Welcome to WebmasterWorld Guest from 184.108.40.206
If your have too much time on your hands for the good of your health, or like to hyperanalyse the various spiders/unique referrals or whatever from your weblogs, then you have something like webtr*nds to see how your site is "ticking over".
Since I dont want to be splattering statistics all over WMW, I just wanted to create a thread about looking at statistics and how you should perceive them. Of course, every site is unique but im sure there are some general rules of thumb that you can relate to.
So if youve read every other new post in WMW, and you want some more WMW, then check out this from my web site statistics, and see what you agree/disagree with and can elaborate on.....Im just taking excerpts and suggesting something useful (hopefully). The timeframe used here is last week, pre-good update and during the holidays for many countries across the world
Unique Visitors 6,182
Visitors Who Visited Once 5,263
I look at this and think
1. Good, a respectable % of people come back to see more of the site after their first visit
2. Any change in relative % could indicate a re-indexing of a popular search index bringing in more "uniques"
3. Any change in relative % could also mean people dont think as highly of the site "on average" as they once did
We have talked plenty about these, so I'll skip it
Pages below my home page seem to get the best page views, which seems obvious being in the "middle" of the site. Is it the same for everyone else? The generic keywords on my second tier target the bigger keywords, pulling in the biggest referrals
The "money" pages at the bottom of the navigation structure are content specific and #1's for niche keywords. The results of this are pretty much reflected in the stats
The home page gets 40% of incoming visitors as their first page load. Obviously most inbound links point to the home page, so we can respect the 40% comprises of a respectable amount of incoming links. Also, the most generic keywords of all are targetted on this page (i.e. one word)....so any increase in ranking overall for the site should boost ranking for this keyword.....keep this in mind.
One of my second tier pages....a dynamic page with a nicely sized db behind it pulls in 15% of entry pages...making the total 55%. Take the following into account
1. Any second tier page is more or less exclusive on the home page, therefore is visible to the 40% mentioned above, as well as people from other areas of the site reaching the home page (the other 45% who entered some other way)
2. The PR from the home page and all the "follow through" from PR linking to the second page makes it rank "better" in the likes of Google
3. The flow of traffic between home > money pages and money pages > home
So...in a nutshell, slightly less generic keywords are targetted here, because with less competition, but alias with less impact as your home page, you want to be ranking well for more generic keywords than badly for a more generic one here.
Im sure you get what im pointing at here....its navigation and keywords.....
The "money" keywords and pages....
Brett made a model where you target your keywords according to the navigation structure. The bottom pages of the structure are known as the "money pages", where your site can harvest in all the niche keywords that your site focuses upon. Basically, the word "content" pretty much comes into play here.
In my instance....good money pages get about 1% of your referrals. One page in particular is a "gem" receiving twice as many hits as any other page in the same part of the nav structure.
In light of this, perhaps I should bring that particular page "up one level" and add more content in this area, thus maximising this keyword and "niching" it further.
Anyways, money pages are what your site is all about, dont play games...just target the keywords that suit the pages and in general you get rewarded for the traffic (believe me I aint been doin it long).
TOP EXIT PAGES
This pretty much looks similar to my top entry pages, which can be a bad thing :o). I could assume that people leave as soon as they go, or perhaps en masse they just leave "on average" after X amount of views....whatever. I guess if you want to study this area more closely web logs and better web stats than mine would answer any queries here :)
SINGLE ACCESS PAGES
My site is content specific, but 28% of people access my home page, and dont go any further. I guess this is pretty negative, considering that is over half of people that enter from that page (the 40% mentioned above). However, on the same note, its easy to click on a site by mistake, or click and change your mind etc etc etc
TOP DIRECTORIES (in order of hits)
Anything thats repeated on the site..ie design I put in a specific folder. These are the directories most accessed on my site. Hopefully much of this gets cached on a users first visit, then a super fast site awaits them :)
Below these cached directories are dynamic pages that are 100% uncached, and therefore like munching bandwidth
Afer that, all the unique text and their respective folders bring up the rear (in order of hits)
TOP FILE TYPES ACCESSED
1 gif 31,470 16,760
2 htm 15,350 172,720
3 js 5,800 9,432
4 asp 5,425 226
5 txt 214 114
6 jpg 199 1,609
I use some gifs for design...but apart from that...content is king on my site :o) and therefore the most accessed.
