Web Marketing Today

Analyzing Your Web Site Traffic

One of the best ways to improve your Web site marketing is to analyze the existing traffic to your Web site. Until you understand who is coming and why, it’ll be difficult to improve. There are four ways to analyze Web site traffic:

  • Install a counter on your Web page
  • Use your ISP’s statistical package
  • Purchase Web traffic analysis software
  • Employ an outside firm to audit your Web traffic
  • We’ll look briefly at the first three options

Install a counter on your Web page

A counter announces to everyone one who visits, “You are the 107th person to visit Aquatic Aerodrome Ltd.” Of course, you could tweak the odometer so it says, “You are the 100,007th person to visit Aquatic Aerodrome Ltd.” Either way, a counter isn’t very professional. You don’t want people to be embarrassed for you, nor do you want them to think you’re a liar. There are ways to put the counter where only youcan see it, not the world at large, but there are much better ways to count visitors.

Use your ISP’s statistical package

Your Internet Service Provider (ISP) keeps log files which record every single “hit” (request for a Web page or graphic) on your Web site. This is a sample of a raw access log file entry in NCSA combined log file format:

cache.is.co.za - - [18/Oct/1996:02:49:22 -0700] "GET /articles/bannerad.htm HTTP/1.0" 200 9657 "http://www.ca-probate.com/comm_net.htm" "Mozilla/3.0 (Win95; I) via Squid Cache version 1.0.17"

Here you see a Web surfer from a domain in South Africa (that is what the .za stands for), who requested (GET) a Web page (/articles/bannerad/ – in this case, my article “Using Banner Ads to Promote Your Web Site”) on October 18. He was referred to my Web page from a link on a California Estate Planning Web page (http://www.ca-probate.com/comm_net.htm) which had linked to this article. He used a Netscape 3.0 Web browser (“Mozilla” was an early name for what became the Netscape browser) on a Windows 95 platform.

While it doesn’t give you someone’s actual e-mail address (the domain name is as close as you get), the log file tells you a great deal about how people are getting to your Web site, and what they find when they get there.

Your ISP is quite likely to have installed a freeware version of a popular statistical analysis package. There are several very good ones which come with some documentation you may find useful.

  • Analog, written by Stephen Turner, of the University of Cambridge Statistical Laboratory
  • Getstats, written by Kevin Hughes of Enterprise Integration Technologies.
  • wwwstat, written by Roy Fielding of the University of California, Irvine.

Your ISP can program his computer so it will produce a report daily, weekly, monthly, etc. Statistics on your Web site may be posted on a Web page, or perhaps e-mailed to you. Contact your ISP and ask how to access system statistics. If statistics are not available, ask “Why not?” Insist. Business-friendly ISPs offer this service free of charge to their customers; others try to bill you for it, while smiling broadly.

Here’s some of the information that the Analog program reports about my Web site:

Analysed requests from Mon-21-Oct-1996 02:06 to Mon-28-Oct-1996 02:05 (7.0 days).

Total completed requests: 94 617
Average completed requests per day: 14 517
Total failed requests: 947
Total redirected requests: 2 367
Number of distinct files requested: 783
Number of distinct hosts served: 5 870
Number of new hosts served in last 7 days: 5 843
Total data transferred: 522 446 kbytes
Average data transferred per day: 75 635 kbytes

Let’s look more carefully. All in all there were 94 thousand “hits” recorded for that week. But wait a moment: a “hit” is any request for a file or graphic. Since each page has several graphics, this can be misleading. Then again, the visitor may look at several Web pages on a visit. So a much more important statistic is the 5,870 “distinct hosts served,” indicating that I had approximately 5,870 different visitors that week. I say approximately, since if a person came back a second time that week, they would probably be recorded as a different entity, but this gives us a pretty good idea.

Finally, you see that about 75 MB of data is downloaded every day — Web pages as well as graphics. Depending upon how much your ISP charges for “traffic” (the total amount of downloads per day or per month), this figure can have important monetary implications for you.

What’s next? Analog tells me that I get the most visitors around noon and the least about 2 am, that the frequency of visitors is least on Saturday and Sunday, and most on Tuesday. (Not very vital, I suppose.)

I am always fascinated by the Domain Report, though it doesn’t have a lot to do with marketing strategy. The report lists data in four columns: percent of total requests, percent of total bytes downloaded, number of kilobytes downloaded, and number of requests. First off you can see that nearly 20% of the visitors used a numerical domain address (the IP number of their ISP’s computer), and 30% of the domains ended in “dot com”. The rest are from all over the world.

%reqs: %bytes: kbytes: #reqs: domain
------ ------ ------ ----- ------
19.86%: 19.84%: 103654: 18790: [unresolved numerical addresses] 0.06%: 0.05%: 273: 58: .at (Austria)
3.15%: 3.43%: 17908: 2982: .au (Australia)
0.12%: 0.13%: 673: 115: .be (Belgium)
3.16%: 3.42%: 17864: 2988: .ca (Canada)
0.20%: 0.18%: 919: 185: .ch (Switzerland)
0.01%: 0.02%: 81: 12: .cn (China)
0.01%: 0.01%: 51: 8: .co (Colombia)
30.68%: 29.56%: 154423: 29029: .com (Commercial (mainly USA))

Don’t tell me that the Internet doesn’t promote a global marketplace!

