Forum Moderators: DixonJones
Say my logs are showing 2,000 "distinct hosts served" in the past 7 days. But how many of that is genuinely unique people?
The pitfalls I'm aware of are dynamic IP's -
1) ISP could give visitors a different IP address each time they log in, thus 1 person visiting every day for 7 days will count as 7 unique visitors.
2) Proxy servers. The log files will count hits from the Proxy and not the unique client. Therefore, if 30 visitors are all using NTL ISP in Townsville then they will count as 1 unique visitor.
So how do you work it out? How do you accurately determine how many are visiting each week?
You won't be able to count those who delete cookies/don't accept cookies very accurately, but it should bring down the number of dialup users who come back with a new IP every time....
(edit)Another trick is to require registration, but that can deter people from visiting your site all together(/edit)
I was just wondering if there is an industry standard formula
e.g. unique hosts / x + y or something.
The real problem with proxy servers (like AOL) is that they serve up a different IP address for nearly every page requested. So instead of 30 visitors looking like one (as in your example) it is more likely that one visitor will look like 30. When I started tracking visitor sessions using an on-page script I was really surprised to see that my logs were far from accurate.
I'm using the AXS visitor tracking system which is great for looking through sessions with a fine tooth comb, but not so great for viewing trends and overall stats.
We're about to implement cookies for this on one site. I also just got a new site with West Host so I can try (or so my code guy can try ;)) using PHP and MySQL to track and log visitors. So while I can speak at length about what does not work, I'm still experimenting to see what does work :)
We may all have them enabled. But I'm not sentimentally attached to the little critters. I have Opera set to "Throw away new cookies on exit".
If I visited Frank's site on separate days, a naive cookie counter would see me as different people.
e.g. if 25% of the population are using AOL, 40% using other dynamic IP, then surely there is a formula which states
best guest unique visitors per week = unique hosts / x + y * n%
where x and y are compensation factors for the above reasons.
Surely, there is a simple(ish) formula for this?
I don't know!
But Frank is angling for a formula. It'll need to take into account:
** awkward people like me;
** those who regularly clear their "temporary internet files" when using IE -- this can also delete cookies;
** people who have cookie management software that purges cookies;
** etc.
It may only be a few percent worldwide, and even less in the actual audience for Frank's site, but it is a factor to consider even if it it doesn't make it into the final formula.
Yes, I help people with their computers all the time, and it is frightening the amount of trashware people download onto their machines. Based on a biased sample, I reckon I'm about the only sane surfer left :)
I'd say that people who change the defaults are 5% or less, sorry but we have to cope with that. I mean browser vendors are complete idiots when it comes to setting up default values.
Cookies are a lot better than IPs, but you'd get better results with visits rather than visitors.
I think the most important part of any web stat is the trend. If you measure the same number the same way week after week, month after month, and bump that up against changes you know were made on the site, then you have value.
If your site offers a product or service, a million (accurate) unique visitors is poor if producing 1 sale.
100 unique visitors where 50 brought something is substantially better or the same returning visitor is making multiple purchases.
Tedster once indicated a client got 45 visitor/month where 6 - 8 turned into sales (relatively large contracts) which is exceptional.
Accurate unique visit counting has only one real value -- if selling media space or traffic re-directing but in this case sessions are still (generally) more important unique visitors.
On a large scale (example -- Amazon.com) returning visitors out number new uniques -- and keeping returning customers who know and trust you is more important and profitable than new ones.
Is it worth knowing if multiple user on the same IP or proxy, the same user on a different browsers, a public computer with mulitple users finding you, an office computer that can never buy or home computer with muliple buyers (mom, dad or young adults).
Tracking uniqueness is less important than what they do -- how long they stay, paths taken, and the outcome per session. :)
I have a lot of free stuff on my site and I think that it is too much.
I know that I have to:
a) cut back on the freebies
b) use more enticing words to direct the punters to the till.
By counting the number of freeloaders each week I can check if the changes I'm doing are working.
The previous poster is right in that it is best to compare trends. i.e. each week get the customer to visitor ratio. But then surely this is hard to do as some customers will signup on impulse, some will scan the site, bookmark, and call back a few weeks later and decide.
Man, this trading on the internet aint easy.
It seems to me that what you are really interested in is your conversion ratio. How many unique visitors come to the site every day vs. how many purchases.
As your site becomes more popular, it is inevitable that the total number of freeloaders will increase. But if you can determine a conversion ratio, you can view the trend over time (great point, Tedster) to see if your percentage of unique visitors who purchase is increasing.
It may also be just as important to learn where the people who buy came from. Not all visitors (or referrals) seem to be equal.
But I think no matter what you do, understanding how many unique visitors you have today and what they buy is important. If anyone has a better suggestion for tracking that than cookies and some sort of database (MySQL?), I am all ears - I'm looking to implement a more sophisticated tracking system myself.