Number of Visits/Number of Visitors/% of Total Unique Visitors
1 visit 5263 85.13%
2 visits 455 7.36%
3 visits 164 2.65%
4 visits 83 1.34%
5 visits 46 0.74%
6 visits 31 0.5%
7 visits 26 0.42%
8 visits 24 0.38%
9 visits 26 0.42%
10 or more 64 1.03%
What I assume here
1. If 1 visit % is high, then im getting lots of new referrals OR people aint interested in the site
2. If 2 visits plus % is high, the site is "sticky"
3. Those that keep coming back might be willing to part with money :o)
SITE ACTIVITY - THE JUICY BIT
Time Interval Hits Page Views Kbytes Visits
Mon 04/01/2002 9,424 3,430 27,049 K 1,056
Tue 04/02/2002 9,754 3,572 30,316 K 1,128
Wed 04/03/2002 10,146 3,720 28,972 K 1,250
Thu 04/04/2002 10,872 3,958 28,881 K 1,108
Fri 04/05/2002 15,489 5,794 38,031 K 1,341
Sat 04/06/2002 9,488 3,752 32,476 K 1,224
Sun 04/07/2002 16,230 8,029 39,140 K 1,570
What can be assumed from this?
1. Since its easter holidays...not much in the long term
2. "cache you later baby"....See how much bandwidth on average people are using
3. Relatively low number of hits because its text based, but a jump in hits out of line can indicate many new visitors (perhaps from a new long term source)
4. The Googly-type update is gonna increase those numbers....(thats just a wish)
ACTIVITY LEVEL BY HOURS
Pretty much when people are awake they will tend to view your site more than when they are asleep :o). From this the only thing I take note of is the busiest hours and reasons why (geographical location and reasons for use)
PAGE VIEW PER VISITOR
Number of Pages Viewed Number of Visits % of Total Visits
0 pages 791 8.81%
1 page 3,680 41.01%
2 pages 1,325 14.76%
3 pages 811 9.03%
4 pages 562 6.26%
5 pages 417 4.64%
6 pages 298 3.32%
7 pages 241 2.68%
8 pages 152 1.69%
9 pages 126 1.4%
10 pages 97 1.08%
11 or more pages 473 5.27%
0 pages? Im stumped by this, im assuming its people who click "back" before they load the page up. Those viewing 1 page either got the info they wanted and sped off/didnt find it OR didnt like the site.
AVERAGE TIME SPENT ON SITE
The longer, the better
The rest of my stats include top browser types, SE referrals, top URL's etc that Id rather just skip, as these areas are pretty much well documented in here. But what about the rest? What about the stuff above.....
I appreciate there are a 100001 reasons for stats being stats...and unique to everyone, but im sure there are some general assumptions that can "reassure" you while you look at them. If you dont find anything useful in stats apart from curing your curiosity, then youre wasting your time :)
I just wanted some feedback/expansion on the things Ive mentioned, and any other method of tracking that you think should be discussed with the people of WMW :)
I mentioned my stats for reader drop-off [webmasterworld.com] in another thread. If I hadn't been watching these numbers, I'd have missed something important. At the very least, I've improved page views (and ad Impressions) with the changes I made.
joined:Sept 20, 2000
For a site that has a conversion event, like sales, memebership, newsletter signups, etc...Top paths would seem to be a very valuable table. Does anyone ahve a good way of extracting meaningful data from it?
Often times the conversion event results in a dymanic page being called as the confirmation page, so it's good to know what page that is in the logs and in WT have the table with top dynamic pages (or whatever it is called) extend at least until that page shows up so that an overall conversion rate for the site can be calculated.
A membership site such as WebMasterWorld can probably get MUCH better information about visitor behavior due to the authentication.
joined:Sept 20, 2000
If so, some people could have a site running in the background for hours, even though they finished browsing ages ago
Just one of the many things to consider at hand ;)
That was exactly what I would have expected, since their sales were basically nowhere. But their stats package wasn't letting them ask the right questions!
You really do need to dig, and if individual stats don't hang together in a coherent picture, find out why.
