Are "unique visitors" really the goal?

[jk321]

In trying to figure out exactly what a visitor is worth to my site on the ppc's, I've been doing some thinking about the "unique visitors" measurement, which seems to be the standard that ratings companies have deemed most desirable.

But, wouldn't the opposite be true? Wouldn't the "non-unique visitor" be most desirable? For example, say your site was an online magazine that generally equated to a print version of the same magazine. Would you really want a unique visitor --one who picked up your magazine one time and then never looked at it again during the measurement period -- or would you rather have a non-unique visitor -- one who picks up your magazine everyday and thumbs through it repeatedly?

Am I looking at this the wrong way?

Thanks,
jk321

[redzone]

jk321:
You are correct in wanting the "repeat" visitor. But you want the repeat visits to come from the visitor:

1. Bookmarking your site
2. Signing up for a contest, sweepstakes
3. Signing up for an email digest, ezine...

You don't want that same visitor repeatedly coming to your site everyday from a click through in a PPC engine...

[chiyo]

important to make a distinction beyween uniqiue and repeat visitors - as suggested.

Teh advantage of counting uniqiue visitors is that it reduces inflated estimates of traffic due to spiders, diffs between sites with lots of gifs etc where hits are far more than actual visitors, and various other problems with hits.

As far as PPC is concerned, I agree with the last comment. With PPC bids on the way up, you only really want PPC's to catch new customers, not to have to pay PPC for people who already know you! People who have already visited your page you haope to have caught already by them bookmarking, signing up for a list, or emailing you. And hope they never have to click on you from a PPC again...

[tedster]

>important to make a distinction beyween uniqiue and repeat visitors

It's also important to be clear about your definition of unique visitor -- and it doesn't really mean the same thing as we intuitively mean by "new visitor". There isn't one standard definition -- this is fuzzy territory.

Over time different visitors will be coming from the same IP address. On the other hand, because of proxy servers, one visitor will also register as several different IP addresses on the same visit.

There are a few different algorithms for counting uniques and they make different assumptions. For instance, FastStats uses something called the I/PRO algorithm. This algo assumes that after 30 minutes of no hits from a given IP, a hit from that previously requesting IP is now another unique visitor -- and FastStats adds it to the count. So, this algo counts a next day return by the same person as a unique.

Cookies will help you determine truly "new" visitors, but even then, cookies can be turned off or deleted between visits. Complicated "ballpark answers" are the best we can do to answer what seem like simple questions.

Edited by: tedster

Edited by: tedster

[chiyo]

yes agree tedster.. your "ball park figure" is probably the best descriptor of what we can hope to get for measuring new visitors.

[seoboy]

>Over time different visitors will be coming from the
>same IP address.

more and more ISPs use dynamic IPs, whereby each visitor is assigned an IP from an IP pool at the time they "log on" to the internet. upon logging off the IP they were assigned is now freed up, and returns to the pool so it can be reassigned.

this is compared to static IPs, wherein a user keeps the same IP address "permanently".

>On the other hand, because of proxy
>servers, one visitor will also register as several
>different IP addresses on the same visit.

not so sure about this. my understanding is that it works the opposite way - proxies allow many users to look like an single user, because they let many users share the same IP address at the *same* time). AOL is a prime culprit here, which unfortunately makes tracking AOL traffic by the time-based algo tedster mentions flawed. but its an ok solution in an imperfect world.

i think its good think of a distinction between "unique visitors and "unique visits". you really get closer to the former with cookies, while the time algo is better at judging the latter. unfortunately, there's no perfect solution. but then again, its a hell of a lot more accurate than Nielson or Media Metrix or any of those micro sample, macro extrapolation measurement techniques.

seoboy

[tedster]

tedster:
>>On the other hand, because of proxy servers, one visitor will also register as several different IP addresses on the same visit.<<

seoboy
>not so sure about this. my understanding is that it works the opposite way - proxies allow many users to look like an single user<

Both are true. Here's how it works with AOL, for instance.
One visitor requests a page. Every element on that page gets a hit. AOL uses 7 or 8 different proxy servers for that one total page -- that is, server1 asks for the HTML, server2 asks for this graphic, server3 asks for a different graphic and so on. This has the effect of inflating visit counts.

But with high traffic, what you said comes into play. Several different visitors at once may be using the same proxy server, the same IP address. Then many different visitors look like one. However, each one of these several visitors is still hitting the site through 7 or 8 proxy servers, so it becomes impossible to untangle.

In short, with AOL, the more traffic you get, the more the "uniques" number is understated. With less traffic, the tendency is toward inflation. And I have no idea why they do things this way at AOL and others -- Deutsche Telekomm for instance. Maybe just because they can.