, primarily as it required the stats to be public. Once my numbers improve, I’d be happy to share, but I want it to be my choice. See below for more info on my plans to share. I just saw a comment from Michael on Darren’s post telling how you can make SiteMeter stats private… might have to go back and try it! Update: after reviewing the SiteMeter site, I realized this is only for the paid version. We will revisit this in the future.

I’m already using Google Analytics, however I decided to cleanup some of my data last week by putting both my blog.cdchase.com site and my old www.cdchase.com under the same account. I wish I had realized I could have done multiple sites under the same account sooner! It would have saved the need for downloading the data from the old account for archival purposes. Of course, I could go back and login, but it’s not convenient and it’s old data. But, I do want to keep it for long-term analysis. So, I downloaded the data and saved it in an Excel file for archival record and merging & comparison with the new site. For fun, I created a graph of the stats for the period I had which was only since October 2006. (I was going to give you a shot of the graph… but, Excel decided to keep crashing. Maybe later…)

I reviewed the metrics package provided by Media Temple, my host, which is Urchin. If you are not aware, Urchin is the predecessor of Google Analytics, and if you look at both side by side you can see the relationship. However, Google Analytics is much more detailed. So, I’ve decided I can pretty much ignore those stats. Besides, the numbers that I receive from my host are based on the logs which merges all my sites, and Urchin does not give a break down of them by site. On top of that, they seemed really out-of-line with the GA numbers. On the very high side, which would have been nice, but not realistic. I think they are a strictly file-based count, so one hit to a WordPress page gives you all the hits for each include page that is used to construct it. That’s not a realistic way to get numbers today. The JavaScript ‘bug’ that is used for Google and the other systems I installed below are much more reliable. Though you do depend on the user having JavaScript turned on or enabled for the Google Analytics site. (I use the NoScript plug-in for FireFox which disables JavaScript unless I explicitly allow it. I globally allow the Google Analytics servers.) So, those numbers may be on the low side. The other point is how does it track robots since they generally don’t execute javascript… I’ll save that for another post.

While I was there, I did setup my local client to download my raw logs so I can do off-line intensive analysis in the future. This is how I process web logs at the university I work for, using Analog. I have raw logs back over six years! Analog is the creation of Dr. Stephen Turner, the CTO of ClickTracks. Once I achieve a little income from this blog, I will be investing in ClickTracks, which starts at $25/month.

I’ve also added 103Bees… it looks like it will be interesting as well.

Somewhere along the way, I ran into StatCounter and decided to try that too.

I’ll give them all a month to gather some numbers, then give you a recap. I have no intention of keeping them all! I’d go crazy since they don’t always agree with each other. I’ll try to analyze and see which one’s are the most useful as far as a quick overview, and which for a detailed analysis.

As mentioned in a few of the ProBlogger comments, there are also my FeedBurner stats. I’ve been running them from Day 1 too. Subscriber numbers are of extreme importance to a blog when not all visitors actually visit your blog to read it.

As I mentioned above, once I have more significant numbers, I would be willing to share here. I can imagine there might come a time where I might stop doing so for privacy considerations–so my competitors don’t know how good (or bad!) I’m doing. However, I’m generally in favor of transparency, so I will save that evaluation for the future. But, I can say now that I understand other blogger’s choices to both be open with their stats, or not. It’s their choice. And mine. Stay tuned for more details on these packages as the numbers roll in! I will also probably evaluate a few others I was introduced to while reading Darren’s post comments.