Log Files vs Javascript Code

Thursday, May 22nd, 2008

I have written in the past about the various mechanics of click fraud software on the market, but I thought I would extend on this and talk about the benefits of log file analysis versus javascript tag implementations.

If you have checked out my consultancy page, you will see I am a fan of log file analaysis, in this post I will say why I hold this position, and it is due to two incredibly important facts

You can Disable javascript on your browser

Users have the option to disable javascript on their browser. As JS is client side scripting language, if the browser is not prepared to run your script, there is no way to capture metrics on the visit. Any tech-savvie click fraudster will disable javascript as their first port of call. This then renders your click fraud tool useless.

Some bots can be in and out before a script is loaded.

Analysis of certain click bots has shown that they can hit your site, register a click with Google and drop off your site, before your javascript has had time to load up and capture the visit details. This visit will however be recorded in your web servers log files.

How to Test This

Don’t belive me, then here is a quick test in another area of log file vs javascript to prove the problems with scripted implementations.

1) Review your traffic data from a log file analysis package such as webalizer

2) Review your traffic data from a javascript service such as Google analytics.

I am prepared to bet that your visits will be higher on the log file to javascript package. This is due to a number of factors including the two outlined above.

As a parting word I would like to say that there are some instances where javascript tags are the only way to analyse your traffic, for example if your ISP does not give you access to your logs in a hosted environment. If that is the case, go with the scripts, the more evidence you have, the better your chance of a refund from the search engines.

Tags: click fraud software
Posted in Features, How To | 1 Comment »

Click Sentinel Click Fraud Solution

Saturday, February 16th, 2008

I thought I would bring this post back up from my archives, I thought Click Sentinel had dropped into the dead pool, and development had stopped, but it looks like click sentinel has risen from the ashes and a new version has been release, I look forward to seeing what they have to offer, and will bring you a full review when it arrives (note it was due for release in Jan 2008)

I came across a new product in the click fraud product portfolio called Click Sentinel.

It is one of the increasing number “community” click fraud products where data is collated from a number of advertisers to help gauge the real level of click fraud which we cannot get due to the lack of transparency from the search engines. I talk about this subject at greater length in my post Click Fraud Caring Sharing Few.

Click Sentinel is a javascript implementation of click fraud software.

I have not yet completed my review of the product but first impressions are a good product.

UPDATE: I have been using the tool and it is really good, but it looks like development has slowed and the promised IP address updates are not coming. This is a real pity, a community click fraud tools at this price is an excellent idea.

Tags: click fraud software
Posted in Features | No Comments »

How Click Fraud Software Boosts the Bottom Line

Monday, November 26th, 2007

I have held the opinion for a long time that click fraud software is an investment rather than an overhead. In this post I intend to publish the figures which I believe support my opinion that click fraud solutions boost the bottom line of a pay per click campaign.

Assumptions

My calculations are based on a number of assumptions, these are of course subjective, please feel free to challenge my assumptions and produce your own figures in the comments of this post. I see this post as a discussion on this topic and I am keen to gather feedback on my calculations.

Assumption 1 - click fraud is 16%

There are many, many opinions in the level of click fraud. Pay per click suppliers such as google suggest it is very low (2-3%) whilst some players in teh click fraud monitoring commuinity suggest it is in the upper 30% bracket. I don’t have the resources to do my own analysis of the click fraud problem, but I know a company which does. Click Forensics published the much respected (except by Google that is :-)) Click Fraud Index, which suggests that click fraud is currently running at 16% of all clicks. After analysis of their methods, I am happy to use their numbers.

Assumption 2 - average cost per click USD 0.05

The second assumption I would like to make in my calculations are an average cost per click value. For the purposes of this post, I am being very conservative with my costs and I am stating that avertage cost per click is USD 0.05. It is my experience that the majority of campaigns cost more than 5 cents per click, so assuming this low amount helps to highlight how click fraud can become a very big problem with high value key words.

Assumption 3 - conversion rate of clicks to sales is 5%

To create an income calculation, I need to give a conversion rate of clicks to sales. Again like the cost per click calculation I would rather err on the cautious side and I will set a value of 5%

Assumption 4 - each sale brings USD 10 income.

The fourth and last variable in this post is the average value of an action. This value is entirly dependant on the type of product or service a website sells. I have therefore taken a completely arbitary value of USD 10 as the value of a sale.

