Why you need to analyze page by page, and how to leverage the greatest effect for the least effort.
Analyze page by page
This means you have to think of your site not as a black box, but as a sequence. We can think of it as a conversation. Your site says something to the visitor and the visitor either replies "tell me more" or leaves. Web analytics is about finding out what the site said to whom, and how they reacted. Applying web analytics means changing the site as a result of what you discover.
The number of people who are changing their site in response to what their web metrics are telling them is so low as to be un-measureable. We need to cease designing sites in a way that makes it difficult or costly, or even a skilled job, to change them. Once the site is built, that is not the end of the matter. While most organizations recognize the need to change content, and therefore have content management systems, few of them drive this with structured processes or any form of analysis.
Rather than examine the site's overall conversion ratio, each page needs to be analyzed individually. After all, we don't change the site-- we change individual pages. Obviously we can't look at every page, but luckily we don't need to. Certain pages are critical: entry pages, major exit pages, forms, shopping carts, and a few main content pages that are directly involved in a specific marketing or sales process. These are the pages that need attention. They may vary from month to month, but it's usually pretty obvious. If we launch a new product, and the page listing for that product suddenly becomes our biggest exit page, we know it needs attention.
It's rarely informative to examine a page's overall stats. A page will work differently for different audiences, and averaging this will only obscure what we need to know. For example, a "special offers" page on an airline site may work wonderfully for people searching for "cheap deals" and yet be a disaster for people searching for "first class flights." If the "first class flights" visitors form only a small percentage of the visits, losing them may not even drag the stats down to a noticeable degree.
The best method is to break the audience performance down according to their sources. Sources tell us what people were looking for. People come to every website looking for something specific. They will stay in that site only while they believe it can satisfy their goals. If we know what led them to the site in the first place we understand their ultimate objective. This is easiest if they came from a search engine, because we can see what phrase they used. That's their goal. The most common reason they leave a site is that they ceased to believe the site could satisfy their goal. The next most common reason is that, while they still believed the site could satisfy their goal, the amount of effort involved became too high. For this second group we need to see which page (or pages) they came from within our site before they exited.
Leverage the greatest effect for the least effort
It's easy to be overwhelmed doing this form of analysis-- there's so much data. Even a small commercial site will have several thousand different search phrases generating visitors, and hundreds of different paths. If we think about it, this is similar to the web. There are millions of websites, and billions of links between them.
So how do we manage to avoid being overwhelmed on the web? The answer is simple-- start with a goal. Follow the scent of that goal and ignore everything else. The same thing applies in web analytics.
Before you start a session of web analysis, decide in advance what you want to know, then follow the scent. This means you may have to allow for discovering the unexpected, and some associative thinking. For example, if you've just launched a new PPC ad, what's the bounce rate on its landing page? How many of the people who click through from it make a purchase? If most click through, but few buy, on which page do they drop out? How does the drop-out rate on this page compare with that for people from other ads?
If you're not sure where to start, start at the back of the conversion funnel-- the shopping cart and the credit card submission forms. Improving the performance of these pages ripples back through the entire system, improving the ROI of everything which has gone before. In addition, since these are a small number of pages, you'll get the greatest impact for the least work.
Conclusions
Web analytics are there to be used, not just so we can make pretty charts for our reports. Using them means changing the site in response to the things you discover. A website is composed of many pages, each of which can be analyzed and tuned individually. This means we should not consider the launch of the site as the end of the design process, but as the beginning.
A website should not be seen as a static monolithic object, but rather as a dynamic process composed of many components. The purpose of web analytics is to examine and improve the fit between each component and the visitors who view it.
Brandt Dainow is an independent web analytics consultant. Read full bio.