Dollars & Sense of Web Analytics
In: Columns > The $ & Sense of IT
Published on October 3, 2005
As Web developers, designers and marketers, we are quick to make recommendations to improve Web sites. Yes, we can look at anecdotal evidence and say, “I did this for another client, and boy did the site take off.” What’s missing is concrete evidence that substantiates the claim and excludes other influences. This is where Web analytics comes into play.
Web analytics, whose origins date back to the invention of the Web, has worked its way from the domain of the technically minded to marketers, thanks in part to software with improved user interfaces and easy-to-understand reports. Despite these improvements, there is still a lack of understanding of the technical side of analytics. Web developers and marketing folks need to share the analytic tool, and each group has specific needs that can be fulfilled by the tool.
Let’s look at some of the terminology used in Web analytics reports. While there are many tools on the market that offer varying degrees of functionality and reporting, most advanced programs can compile and process data into comparable reports.
The History of Web Analytics
Web sites record user activity and generate access logs. An access log is merely a flat file that is either space- or comma-delimited, each line in the log file representing a request of a specific file (a hit).
In the beginning, Web sites contained only Web pages. Web pages were composed of only text. A single hit recorded in an access log was a request for a Web page. Faced with huge log files and a desire to know how many pages were requested, some clever person wrote a utility that scanned the log and reported the number of hits. The result was the birth of Web analytics.
Demystifying the Access Log
Something that frequently stumps people is why reports for different sites detail different data. The answer lies in the access log and how it’s configured. Web analytic tools can only report on data that is contained in the access log. If certain data isn’t captured in the log, there is nothing to report.
A basic log file contains the IP address of the user accessing the Web site, the URI stem of the file requested, the time of the request and the status code. For better reporting, you’ll need to enable some additional options: the referrer (the page the user was on when they requested the page), the user agent (this includes the user operating system and browser) and cookie (if your site sets a cookie). If your site operates on multiple URLs, you should also enable the server name reporting.
The below examples use appropriately configured access logs.
Reporting Visits, Visitors and Unique Visitors
Just as hits are misunderstood, so are the concepts of visits, visitors and unique visitors. To understand the difference, we need to look at how analytic tools determine these numbers.
The Unique Visitor
In the early days of the Web, every user on the Internet had a unique IP address. By counting the number of different IP addresses in an access log, you could determine how many unique computers (not people) came to your Web site. Then along came ISPs and proxy servers and all of a sudden hundreds—if not thousands—of users all shared a handful of IP addresses. Merely reporting on IP addresses was no longer logical.
Better analytic tools now use a combination of the IP address, the operating system and user agent to better approximate that unique computer count—or unique visitors. For better accuracy, try setting a persistent cookie. The more information you capture the more accurate this number will be.
Now that we know what a unique visitor is, understanding the difference between that and a visitor is easier. A visitor is an individual unique visitor whose activity is tracked over time. For example, a unique visitor who visits a site five times during one month is still one visitor.
A visit is an uninterrupted session on a Web site of a certain minimum length. Most analytic tools set the period of interruption at 30 minutes. If a visitor goes for a coffee and returns to his desk in less than 30 minutes, then continues browsing the site, he has paid one visit to the site. If he returns to browsing after a 45-minute lunch break, he has visited twice.
Set the time-out appropriately for the nature of your Web site. Thirty minutes might be appropriate for a large content-rich site, but for a smaller site with limited content, 15 minutes may be better. If your site requires users to log in and has a predefined time-out, then you may want to set your visit time-out to the same length.
Pages vs. Hits
As discussed earlier, hits are interesting from a server perspective if you really need to know how many unique requests were made, but not very useful from a true analytics perspective.
To understand traffic patterns, you really need to look at page views. This report will tell you how many pages (HTML, ASP, JSP, PHP, etc.) were served. While a positive trend is a nice thing, merely looking at the number won’t tell you much. And if the site was redesigned to make it easier for users to find information, you’d expect a downward trend. If the company has been in the press, expect an upward trend.
Simply looking a page count doesn’t tell you much. It’s a barometer just like visitor count. For true analytics, you need to find the reason for the change. That might be found by using some of the advanced techniques below, or you might have to investigate outside of the access log. Simply reporting page count and visitors is like telling someone the temperature is 30 degrees. Is it Fahrenheit, Celsius or Kelvin?
Referrers are perhaps the most intriguing parcel of information to extract from an access log. Referrers tell you what page a user was on immediately before reaching your site. However, since pages within a site refer users to other pages within the site, your site will often be your top referring domain.
By analyzing referrers, you can see which domains are sending traffic your direction. By looking at specific page referrers, you can see which pages are driving traffic to you.
If you see a sudden spike in traffic, check your referrers. You may be surprised to find someone new linking to your site. Is this a good thing or a bad thing? That depends on the content of the page that links to you.
Search Engine-Generated Traffic
A further refinement on referrer traffic is traffic referred by search engines. Web analytic tools maintain a list of search engine domains and group and report on these referrers.
A good search engine report is more than the name of the search engine, but also includes the phrases entered into the search engine that were used to find your site.
With some effort, you can also split organic search engine traffic from paid traffic.
Effect of Search Engine Optimization
If you have taken on a search engine optimization (SEO) project to improve your position on search engine results pages (SERPs), then you’ll want to measure your effectiveness. There are tools that make measuring your improvement on SERPs easy, but what you really want to know is if the improvement affected traffic and sales.
