Forum Moderators: DixonJones
My professor has asked me to submit a paper on click frauds. How to identify it?
I thought it is quite easy but after few days of work I can smell the difficulty level.
My Question is how can I identify a person surfing the site? Say my professor has a blog (widget.blogspot.com). He is doing some PPC marketing and paying as high as $2 pc.
Say his ads appears at
1) Google
2) Overture
3) Yahoo
when some come from these sites these are the possible parameters I can store
1) user IP
2) user OS
3) user Browser Type
4) unique ID no (using cookie or sessions)
5) Screen Resolution
6) screen color settings
(do we other parameters assosiated with user's identity)
Now say person X visits the blog through Google , after some time another person visits the blog through Overture. How do I check whether it is Person X or not?
Few possible ways are
1) If I make a cookie with some UID which can be rechecked, this can be almost 100% check. This will fail only if one machine is used by many others in case of internet cafes. but still its very a possibility.
2) Using IP (but many people can share the same IP) , so IP+resolution+browser+....(other parmeters) will help me check for person X.
My Questions
Other Issue
I cannt use php on blogspot , so I am using our university website to provide the basic php and DB support. I am using Js for that.
I real appreciate for your time and help.
Thanks
AjiNIMC
Well hell, if click-fraud were performed by Johnnie-was-born-yesterday, that would be no problem. But the majority of click-fraud is done by folk who've been in the business for a long time, and they've perfected the business to an art. To catch a click-fraudster of this nature would be about as easy as telling which one of your car's tires has a pebble in it judging by the vibration in the radio antenna.
That, and click-fraud usually originates at a nice place as a matter of revenue, since it pays per click.
Not to say it is the intention to mislead visitors, the link is labeled as appropriately and accurately as possible, but in the end it is still fraud.
Let me explain:
Click fraud is best defined as misleading a visitor.
It is done most frequently by pay-per-click (ppc) engines as the only solution to delivering traffic to sites that otherwise would get no clicks. Before anyone looks down on click-fraud with too serious an eye, lets keep some things in mind: PPC engines have a tough job supplying millions of clicks to a large network of pharmaceutical, gambling, mortgage and other sites whose owners fail to realize the supply-demand ratio is not appropriate (i.e.: too many sites, not enough interest). Furthermore, the owners pay money per click, and when someone pays per click you deliver traffic one way, or another. Unfortunately, owners of good sites get caught in this mess as well. After several years on the Web, the best way I've learned to avoid the worst of the click-fraud is to find specific keyphrases unused by most but searched by visitors, and bid 1 cent/click. The lowest bids are least affected by the fraud, but affected they still are. Then you have to wait at least a year (2 or more years is better) as your listing slowly gains ranking as all the rest of the 1-cent listings above you drop out throughout time.
Before long (well, 2-3 years or so), you have a nice number one listing for a term the usual jokers don't use but visitors search frequently. That is what you want, stay out of the comedy.
If you pay 2 cents/click, the click-fraud doubles. 4 cents/click is really bad, I do not know how some folk pay as much as 20 and 50 cents/click, still that is reasonable I have seen folk paying dollars/click.
FTS!
But one wonders exactly how ppc engines go about their business of acquiring this type of traffic.
Some get affiliates to label links such as:
Click here to search for <% terms %> on Enginename.com
somewhere on their page. This then sends the visitor to ppc engine results page with pre-defined terms. This works well for a niche site specializing in one topic or subject. For general sites, a dynamic link fed with the visitor's preferential terms (meaning the site owner gains the terms via which a visitor might have arrived at their site) works well.
Others get affiliates with search-enabled sites to display the top 1-5 results on their search result page as sponsor links via a dynamic java or xml-feed. Again, these are per-click, and what makes this fraudulent is that it is usually a content site with a large number of regular visitors who trust the site and are caught by surprise when they click such a link. This costs the site owner a slight loss of trust, but when a visitor is caught by surprise, most react by closing the end window and returning to the site to start fresh ... They won't be clicking THAT link again anytime soon. Meanwhile, the affiliate, the sponsoring ppc AND the site owner paying for the click all got a recorded click. Job accomplished, the result of this type of click-fraud is obviously poor conversion (less than 1 in 100,000 for the usual garbage, but even for a great target site {for the few good sites advertising on ppc's} conversions are no better than 1 in 1,000 to 1 in 10,000).
The next way to acquire a large amount of traffic, and I am not sure how this works exactly but these places like searchsponsor and domainsponsor acquire recently expired domains and link them to their landing page, full of cash links.
You will see these pages everywhere.
Peace
Pascal, mgr.
[edited by: webdiversity at 3:44 pm (utc) on Mar. 13, 2005]
[edit reason] Amended to comply with Terms of Service [/edit]
I have noticed that few click fraud employees do these work at a specific time (like every wednesday 6:00), they are experts as they change their resolutions , proxy and other settings also.
Your suggestions will help me get more dimensions to it.
but the very question,
to identify the person is getting tougher as from the sample data I am getting a one to many mapping
for each access entry it is leading (or can lead) to many earlier visitor.
Lets see what all can be done.
Thanks
Aji