The Figures

Our imaginary campaign is reasonably large, it has 50,000 paid clicks per month, below are the sums:

Cost of Clicks

50,000 clicks / month @ 0.05 =usd 2,500

Projected Income from Clicks

( 5% of 50,000 ) * 10 = USD 25,000

Cost of fradulent clicks

16% USD 2,500 = 400

Cost of lost business

16% of 25,000 = USD 4000

Total Potential Losses

4000 + 400 = USD 4,400

Conclusion

These are very large amounts, and when multiplied to very large campaigns, the amounts become very substantial. When considered that the average cost of click fraud for 50k click per month is only USD 50 (source Adwatcher and Who’s Clicking Who) I reiterate my claim that click fraud software is not a cost but an investment in your pay per click marketing campaign. Using this model, it is feasible that a tuned campaign with click fraud software installed for monitoring could make an additional 4k USD per month.

Caveat

The glaring ommission from this model is that pay per click suppliers filter invalid clicks at source and either do not charge for clicks or refund any invalid credits which get past the filters. The problem is, in my opinion, the filters do not catch all click fraud attacks. In support of this statement I set my readers a task. Check the credits on your account, do they look like the figures I am quoting? I doubt it.

Tags: click fraud software
Posted in Features | No Comments »

Adwatcher Click Fraud Software Review

Sunday, November 18th, 2007

In this post I review Adwatcher a Click Fraud software solution produced by MordComm Inc.

I have been using Adwatcher for some time now, and thought a review of their click fraud software may be of interest to readers of this blog.

How it works

Adwatcher like many click fraud offerings is first and foremost an analytic package, but it has a large number of click fraud monitoring functions.

Adwatcher comes in one of two offerings. The first is a hosted solution where all installation and configuration issues are dealt with by the Adwatcher staff. The second offering is a downloadable version which is hosted on a companies own server.

The downloadable version is a PHP and MySQL application which also requires the Zend optimizer, a free optimisation and PHP encryption solution.

Adwatcher works via the redirection method as discussed in Mechanics of Click Fraud Software. Changes are made to your PPC campaigns to redirect all ads to a specific landing page. This page captures the metrics of the click before redirecting to the true target page. For each target page required, a landing page must be created in the Adwatcher console.

Features

As previously mentioned, Adwatcher has many functions over and above click fraud monitoring, but for the purposes of this review, I will only be highlighting the features which are specifically relevant to click fraud detection.

Fraud Settings

Adwatcher enabled you to set your own tolerance for click fraud. The fraud setting specifies the number of clicks in a defined time period which you would like to set as your tolerance for click fraud.

Suspicious Activity Alert

One of the features which differentiates Adwatcher from other click fraud solutions is the suspicious activity alert. If adwatcher detects multiple clicks, a warning error will be displayed (an example is shown below) to the person clicking on your ads. This type of interception of click fraud at source and issuing a deterring warning would be very difficult to do with javascript type click fraud solutions.

WARNING – Suspicious Behaviour Detected

Your internet location has been detected visiting this site more than 5 times over the past 24 hours. Your information and IP address has been logged and sent to the owner of the advertising campaign.

We appreciate your interest and thank you for your visits. However, to protect our customers from higher prices by keeping our advertising costs down, we routinely examine recurrent visitations from their advertising campaigns.

Please help us pass the savings on to you by bookmarking our site for future reference.

E-mail Alerts

Adwatcher has an e-mail alerting function to warn when suspected click fraud has occurred. There is also an SMS function to send text messages to cell phones. NB this only currently works with US carriers.

Generate Refund Request

Probably one of the most powerful functions of the software is the automatic generation of refund requests. At the click of a button, a report is generated showing fraudulent activity which can then be passed onto your PPC supplier as part of a refund or credit request.

Spike Analysis

Spike analysis is an interesting function of Adwatcher. It trends your normal campaign activity and enables analysis of spikes in activity. For example if your normal click rate was 100 clicks per day, running a spike analysis it is very easy to spot anomalies where 2x , 4x up to 10x the normal click activity is happening.

Concerns

Some of the issues I have noted with Adwatcher are shown below:

Activity from Google is classed as click fraud. By default I would recommend adding Google to the list of IPs which can create multiple clicks. The testing tools on your Adwords account appear as click fraud.

I used the hosted version and I was concerned that referrer information would be lost as all clicks would be referred from an adwatcher server. Being a bit obsessive about my stats this was a concern. An investigation of the advanced options showed that there is an option to retain referrer information and pass it through to your other analytic tools.

My final concern is with the hosted version of the software. If for any reason, the hosting servers go down, your campaign clicks will return a 404 file not found error. I must quickly add I have no evidence of this happening, and I trust a company of Mordcomm’s integrity to have the appropriate fault tolerant systems in place.

On-Line Demo

Adwatcher has a demo version of their hosted solution for review. It is useful to check out the features mentioned above. There are also numerous flash training resources.