All too often, organizations focus on improving ranking for the wrong phrases. Sure, it’s great to be No. 1, but what good does it do if no one clicks on the link? Web analytics won’t tell you if you picked the wrong phrase, but it will tell you if the phrase is driving traffic to your site. If it’s not, look at information presented on the SERP. Perhaps your page needs some tuning to display a better message on the SERP.
Measuring Search Engine and Online Marketing Success
If you’re involved in running SEO, search engine marketing (SEM) or other marketing campaigns to drive traffic to your Web site, it’s critical to be able to identify the segmented traffic to monitor the effectiveness of different efforts.
Many organizations run multiple simultaneous marketing efforts, so it’s not always so easy to point to a peak in Web site traffic and attribute it to specific efforts. Fortunately, there are several easy ways to track marketing efforts.
Tracking SEM and other Web-based traffic is easy. First, when configuring your SEM ads, modify the URL to include a tracking parameter (WebTrends and other tools may provide proprietary tags). For example, instead of using a URL like:
… add a parameter:
Now you can track the activity generated by everyone clicking through that specific ad. Set up a spreadsheet to track campaigns by parameter.
To track activity from offline marketing efforts, create a landing page that has no inbound Web links, then use this URL (something simple to type) in offline advertising, for example:
By subtracting the paid traffic generated by SEM from the total traffic that was referred by a search engines, you can monitor the effectiveness of your SEO efforts. Higher-end analytic tools will break out organic search engine into a separate report.
It’s imperative to be able to separate ads that merely bring visitors to your site from the ones that bring buyers to your site. Web analytic tools can make this job easy, but you’ll need persistent cookies on your site.
The process involves two steps. First, set a URL parameter and persistent cookie so the analytic tool can track the specific user activity, including return visits.
Next, you need to set a specific parameter (provided by the Web analytic tool) that will appear at the end of the URL of the final page in the conversion process, i.e. your “Thank you” or acknowledgment page. When this parameter is present in the URI stem, the analytic tool will look up the parameterized URL (a campaign landing page, if applicable) the visitor entered the site on, and attribute the conversion to that campaign.
By tracking conversion rates by campaign, you’ll be able to do a proper return on investment (ROI) analysis and start targeting those marketing dollars appropriately.
Even if you don’t sell anything on your Web site, you can use this technique to track the effectiveness of your conversion goals, such as downloading a white paper, signing up for a newsletter, or requesting more information. Simply add the tracking code to the final pages of these tasks.
Using Filters for Detailed Reporting
If you’re using a Web analytic tool that doesn’t provide conversion tracking or some of the other advanced features described above, don’t worry. All Web analytic tools allow for the use of filters. Filters allow you to either include or exclude specific types of activity from generated reports.
For example, you may want a filter to exclude reporting traffic from your office or search engine bots. By using filters, you can track activity based on a specific entry page. And by using multiple analysis definitions based on different entry pages, you can compare visitor behavior and conversions for different ads. If you’re running more than two or three campaigns, it may be worth investing in a better Web analytics product.
Using Web Analytics to Identify Design Flaws
Are you ever completely sure you’ve found all broken links or uploaded every graphic for a new page or site? By reviewing the 404 (file not found) report, you can see requested files that don’t exist.
The 404 error may be caused not by the site itself, but by external sites that link to a page that doesn’t exist (for example, a deleted file). The next thing to check is the referring page that called the missing file.
With this information in hand, you can quickly repair your site or ask the owner of the other site to update the link.
Path analysis allows you to see the most common paths through your site. How many visitors come to the home page and go nowhere else? What is the most common second page? Perhaps the most common path doesn’t include your home page. These are all things that can be derived from path analysis.
You can also use a path analysis to see if people are clicking where your want them to on a specific page. Simply search through your path analysis for a specific page and then view the most common next pages. This type of analysis is sometimes called a single path analysis.
Monitor and evaluate scenarios by defining a sequence of pages. Some advanced products will even show you where visitors dropped off. So now you’ll be able to see in a single report how many visitors went from page 1 to page 2 and how many went from page 2 to page 3 within your scenario. You can also see how visitors strayed from the path.
Scenario analysis is ideal for validating a set journey (for example, a checkout process or sign-up form). By analyzing where the drop-off occurs and why (where’d they go instead?), you may be able to improve the usability of specific paths or at least test different scenarios to see which one works the best.
Measuring Impact of Design Changes vs. Other Activity
As we’ve seen, there are numerous types of reports available from Web analytic tools beyond the hits, visitors and page views. By using these other reports and combining them with the use of exclude and include filters, you can isolate and evaluate specific events and their impact on your Web site.
Think about monitoring conversions before and after a Web site redesign. Create a scenario analysis to see if there is a problem with your checkout process.
Web analytics is much more than hits, visits and page views. Web analytic tools are worthy of more attention than a mere once-a-month look for a quick summary. By monitoring and tracking these bits of information frequently, we can make recommendations, prove they worked and accurately report on the true effectiveness of changes.
Alan K’necht operates K’nechtology Inc., a search engine optimization and marketing and web development company. He is also a freelance writer, project manager, and accomplished speaker at conferences throughout the world. When he’s not busy working, he can be found chasing his small children or trying to catch some wind while windsurfing or ice/snow sailing.