| Resources By Subject |
| Technical |
|
|
|
by Bruce Morris
Undoubtedly it's good to know how many people visit your site. Better yet is knowing just what they do when they get there - which pages they linger over, where they came from, which pages they are on when they decide to leave, which pages are the most popular - and site analysis programs can tell you all this and much, much more.
January, 1998
How many hits do you get? For a long time the number of "hits" your Web site received essentially measured how cool or successful a site was. Soon however, people began to realise that "hits" is an imperfect measurement and that it might be nice to know just what people do and where they go when they "hit" a site. Thus site analysis programs were developed and have become so sophisticated that few but the best-funded sites have the resources to interpret even a fraction of data these programs can provide.
The good news is that there are quite good shareware or inexpensive programs available that do a great job of providing more such data than you are likely ever to be able to digest. The bad news comes in several flavours:
Although a wealth of information is provided, the inexpensive programs probably do not provide some particular piece of information your marketing manager wants to know about your site traffic.
- Although almost all of these types of programs can be automated, the better programs can be a real beast to run. They can be slow and so processor intensive that they may need to run on their own dedicated computer - it might not be a good idea to run them on your Web server.
- Once again, they provide so much data that even though it can all be arranged in pretty graphs and bar charts, there is just way too much of it to digest easily.
- Implementing the changes suggested by finally digesting the results, could be a huge project in itself.
- To use the programs, your Web server needs to collect and store site log files. These can get rather huge in a hurry and can easily take up much more server space than the site itself - you may need to add to the amount of storage space you have on your server.
- Even using the best of these programs does not guarantee exactly accurate results. Due to things like caching and proxy servers it is impossible to tell exactly how many people come to a site. Generating an educated 'best guess' is what they do.
But let's stop whinging about it. These programs are great stuff and can, with a bit of work, help make your sites much more efficient and easy to use. They can also serve as a basis for proving the usefulness of a Web site - demonstrating to the suits in your company that the Web site is seen and used by many thousands of greedy customers with real money in their virtual wallets.
What Traffic Analysis Programs Can Do
At the very least, any Webmaster and the suits he or she reports to are going to want to know how many people come to the site. But so much more can be done. We'll make a list of all the data these programs can provide but first let's get ourselves in the right frame of mind so we can be thinking about how to use these programs to make the sites better. As you go through the list of data you can get, be thinking about how comparing one type of data to another can tell you things about your site.
Here's a specific example of what I'm talking about.
For two years I was Global Internet Manager at computer manufacturer Gateway 2000. On our Web site main page we had a little drop down list with a button that said, "GO!" linked straight to the pages where visitors could order a computer. I don't remember exactly why we decided to use the word "GO!" but it worked well and we sold millions of dollars worth of computers every day from our Web site (honest). Well I got the bright idea that if we changed "GO!" to something like "BUY NOW!" our sales would go through the roof and issued the order to try the new wording. I was wrong. Sales fell dramatically. As quickly as possible we switched back to "GO!" but I became determined to figure out exactly the best word, colour, page placement, etc. that would get the best result.
Over the next few days we tried several different wordings, graphics, colours, moved the thing around on the page, etc. and learned that "GO!" worked better than any other word we could think of. Any colour besides black was good as long as it was a bright colour. Generally the upper left corner of the main page was the best place to put the link. I don't think I need to further point out the possible benefits of using site analysis software. Obviously you can use this technique to fine-tune many aspects of a site. With a bit of effort you can find out just how to lay out a page to achieve the result you want.
Please pardon the digression. Here's the list of typical features included in site analysis programs. Look at them with an eye to how you can use them, or a combination of them, to make your site work better or convince the suits and marketing dweebs at your company that the site is worth the multi-million pound budget you are proposing.
Daily, weekly, or average hits, visits or page views on the overall site. Pretty graphs showing activity by day of the week or time of day can be generated. To explain this properly we need to do some definitions:
- Hits or visits - when a site visitor clicks to get a page every separate item on the page generates a hit. So any HTML page counts as a hit along with every graphic on that page. A page with seven graphics would generate eight hits - seven for the graphics and one of the page itself. Obviously, if you're trying to figure out how many people come to your site, this could be misleading.
- Page views - this is a better way to measure site activity. The term "page views" usually refers to the number of HTML pages served - no matter how many graphics or other things may be on the page. If a visitor goes to ten pages, ten page views will be generated.
- Visits - a visit is a trip to the site by one individual in a ""session". If a visitor comes to your site and hangs around for 20 minutes looking at various pages a "visit" is generated. If they come back a bit later another "visit" is generated. Analysis programs calculate a visit by tracking a visitor (based on IP address) and deciding that any activity by that visitor during a pre-determined period of time is a "visit" by that person. You can change the number of visits reported by changing the period of time you use to define the visit. As far as I can tell, there is no agreed industry-standard amount of time to define a visit. Different site analysis programs have different time periods set as the default for making the "visit" calculation.
