ThinkMetrics' CEO makes the case for distinguishing the unique actions behind these two types of page requests.

There's a very common behavior I see when people visit a site from a search engine. For example, Bob (not his real name) wants to buy an MP3 player. He starts by searching the Uoodle search engine (not its real name) for "buy MP3 player." Uoodle offers three relevant results and a pile of rubbish. Bob comes to my MP3 player site via the home page and then reviews my MP3 product pages for five minutes. He then uses his back button to return via my home page to Uoodle and review the other two sites. About 20 minutes after leaving, he returns to my site, repeats his navigation path and buys. 

This is a pretty typical purchase pattern, but no system in web analytics currently exists that can accurately record this behavior. There used to be, but the standards bodies have lost it. This article is a call to recover that ability.
We've all seen web analytics reports talking about "visit sessions" as if they are one and the same. Both the International Web Analytics Association (WAA) and the Internet Advertising Bureau (IAB) are now using the words "visit" and "session" interchangeably. But they used to be considered two separate things.

A "visit" is defined as a series of page requests with a gap of no more than 30 minutes between each one. This bit about the duration is key -- if there was no limit on the amount of time between page requests, I could return to your site after a week and it would still be the same visit. Why 30 minutes instead of 20 or 45? It's arbitrary, but it seems to work for most people.

Using this definition, Bob's actions constitute just one visit to my site, even though he left and then came back again. This is because the gap between page requests was less than 30 minutes. In other words, even though Bob left the site and visited multiple other sites, it all counts as just one visit because he returned in less than 30 minutes. The fact Bob left the site at all is lost.

It used to be different. The IAB had a separate definition for a session: a series of page requests from the same domain. In other words, if you left the site, the session would be over; under this system, Bob's one visit would be composed of two sessions. IAB dropped this definition because no one was using it, and they thought it was confusing to people. WAA has not formally ruled on this, but current standards being developed intentionally treat visit and session as the same thing, also to avoid confusion.

However, in my experience, this distinction is essential. For example, if I want to improve the conversion ratio of a site, I need to break my shoppers into three groups -- those who buy on their initial visit, those who buy in the space of an hour or two and those who like to think about it overnight (or longer). Each of these types thinks differently, and therefore needs different features in the site.  

First-visit purchasers don't care about you, or what other products are available -- they just want to know the product is good enough to do what they need. If it is, it'll do. These customers typically need features designed to maximize speed -- obvious navigation, scannable product pages and so on. 

People who don't return for days need time to think about their purchase. These are the people who are most sensitive to issues of trust in the company behind the site; the ones who want to know who they're dealing with. They require features focused around remembering them and reminding them about the site.

People like Bob, who are going to do a quick comparison with a few other sites, don't care about your company so much, but they are concerned that the product and/or price is the best for them. They want to quickly compare your offering with a few competitors before making a decision, but they want to get the purchase done before they leave their desk.

In order to maximize the site for any one of these groups, metrics need to be available to separate these users and analyze their behavior. This can't be done unless separate numbers can be put against sessions and visits. Let's look at what happens to Bob's data if I look at his visit and ignore his sessions:

Bob arrives on my home page, goes to my product page, then backs out to the home page. Twenty minutes later, he goes back to the product page. If I ignore the change of session, it just looks like he spends 20 minutes looking at the home page in the middle of his visit. The conclusions I draw will vary from site to site and will be based on what's in the home page, but no matter what they are, they'll be wrong. It's important to me to know that Bob left the site for a while. Not only will I be able to understand that Bob is comparison shopping, comparison of his behavior between the two sessions may be important. Let's assume Bob spends 30 seconds on the home page on the first session, but he only spends five seconds there in the second session. That tells me the navigation is easy to follow.

There are a number of derived metrics that are of potential value once you separate visits and sessions. How many sessions to a visit? This may tell me comparison shopping is an issue. How does the duration of the first session compare with subsequent sessions -- are they shorter or longer? Longer tends to indicate a person's first visit was primarily to identify sites of interest, and that they're thinking more deeply on subsequent visits. If subsequent sessions are shorter, it suggests that they have done most of their thinking on the first visit and are primarily comparing you with other sites between sessions. What is the interplay of visits and sessions in repeat visitors? These are people who come in over days or weeks. Do they also make multiple sessions per visit? Does the number of sessions per visit increase or decrease as they continue to return?

Suppose the visitor enters from different sources on each session -- I've seen behavior where someone will come from a PPC ad, then come back a few minutes later from a search listing. What does that tell me? It could well indicate the branding of the site is not memorable, or that the landing page is badly designed.

Not everyone is analyzing behavior to this level. When you first start to analyze a site you can usually improve performance without going into this much depth. However, it is almost inevitable that at some stage this behavior begins to matter. Whether you analyze the relationship between sessions and visits now or not, it's important you have the distinction available should you ever need it. If people are leaving our site and returning in less than 30 minutes, we need to know. A visit is not a session, and it's important to understand and maintain the difference.

Brandt Dainow is CEO of Think Metrics, creator of the InSite Web reporting system. Read full bio.