I also find the browser type very important, have seen in other threads here at wmw that everyone is using 5x and 6x with netscape dying, not the case with my sites.
The mix varies, but I even have NS 2x and 3x still visiting, so as you can imagine, I don't experiment much with the "bleeding edge, you beaut bells and whistles" stuff. (Have to say though, NS even at it's best, is tracking less than 22% total traffic on our sites, seems to be losing out to AOL, who've had a tremendous ad campaign here in Oz for the last 12 months.)
I must say though, when I look at my Session Origins for Week:
Total search engine referred sessions : 14.18%
Total referred session : 65.91%
Total blind (no referrer) sessions : 19.91%
I wonder if I should continue to devote as much time as I do on search engine submissions, or would finally giving into PFI or PPC improve this?
Intersting thread, as always.
Cheers and hooroo
SE Referrals 2,014 (for some reason its been a quiet google week)
No Referrer 3,035
Rest 5401 (must be links then eh!)
Yep Google rules the SE world, but bookmarking and links from other sites without a shadow of a doubt play their part in referrals
joined:Oct 8, 2000
"Actions" that first come to mind could be:
1. lenght of stay
2. visit to your action page, e.g. download, subscribe, buy, whatever.
"Visitor data" can be lots more, just most obvious:
3. Entry page
4. Last page seen
5. individual referrer
6. all search engines
and probably most valuable
Then combine these sets to your needs.
Wouldn't be nice to spot things like, for example:
keyphrase1 coming from search engine1 lands on "action" page more often than same keyphrase1 coming from search engine2, or
specific country tends to leave site after seeing one specific page more than average, or
entry page5 does wonders for browser1 but browser2 users leave right away, or
keyphrase3, that brings n% of traffic, rarely ends up on "action" page...
I'll go to a webpage, e.g. Webmasterworld. I browse the links, and shift-click any link I want to read. I do this until some arbitrary moment when my brain says, "Hey, you have a lot of open windows, why not just start reading the content." Now, I'll start surfing these open windows, closing them one at a time when finished reading them. I may click on external links, and be on five different websites at the same time.
If I do this at night, which I frequently do, and decide to go to bed, I leave the unread or important webpages up overnight. It may be 12+ hours.
What could a developer do to combat this? Maybe eliminate large outliers, which is probably the best option, but what if someone really is surfing the site for 45mins? Maybe a surf time vs. page view factor, which eliminates an outlier which has been on the site for 45mins but only seen two pages. In my example, however, I have many page views and a long length of stay.
Visitors need to be looked at in three groups. New, Returning within 30 Days, and Returning After 30 Days. Any Visitor you haven't seen in over a year should be treated as New.
Bounce-offs are those Visitors that just view one page. They should be classified as New, Returning within 30 Days and Returning after 30 Days.
Some New Bounce-offs probably are lost opportunities while others just wandered in by mistake.
Returning within 30 Days Bounce-offs might be indicating that your content isn't changing fast enough.
I use summary for crunching - but all image, js, css, and any other no-html hit is filtered out. Reduce to the HTML-Hits (or simply ad-place-impressions if your business is selling ads). Or do you sell gif's? ;)
So putting too much importance on the absolute value of a number in any given report is a questionable practice. But looking for % changes from one period to the next is exceptionally useful, especially when bumped up against a timeline of website changes.
I also find it's very important to "clean" or filter the raw logs intelligently to come as close as you can to comparing apples to apples on a clean baseline. For example, for many stats, it's good to eliminate spider hits, or "in-house" website use (filtered out by IP).
At best, a server's access_log gives vague, inaccurate statistics. Still, I choose this method over cookies.
A cookie and database combination can store the date of last visit by IP. This combination I have found works over 99% of the time. Knowing the date of last visit allows you to determine whether the visit is New, Within 30 Days or After 30 Days.
The same tracking software can determine if which pages were viewed by session or what I called a bounce-off.
I think both of the top of the line offerings by WebTrends and Hitbox can get you these stats along with several other tracking services.
I think both trends and changes to underlining stats are more important than the actual numbers. There is a constant error rate that keeps the numbers from being perfect but trends and changes are fairly reliable.
Cacheing problems can be reduced by employing cache-busting schemes. I bet people on this forum can supply information about that.