- Requests for each page of the site. Most programs will prepare a list of the most popular and least popular pages. This lets you see what visitors seem to like the most on your site. I said, "seem" to since pages may get clicked on a lot simply because they are in what I call the "action aisle" on the home page. A page may seem unpopular but actually simply be hard to find - the links to it may be poorly placed. The "action aisle" is the part of the home page that first appears in a browser without having to scroll. Experience has shown that, other things like colour, size, etc. being equal, links in the upper left portion of the main page get clicked on more than anything else simply because of their convenient placement. This is really useful data. You can fiddle with layout on your pages to direct people where you would like them to go or figure out what you have on your site people especially like or dislike. You do need to be careful making such assumptions based solely on this data though.
- Average number of HTML pages per visit. This figure shows how many pages the average visitor goes to before leaving. If this figure is one or two pages you know your site sucks. If you spruce up your content and improve the layout and placement of the most interesting links, this number should go up.
- Average time per visit (usually in seconds). This tells you how much time the average visitor spends on the site.
- Most common referring sites. This tells you where most people come from to get to your site. This tells you who has links to your site and which search engines you are properly listed with. A very useful bit of info.
- Browsers used. This breaks down what type of browsers people tend to use. Site analysis programs can usually be customised to prepare charts and graphs and it is interesting to see how Microsoft Internet Explorer is slowly but steadily gaining ground on Netscape. If you notice many of your visitors are using a particular browser you might make the decision to use that browser to check your pages before publishing them. If you have a site pandering to Mac heads you will probably see many more visitors using a Mac browser than would most sites. You may want to use this data to decide which advanced HTML and plug in features you want to have on your site. If a large number of your visitors are using very old browsers you may want to hold off using advanced features. I actually had someone come to our site a couple of days ago with the first Netscape beta browser - not even 1.0!
- Most popular operating systems. If you know how many of your visitors are UNIX geeks versus Windows or Mac users you can tailor your content and page development to accommodate them - or you might choose not to depending on what you are trying to accomplish.
- Most popular visitor's organisations. This will show you the percentage of com, net, org, etc. visitors.
- Most common countries. This tells you which countries visitors come from. If you see a large portion of your visitors are coming from Lower Slobovia you might choose to translate some content into Slovobian (?). Then again you might not.
- Geography of visitors. Some of the more sophisticated programs can tell you what cities or even zip code of the US your visitors come from. This can be useful. To get really sophisticated you might be able to determine that a large number of your visitors come from the south and start including expressions like "ya'll" and "hound dog" more frequently than you would normally do. Most sites will notice an enormous number of visitors from Northern Virginia. This has nothing to do with whether or not your visitors like fried chicken - AOL has a main hub located in Virginia which many of their users are funnelled through skewing your data. So be careful how you look at all this data - there are some 'gotchas' that may not be so obvious at first.
What Traffic Analysis Software Can't Do
The site analysis programs we are talking about analyse site traffic - not server reliability or site integrity. They will not tell you where you have broken links or missing graphics. For that you need another type of program altogether.
The type of programs covered here do not provide independent auditing of your site traffic. There are services available dedicated to providing independently audited usage statistics to you and your advertisers.
What Features Should You Look For?
This depends on what type of site you have and what you need to know about it and your visitors (also how much you want to spend). Here are a few of the major features you need to be sure you are covering during your selection process.
- Operating system the program runs on. You need to run the software on something. The simple ones can run on your Web server and send out an automatic report. The more advanced programs can easily bog down your Web server if you try to run them on the same system. These types need to run on a separate system after receiving the log files from the Web sever. You could take data from a UNIX server and analyse it on an NT machine if the log file output is the correct format.
- Type of log file output it needs to do its thing - what kind of output does your server software generate? What kind of log file output does the analysis program need?.
- Resolving domain names. If you want to see more advanced data about site visitor activity expect to pay a bit more and tie up a stand-alone computer. The better programs will attempt to "resolve" domain names. This allows you to try to track an individual's behaviour on a site - where they come from, how long they linger and where they go.
- Ease of use. Some of these programs are so quirky they approach being unusable. Most are undergoing constant updating. One quite good program I used had almost weekly updates and patches issued. Reviews I read often are obviously rather shallow - the reviewer seems to have "fiddled around" with the software rather than really giving it a good, real-world workout. Installing these programs, configuring them and giving them really significant, huge log files to crank on is not a trivial matter and most reviewers are not going to go to the effort really needed to check out these programs thoroughly. Sorry I can't help on this one.
GETSTATS
This is a free UNIX program that outputs data in ASCII format - the charts it creates are a series of ########### formed into bar graphs. It is primitive but quite effective. I have it set up on one site to automatically email myself and a couple of strategic suits a daily summary of the site activity. At a glance I can tell what's going on and if I really want to know how my site is being used I look at the more detailed reports generated by other programs.
Microsoft Internet Information Server
Microsoft conveniently purchased the very best analysis program I've found which was published by a company called Intersé. This is good news for those of you whose sites are running on NT - the program is now included with NT's Internet Information Server. This program gives you every imaginable type of data about site visitors and incorporates its own special database that lets you narrow down and count visitors by state and even city.
HitList
This program produces pretty graphs and can resolve domain names. Be aware it is a bitch to run and quite delicate when attempting to crunch large log files. As with most programs that do this much (especially resolving domain names) they can take days or even weeks to crunch data from a few months of log files. They need a very powerful PC to do their thing. HitList is published by Marketwave,601 Union Street, Suite 4601, Seattle, WA 98101 Phone: 1 (800) 521 8176 (+1 206 682 6801).
|
