Calrissian.com : Techguides : Website Design : Interpreting Web Visitor Statistics

Summary : Website statistics are commonly misunderstood by most webmasters, site visitors, and in particular the media.

Recommendation : A webmaster should be very careful in how they interpret web visitor statistics. Any presentation of data in support of a websites popularity should be in an honest manner- 100,000 hits will usually not equal 100,000 people. All data should be clearly described.

The following guide is based upon usage of Webalizer Version 2.01 , as provided as part of the standard Phpwebhosting service.

A few basic concepts should at least be understood about interpreting website statistics. Note that definitions may vary depending on the type of statistical package that you have on your host server. The following notes apply to phpwebhosting web service, but should also be applicable for most web hosts.

A web page may contain numerous opportunities to register a 'hit'. A page may contain, 3 pictures, a header gif image, a few strap lines. Typically if a visitor has requested a page for the first time, many 'hits' may register, even though they are requesting just one page.

The notion of multiple hits per web page is GREATLY overlooked by people. Just because a site has registered many hits, does not mean it has been visited by a great many people.

-'Visits' : A visit constitutes, where a visitor has requested page/s from the server. There is a time-clock issue to be aware of. The default max. time allowed between requesting pages in a single visit is <30 mins. Note, if the gap between requesting any given page is >30mins, then this constitutes a NEW visit.
So, if a web visitor request a page once every 31 mins four times, over 124 mins, then that would register as 4 separate visits.

-'Pages' : Where a page has been requested. For a page to be registered as downloaded, does not mean ALL the graphics necessarily have to be sent. What matters is that the general 'frame' of the page is sent.

Notes

1. Not every hit will result in the server sending the web visitor data,
a. some pages, graphics, files, will already be in the visitors cache (cache may be browser cache, local ISP cache)
b. 404, page not found errors do not register as hits.

2. Repeat visitors : can be discerned by analysing the diff. between the hits and files totals. The larger the diff. between the two, meaning more of your visitors are requesting pages that they have ALREADY viewed.

So, a big difference in hits and files means your site has more 'regulars' - and regulars are always a good thing (aren't they ?)

Okay, so lets take an example. A classic case is where a new website has just sprung up, and the webmaster has stuck a hit counter on the main index page.

Lets say our site is about banning smoking in public. The anti-smoking site soon attracts media interest, and as part of a news story, the reporter says "www.nosmoke.com has already got 50000 hits in just one week, which shows the immense support which exist in banning smoking in public."

Well, on what basis does the reporter, and indeed the webmaster justify their statement that the site has huge visitor numbers ?
The only thing on the webpage that indicates supposed high visitor numbers is the following....

Screen shot taken March 1'st 2003, from the site www.dont-pay-ntl.co.uk (site now dead)- a typical example of how a webmaster should not be trying to inform us of visitor numbers/site popularity.

As we can see, from 8/2/03 to 1/03/03, this new protest site has a total hit count of 117718. But does this mean 117718 different people have visited ?

NO NO , hell NO !

ALL this figure of 117718 means is this ..... 117718 'hits' for files from the website. 'Files' is the absolute key to web hits (in most cases)

A file could be... a jpg, bmp, or any image file, an add in web component such as a messenger status program.

Typically, most WebPages will contain at least one or two pictures, and maybe a strap line/header. So, overall, for each page that someone downloads to view on their computer, the webserver is 'serving them' a number of files - each of which registers as a 'hit'.

So, each page downloaded will often register as multiple hits.

Making the most of website statistics

In this section, we shall briefly look at what CAN be derived from website statistics.
Might as well use my own web stats, -what better an example could i use ? ;)

Okay, firstly, with reference to figure 1.0, we have just 5 months of stats.
Well, what if anything can be gained from this ?

1. The 'general trend' is upwards in terms of pages, files, and hits

2. The difference in hits/files ratio has changed. There were more regulars as a % of total hits/files in Jan, rather than Dec.

3. Although the number of visitors was sharply up in Feb, the number of 'hits' actually was less*

*The reason for this was due to a re-structured website (done in late January), less gif-hyperlinks - each of which registered as a hit. Overall number of small gifs/jpgs is sharply down, thus accounting for this anomaly.

4. Total data downloaded shows a broad increase over the period.

5. Visit numbers are broadly in sync with total site numbers.

Summary by Month

Month	Daily Avg	Monthly Totals
Hits	Files	Pages	Visits	Sites	KBytes	Visits	Pages	Files	Hits

Feb 2003	552	378	106	53	1464	104395	1506	2981	10602	15472
Jan 2003	523	297	68	26	710	57595	817	2131	9218	16236
Dec 2002	298	198	45	17	467	28272	548	1409	6153	9261
Nov 2002	119	90	32	11	249	40966	346	973	2703	3580
Oct 2002	85	44	19	4	49	3432	81	351	808	1533

Totals	234660	3298	7845	29484	46082

Well, Figure 1.1 gives a simple summary of typical results from a starter website that is less than a year old. Numbers are small across the board, although a discernable trend can be seen.
However, the mean daily page - visit ratio for Feb. is only 2 pages per visit. Clearly, the majority of visitors (inc robots) are not trawling the site across many pages.

Key points

-A trend CAN be assumed from the data.
-The diff. in hits/files represents how many 'regulars' the site has. Robots can indeed also be regular visitors, which further complicates matters.
-The most important numbers are arguably average visit and page total numbers. Also, the page-visit ratio is important to calculate.

Robots - implications for visitor data

Search engines, using automated 'robots' which trawl the net's millions of websites and indexing billions of pages, can really make web visitor statistics very much more harder to analyse.

As the following screenshot shows, visitor data for Mar 2'nd 2003 - for Calrissian.com.

With 111 total pages requested on that day, 54 pages were due to known robots ! Calrissian.com, being a new site has VERY small visitor numbers, and there are times when more up to 75% of all pages requested are not even by real people !
Clearly, young and small web sites will look more bizarre in this way, than the large global net sites.

Typically, i have found that for a web site with less than 100 visitors a day, it is likely that on average 20% will be robots on an average day. In my experience the range can be as low as 5% or as high as around 90%. Naturally the number of robot visitors will depend upon how well search engines have managed to discover that the website exist. It may take a number of months for most mainstream search engines to even catelog the index/home page for a personal/small scale website.

Summary : In the early days of a website, robots may well make up a considerable % of all web visitors.

A good webmaster will want to know who is visiting their site, what pages they visit, etc. However, just reviewing a few raw total numbers like hits and total visits is simply not adequate enough for gaining even a rough understanding. The important thing is the more data the better when forming any level of analysis.
Web robot visits are particularly important to consider, they can easily distort visitor data. Such robots are more of a problem for small scale web sites, where the proportion of robots to real people can be VERY high.

Key points

-Hit counters on web pages are a notoriously unreliable means of forming judgement on the success/popularity of a website.

-Suggestions : Hit counters should rarely if ever be used on webpages.
If web visitor data is presented on a website, the data should be presented at least in a fair manner with some background/history info.

-Web robots must be considered when forming any appreciation of web visitor data.