Forum Moderators: open
I'm ignoring sectors. I'm ignoring the fact that the links that give you the most Pagerank are likely to get a lot of traffic from Google, since they have high Pagerank.
I'll start off:
PR 7 (main page)
2701 unique ips in a WEEK (all pages)
709 unique ips in the WEEK had a Google referer
1992 unique ips in the WEEK never had a Google referer
To calculate this, I dumped a week of Apache logs into a directory and ran a batch file that doubles as my daily log-analysis batch file. The batch file requires Cygwin, a free collection of unix utilities for Windows. Here are the relevant lines:
path=c:\cygwin\bin
cat acc* > all.txt
echo Raw hits:
wc -l all.txt
echo Unique hits:
cut -d" " -f1 all.txt ¦ sort ¦ uniq -c ¦ wc -l
echo Unique hits with Google referers:
grep -i -E "(Google¦Yahoo)" all.txt ¦ cut -d" " -f1 ¦ sort ¦ uniq -c ¦ wc -l
I set the path to the Cygwin directory in this batch file to avoid getting the DOS version of "sort", which is case-insensitive, since my computer has both DOS sort and Cygwin sort.
Since Pagerank is logarithmic, any mathematical analysis I do will involve taking natural logs of traffic numbers.
Not at all. You could have high page rank with few pages indexed. The more relevant question is how does site design relate to traffic?
IMHO I would spend time wondering why your Google hits are low compared to your non google hits. If your site is well optimised, you should expect at least 55% of your traffic from Google and then a lot from google partners.
Also some google traffic may show with no referrer..... stats annalysis is a mine field of red herrings, such as isp proxy caching etc. I would never take stats too seriously, just a useful guide.
Cannot say I agree with that. Depends on all sorts of factors, like the market sector you are in, how well you are positioned on other search engines, what sort of referral sites are sending you traffic, and so on.
I specialise in tourism, so if I have a new hotel site, will compare the stats for the new site with the log file analysis for other hotel sites I have.
>> stats annalysis is a mine field of red herrings, such as isp proxy caching etc. I would never take stats too seriously, just a useful guide
Spot on. If you look at the Tracking and Logging forum you can pick up a number of good discussions we have had on this subject.
Having said that, log file analysis is relative not absolute, so you can compare trends, and differences between sites in the same market sector.
I knew when I wrote that I was being a bit contentious, however, we have sites covering all commercial sectors except adult, and all the sites get 55%+ traffic from google. Is this because we are not well placed in other engines, or because we are doing well in google? In otherwords, get it right in google and no matter where you rank elsewhere, google will be 55%+ traffic.
"Depends on all sorts of factors, like the market sector you are in, how well you are positioned on other search engines, what sort of referral sites are sending you traffic, and so on. "
I can't think of a market sector except adult where the traffic is elsewhere! Any good links in may be supplying traffic, but I bet 55%+ originated from google. We have experimented with ads in the paper media etc. but all pales into insignificance compared to daily traffic from google.
People used to say yahoo was very important but I always suspected this was because you paid to get a listing, and if your site was not search engine friendly, the traffic would dominate from there. But compared to google it was always a lot lower, unless you got lucky and were top of a good category. These guys must really be feeling the pinch since yahoo started using google results.
The situation now must be that with most of the top engines using google results, the remaining source of traffic is tiny. Personally I would love this to change, but is it not the case?
I would agree with you if you said "new traffic". There are a lot of repeat visitor sites out there. It's something that webmasters should cultivate.
I doubt very much that amazon gets 55% of their visitors from Google.
On Jesse's question,, I have examples from recent experience of a PR5 with 3k unique IPs per week, and a PR8 with a small fraction of that. These, I would say, are exceptions to the norm.
With a large enough and diverse enough sample, I would be amazed if you didn't find a close correlation of the PR of each page against its traffic.
Since Pagerank is logarithmic, any mathematical analysis I do will involve taking natural logs of traffic numbers.
Then I think you would get a straight line trend, but only if you can do this by page rather than by domain.
I think this will depend on site structure. If you get the pr flow wrong you may send too much to a page targeting non competitive phrases. e.g.
Home page pr7 points directly to a page on 'red and pink widgets with blue spots', giving it pr6. No one searches for this so it gets no traffic, but achieves number 1 spot. Meanwhile you have another page on 'sexy widgets' which is three clicks away from your home page and has pr4. It has lots of good content for a VERY popular search term..... gets loads of traffic for all the various search phrases it is picked up on, even if it ranks on the second page..... low pr but high traffic.
I except your comment about having a large sample of data, but many low pr pages deep in a site can pick up as much traffic or more than top level pages..... if the spider is allowed in! People are searching in a more sophisticated way, using longer search phrases. The result of this is pr no longer matters so much, and a low pr page, if you get the on page optimisation right, will rank high for these varied 3 or 4 word phrases.... and these more sophisticated searchers are probably worth more, as they know exactly what they want.
Really I'm just assuming the same Zipf/Pareto type relationship as with server logs. There has been some work on Zipf curves and Web site popularity.
VISITORS is proportional to PAGERANK x NUMBER OF PAGES
I have a site which has PR 7 (fluctuates between very high 6 and very low 7) and I get about 1200 visitors a day.
My site is 100+ pages big.
The average visitor looks at 4 pages of my site.
Add those to your Google results, and then let's see how much Google sends you.
On average my sites get anywhere from 45% - 80% if their traffic from Google.
For instance I have a travel related site ranked #2 for the destination name in Google. I currently get, on average 800 uniques daily.
Here is my breakdown of visitors for the past 17 months or so:
Hits from search engines:
google.com 27766 (50%)
yahoo 15805 (28%)
msn 7736 (14%)
aol.com 2451 (4%)
altavista 1083 (1%)
lycos 200 (0%)
iwon.com 94 (0%)
looksmart.com 31 (0%)
mamma.com 30 (0%)
netscape.com 9 (0%)
alltheweb.com 3 (0%)
go2net.com 2 (0%)
-----
This site is not in the yahoo directory, so any referrers from there are from Google. I also do not pay for any PPC (which would inflate my Altavista score, at least).
This site is #1 in Alltheweb for its primary search term. It is #3 in altavista (after the overture dreck).
It is #1 in MSN (after the looksmart listings)
It is #1 in Mamma.com
It is #1 in ask jeeves
According to my stats and math skills, the site is getting ~ 82% of its referrers from Google and google partners.
Obviously different search engines have slightly different user stats, according to their market share / niche.
YMMV, but Google is sending me my visitors, and the other sites, though ranking me well, aren't sending anyone worth mentioning.
- edit - Forgot to mention I am page rank 6.
My top level page gets the most hits because every page on my site links to it. If someone finds the site and likes it, they are likely to go have a look at the home page it see what the site is all about.
The home page, with a PR6, gets less than twice the hits from SEs than some of my PR3 pages that have the content.
My top search term so far this month, is for a PR2 page. I am in the middle of page 2 in the SERPs with only 5000 results.
Second best term is on page one with 72,000 results and a PR3 page.
Third best keyphrase goes to the home page with PR6. 74,000 results, bottom of the first page.
Fourth goes to PR2. 5,200, top of second page.
fifth goes to PR4. 832,000, top of second page.
There simply is no mathematically direct correlation between any of these number and how often someone will search on a term, or how good your snippet is at getting their attention.
My number one keyphrasse is only number one because two magazines have mentioned the product in their new issues. In March I was on the front page for that keyword and it was my number 51 on my most popular list.
