Forum Moderators: DixonJones
I'm working our user tracking system into a website right now - this will be a fairly complicated set-up.
Using url-based methods to track where the click came from would be quite useful, and I would guess more reliable than using referer data.
For example, an url to page x from link y may read www.mydomain.com/x/y/, while the same destination from link z may read www.mydomain.com/x/z/.
However this could be abused, and part of me thinks 'keep it packed up in the server side'.
Anyone have any thoughts/experience on this?
For the above, I was thinking about 'cloaking' the extra 'url bits' from bots and other bot-like creatures.
Basically have a variable turned on or off depending on if sessions were on our off (ie, if yer a bot or not!).
On our latest project, we're working with the clients in the same way one may create an adsense campaign - building action phrases etc that will be used on the site and evaluated with the tracking...
What you do lose though, (unless you collect it another way) is the original referrer all the way through to the completion page - so... 20 people from source a typing in "widget" and 20 from source b typing in "Gillets" and 20 of these buy... but which 20 were they?
Neat idea though...
Sounds an interesting idea - and as for abuse I think it is so unique that it is unlikely anyone would try to abuse it...
Thats' what I thought too, but I've been wonrg before ;-]
What you do lose though, (unless you collect it another way) is the original referrer all the way through to the completion page - so... 20 people from source a typing in "widget" and 20 from source b typing in "Gillets" and 20 of these buy... but which 20 were they?
Let me be more specific - I'm talking about within site stuff. This particular client has links all over their site that lead to a feedback form - the form adapts it's message depending on where they came from.
The form is submitted to their call center and from that point the website has finished it's work. This is how they generate leads with their site.
So, the site has these 'action links' all over the place, in all sorts of pages and content. There are 2-3-4 different variations in wording etc. for each of these action links. We will be trying the different styles and collecting data on click-thrus, and what I wanted to do was to find an easy way keep track of the different styles so that we can use the most successful variation.
We all know that referer data can be unreliable, so rather than "if referer = x then store this data here" I'll have them land on the 'feedback' page using /x/y/ where 'x' denotes the message to use and /y/ is the link identifier. Some form of data collection occurs server side, depending on the switch (/y/) that it sees.
Anyway, I'd definately like to hear your thoughts on this, though now I have it implemented and know that it works okay. I just didn't want some turkey coming along and 'monkeying' with the switches. I suppose there is no way around this...
A down side is from a search engin point of view: All links into the page now look very different to a search engine, so may be indexed differently or the page may be seen as duplicate content and that page dropped from the engine's index, but since these are links to the conversion pages, not the "marketing" pages I guess that is not so relevent.
Your only real danger now is having too much data, not enough information. Depends on how you slice and dice the database I guess.
Bravo for taking a step forward.
iisrewrite
I guess that's like mod_rewrite? mod-rewrite is one of my best friends on this project!
A down side is from a search engin point of view
I would check (am checking) the user agent and then apply(ing) the link ID (/y/) if it's not-a-bot.
Though you are right that they are conversion pages, they do have descriptive text in them that I want the engines to spider; you never know what someone will put in that search-bar.
Your only real danger now is having too much data, not enough information. Depends on how you slice and dice the database I guess.
We're looking at it as a method for evaluating site copy, what works and what works better. What we need to be careful of is not changing too many variables on each page.
The sucecss of a 'module' (each individual link and it's text is a module) will need to be weighted against other page factors:
The page has all I can give it wrt search engine optimization - it's clean, valid, and semantically I did the best I could. I've also leveraged the data here and there...
So it should get traffic, and with this 'tool' we can evaluate how successful the site is once the user is there. Or so I hope.
I can't wait to see this running.
I guess I should have added that the data will be stored in a db - a whole other story really. Actually, the db info is the basis of the user-tracking side of things while the GET data or tags are used for click tracking.
It may seem a bit complex, but it's for a purpose, really.
I see two ways that I can do what I am trying to do. One is by using referer data, to see when who clicked from this page to that page. I could say "if referer = x then record action z".
This method has problems though. Referer data is unreliable, and what if there is >1 module/add on a page pointing to the same target page?
- mipapage
