Learning how our web site works is an important task. Learning how to properly measure it is even more so. We have to look at not only the data, but we also need to know what questions to ask when looking at the data to put it into the correct perspective.There are a couple of “obvious” measurements mentioned in the last article (visits, page views, and unique visitors). While that information is often raw data, and easy to display, we will now look at some data which most systems will calculate for you. Once people move out of the “visitors” and “page views” mentality, they feel they take a large step forward with the secondary measurements.
The two metrics we’ll look at can easily mislead, or provide “no real information” to the user. I’ve included these early on because people often want to look at them, as they are “easy to understand”, but harder to get information from. We’ll look at them, and look at why they are not as important as one might think.
The secondary obvious measurements
- Pages per Visit, and
- Time on Site.
The Pages per Visit is an easy enough metric. Generally, it is displayed as an average by taking the total number of pages from a site and dividing it by the visits to the site. So if you had 1,000 pages in one day, and 100 visits (remember unique visitors are different than visits) you would have an average of 10 page visits per visitor.
Why this matters: By watching how many pages people are visiting, we can begin to see a couple of things. First, under most circumstances, people will not reach their end goal in a single page, or even two. Therefore, as we see people tracking numerous pages, we know that they are looking for information, and believe that they can find it on our site. Second, if we see the number growing, we might see we have a problem. For example, let’s say it take a minimum of 3 pages to reach an end goal, and that people on average visit 10 pages to compare information on different offerings, may be they look in the wrong area, et cetera. However, if we see that number go up, 15, 20, and 30 pages, we now can start to look to see if the site has an issue where it isn’t clear. With this information, we can start to look at areas more specifically, to know if we need to simplify the pages.
Why this can be misleading: This site is an average, and if the site only has few people one day (holidays, weekends, etc), then we might see artificially high or low numbers in this area. Second, we don’t know how much “comparison shopping” a person may or may not want to do. The user may not be lost, just curious. These are things that just looking at the numbers cannot answer for us.
Time on Site is a metric that can also easily mislead. It is usually displayed as an average of users by adding the total time each user spends on a site from the time they start, to the time they leave. From there, they divide that by number of users, and you have your average…sort of. I’ll explain why the number can mislead in a little bit. First, let’s talk about why it matters.
Why this matters: The theory goes, the longer that someone is on the site, the more likely they will convert into making good on the end-goal. The theory is that you are building a trust with the user. The truth might be your site is slow, and/or confusing – and we should know which one it is. Each site should determine a base line for this metric and see if it rises or falls and know how that equates to the success of the site. For example, a site that has 2 to5 minute long video clips would expect longer time on site, than a site that allows you to download product information sheets as a PDF.
Why this can be misleading: First of, the numbers we have to work with are not accurate, and this is beyond the normal accuracy of the numbers. We only know how long a person is on a page given the time they view from one page to the next. However, if there is no follow up page, we don’t know how long the person was on the last page. It might be very long, or very short, but we always see that as a zero time length, so nothing is added, even if they view a 5-minute video as their last action.
Second, people who only view one or two pages can scale this average in ways not expected. Let’s consider 100 people visiting your site for an average of 1 minute. This means the total amount of time tracked was 100 minutes. However, lets assume that 30 of those people visited only 1 page on your site. This means the 100 minutes should be divided by 70 people, not 100. This gives you a new average time on site of: 1 minute 25 seconds.
Third, if your site is a little slower on any given day, even by only a few percentage points, it will slow down the user viewing the site, and they may either leave early (good to find out why) or spend longer on your site waiting to check out.
And finally, people may be on your site long, if they can’t find what they are looking for, but know that this is where they need to get it. (This directly correlates with too many page views.)
Because of this, I like to look at time on site from a trending point of view, and figure out the cause of the effect. Consider a site where the person is looking for information in the FAQ. The person is directly linked from a search engine, or finds it in just a couple of clicks, and finds the information.
Next we’ll look at information that can start to tell us something.
About Walter Wimberly