The Host Reportis next. Here are the first few, from Austria and Australia:

%reqs: %bytes: kbytes: #reqs: host
------ ------ ------ ----- ----
0.01%: 0.01%: 41: 11: buddy.wu-wien.ac.at
0.02%: 0.01%: 66: 17: www.compass.co.at
0.01%: 0.01%: 63: 11: ibmvie.ibm.co.at
0.02%: 0.02%: 86: 17: junisoft.co.at
0.05%: 0.04%: 228: 51: logic10.allogic.com.au
0.02%: 0.02%: 96: 15: modem001.ts.comcen.com.au
0.02%: 0.01%: 77: 15: modem009.ts.comcen.com.au

With 5,800 “distinct hosts,” this goes on for page after page. Let’s say you wanted to see if your competitor was studying your Web site. You’d look for his domain name (unless he’s using only his IP number to disguise his presence).

A Browser Reportreveals that fully 95% of the visitors to my Web site that week used some version of a Netscape browser. The next most popular browser was Mosaic with half a percent. Microsoft Internet Explorer came in with about a quarter of one percent, followed by the American Online browser. Guess which browsers I take most seriously in designing Web sites.

The Request Report, however, is the most helpful of all. It tells exactly how many requests or “hits” were made for each Web page, cgi program, and image file. For example, here is the report on various articles from my Web site:

0.07%: 0.06%: 323: 64: /articles/
0.70%: 3.23%: 16874: 664: /articles/12design.htm
0.31%: 0.54%: 2827: 298: /articles/attract.htm
0.24%: 0.36%: 1858: 230: /articles/bannerad.htm
0.01%: 0.02%: 81: 12: /articles/branchoffice.htm
0.08%: 0.10%: 540: 79: /articles/compete.htm
0.03%: 0.04%: 217: 29: /articles/domain.htm

My most popular article, “12 Web Page Design Decisions your Organization or Business Will Need to Make,” was requested 664 times that week, while “Billboard or Branch Office?” had only 12 requests. Why? If I knew the answer I might learn how to increase hits on many Web pages. Unfortunately, tools like Analog only allow me to speculate. To get better answers I would need to purchase special software.

Purchase Web Traffic Analysis Software

In the past six months a whole new genre of software has been produced which provides detailed analysis of Web traffic right on your desktop computer. To perform an analysis, you download the access logs from your ISP’s computer via FTP to your own hard disk. (Ask your ISP in what directory your log files can be found.) The analysis software will typically “parse” or interpret the raw log file line by line, and place the various elements into separate fields in a relational database.

Perhaps I’ll find some clues to why “12 Decisions” gets so many hits. I load up the 30-day evaluation version of Intersé Webmarket Focus 2 and begin to play. Whee! All sorts of analyses are available with the ability to filter information for any single Web page, surfer domain name, referral source, etc. I do a traffic analysis for my “12 Decisions” page to see which are the main Web sites which refer surfers to that article.

  • Yahoo’s Web design and layout category, 23% (31% total from Yahoo)
  • Infoseek, 4.5%

Not a complete answer perhaps, but helpful. I also learned that 51% of the people who visit my Web Marketing Info Center come from Infoseek search engine. Interesting. I also learned that American Online subscribers made up the largest single group of visitors to my Web site (5.6%, though nearly all used the Netscape browser), a clue to some of the customers I need to cater to.

While large corporations can spend $5,000 or more for high end solutions, several software packages — with varying degrees of power — may be within reach of the smaller serious Webmarketer:

  • Market Focus 2 (standard edition) from Intersé, $695. An excellentproduct!
  • net.Analysis Desktop 1.1a from net.Genesis, introductory price, $295
  • WebTrends 2.1 from e.g. Software, $299

So why are Web traffic statistics important? you ask, after enduring a rather technical explanation. Just this: they make you a much more savvy marketer by telling you:

  • Which Web pages are most popular, which are least used
  • Who is visiting your Web site
  • Which Web browsers to optimize your Web pages for
  • Which Web search engines are most useful to you, and which are the least useful
  • Which banner ads are bringing the most visitors
  • Where errors or bad links may be occurring in your Web pages

In short, traffic analysis allows you to ask questions which help you fine-tune your Web marketing strategy. Web marketing without statistical analysis is like learning to drive blindfolded. At best, ineffective, at worst, dangerous.


[Note: I answer questions on these matters for my clients only. Please refer questions to your ISP or Web page designer.]

Dr. Ralph F. Wilson
Dr. Ralph F. Wilson
Bio  |  RSS Feed


Email Updates

Sign up to receive our email newsletter
And receive a free ebook
50 Great Local Marketing Ideas