Counting What Counts
If you want to measure site traffic, you'll have to count what counts. So... What counts? What tools count what counts? What measurements do the tools use? And just why are you doing all this counting in the first place?
If you want to measure site traffic, you'll have to count what counts. So... What counts? What tools count what counts? What measurements do the tools use? And just why are you doing all this counting in the first place?
Today’s column is the one promised earlier about site traffic measurement. As usual, I’ll focus on measurement concepts rather than a lot of detail about implementation. With some of these key concepts in mind, you’ll be better equipped to use whatever tool is appropriate for your Web business.
Tools of the Trade
There are three main ways to measure site traffic:
Units of Measurement
No standard traffic measurement is best for all purposes. Some of the common measurement standards are hits, unique visitors, and unique visits. These all have slightly different meanings.
A “hit” usually means a single downloaded file (each of which occupies a separate line in the server logs). A single page with many images would register many hits.
“Unique visitors” is exactly what it sounds like: the number of distinct human beings who have visited your site in a particular period. Traffic analysis software usually doesn’t measure this directly. Instead the number of unique visitors is estimated by lumping together hits from a single IP address in a short time. This isn’t always accurate for a number of reasons, including the fact that many users can share a single IP address in a short period. If you are willing to cookie users, you can get a relatively close count of unique visitors.
“Unique visits” is a separate counting of an initial visit and of repeat visits from a single visitor.
Matching Means to Ends
If you’ve been reading my columns, you’ll notice a running theme: The most important aspect of any data collection is a clear purpose at the outset. Indiscriminate data hoarding with an unfocused eye toward mining the data at a later date is a recipe for confusion and wasted time. Traffic measurement is no exception: The proper measure of site usage depends on your purpose for measuring the traffic in the first place.
The number of hits is a bad metric in most cases. For example, a facelift that adds a lot of small images to a site would result in a sharp increase in hits but does not necessarily represent any real increase in usage. It might just represent something irrelevant, such as more downloads of pages that are image heavy.
If you’re simply following a trend line to get a rough idea of when traffic increases or decreases, almost any reasonable measure will do. Unique visits and unique visitors are both reasonable. (For sites with low incidences of repeat visits, these two measures should be roughly equivalent.)
Usually a specific business issue will suggest a specific usage statistic. For example, suppose you are interested in your buyer-to-browser conversion ratio. In most cases the appropriate starting point would be the visit. Why? Because each visit represents a potential sale.
On the other hand, a different kind of campaign may suggest a different metric. A provider of subscription-based services may be interested in the sign-up to browser conversion. In this case, since a single user is only allowed to sign up once, the unique visitor is the appropriate starting point.
In my next column, I’ll talk more about conversion ratios, interesting in and of themselves.
Correction: The explanation of confidence intervals in my last article is not technically correct. Before reaching ClickZ, the article was edited by someone with good intentions but without a statistics Ph.D. To correct it here would take up too much space, so if you’re interested, send me an email, and I’ll be happy to explain the details.