Since my last posting about unique visitors and visits, I’ve received a great deal of feedback. Much of this feedback praised my blog, with one reader saying it was “a voice of reason” in an otherwise “ill informed web analytics industry”. That may be a bit strong, but I love it nonetheless!

As might be expected, my post was also met with some pushback – which is also great. I’d hate to post something everyone agreed with…what’s the point of that? Now these folks had some questions and indeed some misunderstandings that I wanted to clear up.

So the theme of this posting is simply that all unique visitors are not created equal.

To some of you, this may be common sense. To others, this is new information. And to others, this is purely heretical. “How dare you question my unique visitors?! Such blasphemy!”

I know, I know…I’m always rocking the boat. But the web analytics industry is so young and clouded by misperceptions that someone needs to step in and put a stake in the ground. And why not me? So here it goes…

Rather than having an emotional debate about the accuracy of unique visitors and whether you should or shouldn’t use them, let’s start with some basic observations:

  • Fact #1: Most web analytics systems rely upon cookies to identify unique visitors.
  • Fact #2: Cookies are set at the browser-level by domains like
  • Fact #3: If cookies are not accepted or used by a domain, web analytics systems mostly rely on IP+user agent combinations and/or universal visitor identifiers (i.e. a registered user ID)

Now assuming you agree with those observations, let’s look at 15 scenarios that impact the accuracy of your unique visitors counts:

  • Scenario 1: One person uses multiple browsers from one computer from one location
  • Scenario 2: One person uses multiple browsers from multiple computers at multiple locations
  • Scenario 3: Several people use one computer with one browser from one location (i.e. a typical 4 person household)
  • Scenario 4: Several people use one computer with several browsers from one location (i.e. a typical 4 person household)
  • Scenario 5: Several people use several computers with several browsers from one location (i.e. an increasing trend among households)
  • Scenario 6: One person deletes cookies for one browser
  • Scenario 7: One person deletes cookies for multiple browsers
  • Scenario 8: Your web analytics package reports both cookied and non-cookied users in the unique visitor counts
  • Scenario 9: Your web analytics package doesn’t use cookies or registered IDs to identify unique visitors
  • Scenario 10: One person rejects cookies on one browser
  • Scenario 11: One person rejects cookies on one browser, but accept cookies on another (for example, Internet Explorer vs. Firefox)
  • Scenario 12: Many people from one computer reject cookies
  • Scenario 13: One person rejects cookies from one computer but accepts cookies on another computer
  • Scenario 14: IP pooling by major ISPs like AOL means many IP addresses for one visitor as they move from page to page on your website
  • Scenario 15: Dedicated Corporate IP addresses and standardized browser configurations mean one IP address for many people

At any given point in time, all of these scenarios are playing out on your website and are inherent in your unique visitor counts. Whether you like it or not. And just for the record – these are not vendor specific issues as one of my readers suggested. Hopefully you can see that every scenario I’ve listed above has little to do with your web analytics platform. Rather, they are byproducts of web measurement in general.

So what is your definition of a unique visitor? How many “uniques” do you really have? More importantly, how many “people” or prospective customers are you really reaching?

If you get 1 million unique visitors in a given month, is that 1 million opportunities to sell? Or is it just 500,000? Maybe Scenario 3-5 are really heavy in your site traffic mix, and you have 1.5 million unique “prospects” that you could convert.

No matter what the case, given all 10 scenarios above, what number would you feel confident reporting back to your CEO? Can you have confidence in a unique visitor-based conversion rate? Or even a unique visitor count? Some of the scenarios I’ve listed above will inflate your unique visitor counts and some of the scenarios will decrease it. Some will actually do both! And it’s nearly impossible to measure the net impact of all these factors.

For these reasons, as I suggested in my first post on Visits and Unique Visitors, I almost always use Visits as a strategic measure of how well my sites are converting prospects. If nothing else, each Visit represents an opportunity to convert a prospective customer. It’s no more complicated than that. And because visits (or sessions) are generally measured based on cookied-visitors only, they are much more accurate than unique visitors (visits are in effect a subset of unique visitors).

Visits, by definition, also do not require you to determine “uniqueness”. In other words, all Visits are created equal.

Of course, relying on visits has drawbacks in marketing analysis, segmentation, latent conversion, lifetime value, etc, etc. But that’s OK because by the time you get to these drawbacks, you shouldn’t be focusing on unique visitors anyway – you should be focusing on unique customers.

So what is a unique customer? Why is it better than unique visitors? And why won’t it suffer from the same inaccuracy issues? As much as I’d love to dive into that now, you’ll have to stay tuned until my next post – I’m all out of time today. In the meantime I welcome your feedback and as always, if you’d like assistance understanding how to leverage web analytics to maximize ROI, please do not hesitate to contact me and the Omniture Best Practices Group.


I do actually, although as you might imagine I can’t mention their names. I also did a similar analysis myself when I was a customer of analytics many years ago. This was well before the unique visitor/cookie debate, I was actually more interested in understanding multi-channel behavior. In either case, my exposure to these initiatives has confirmed one thing – variances in the accuracy of unique will differ by website. There really isn’t a hard and fast number you can use to estimate it. Quite frankly I’ve seen numbers in the low single digits, all the way up to high double-digits. All of these are based on a registration-based analysis of uniques. If you don’t require registration on your site, look to some of your peers in Media or eCommerce, as registration is quite pervasive there.


Hi Matt. Do you know of any companies who require registration that have compared their Omniture uniques against their more accurate registration logs to measure how off the former numbers might be? Thanks