Pricing

At the time of writing (Nov 2007) the pricing of Adwatcher was based around the number of paid clicks a site receives and if you use the hosted or self hosted version, the prices are taken from Adwatcher’s own site

Free Trial

Adwatcher supply a 30 day free trial demo. This is a fully functional version of the system. I took on a demo of the hosted option before opting for another option and forgot to cancel the credit card authorisation once the demo was up. I contacted the support team and was very quickly reimbursed, something not so quickly forthcoming from some companies offering 30 day free trials.

Conclusion

Adwatcher is one of my favourite click fraud software solutions, it produces huge amounts of top quality evidence to supply to your PPC supplier, in fact I used Adwatcher to capture the click fraud evidence I supplied to Google in the post Click Fraud a story of intrigue.

There will be an effort to retro fit existing campaigns to meet the new requirements, and this could be a large task for companies with big campaigns, but I believe this is true of all click fraud solutions.

If you don’t already have a click fraud software supplier, I highly recommend you try a demo of their solution.

Tags: click fraud software
Posted in Features, Reviews | 1 Comment »

3FN Marketing Enhances Their Fraud Detection System

Tuesday, November 13th, 2007

3FN a second tier pay per click company has boldy announed that it now detects 98% of click fraud attempts through new fraud detection software.

Read the full press release.

It is my opinion that smaller second tier companies need to introduce inovative practices like this to enhance their product and help take market share from the PPC giants.

Tags: click fraud software
Posted in News | No Comments »

Intel Capital Invest $10M in company producing Click Fraud Software

Monday, November 12th, 2007

Intel capital today announced a $10 million investment in Iovation, a company specialising in fraud and security solutions, including click fraud.

The complete press release can be seen at the Iovation web site, but the part which caught my eye was:

The company provides the first device reputation service aimed at protecting business on the Internet. The real-time service prevents online fraud including charge-backs, identity theft, phishing, click fraud and other abuses of online services such as child predation, stealing or cheating, posting unwanted content, and chat abuse.

This is a new company to me, and one I will be watching in the future.

Tags: click fraud software
Posted in News | No Comments »

Who’s Clicking Who Please Don’t Sue

Monday, October 1st, 2007

In noticed the following disclaimer on click fraud software supplier Who’s Clicking Who website. I wonder if they had a visit from a certain search engines attorney?

WhosClickingWho?™ Independent Auditing Service is a tool developed solely to help identify abusers fraudulently clicking on Pay-per-click listings, and in no way implies any abuses by the Pay-per-click Search Engines themselves. We believe legitimate Pay-per-click Service companies are also as keenly interested in preventing fraudulent click activity as their customers are. PPC Audit Inc. makes no assertions as to what constitutes fraudulent click activity. Such determinations are to be made between the PPC search engine user and PPC search engine provider, based on the user’s data provided.

Tags: click fraud software
Posted in Features | No Comments »

Mechanics of Click Fraud Software

Wednesday, September 26th, 2007

This post discusses the three main methods used by click fraud monitoring software to ensure your pay per click campaign is free from fraud. It also discusses the requirements to retrofit each method into your existing ads.

JavaScript

This method requires a small piece of javascript code to be added to the landing page of your website. This javascript then reports back to a central server when a click through from a PPC ad is made. The server is hosted either internally or externally by a click fraud service provider. Increasingly companies are selecting hosted options to reduce complexity.

This method require very little change to the actual pay per click campaign. Editing the landing page of the campaign and adding javascript code will take very little time.

Some service collect keyword data and require a change to the ads as well. This is in the form of adding a keyword tracking tag. An example of this would be on Google Adwords, the landing page would need to be changed from

www.landingpage.com/product.html

to

www.landingpage.com/product.html?kw={keyword}

The amount of time to implement this type of change will increase depending upon the number of ads your company is running.

Re-direction

The second method of data capture is via redirection.

The landing page of the advertisement needs to be changed to point to a gateway server. Once a user has clicked through on an ad, they are briefly sent to the gateway server where the pertinent data is collected. The user is then seamlessly redirected to the real target page.

This method required more retrofitting for existing campaigns as every landing page needs to be changed to point to the new gateway servers.

Log File Analysis

The last method I will discuss is log file analysis. Using this technique, web server log file are imported into an analysis program and analytical analysis of the log is made searching for tell tale patterns in the data.

Some system provide an automatic upload of the log files via ftp every 15 minutes or so to provide near-line analysis and reporting of problems. I am not aware of a product which supplies real time analysis via this method.

This method requires no changes to campaigns as analysis is based entirely on raw log data.

In a future post I will discuss my preferred method of evaluating and choosing a click fraud monitor.

Tags: click fraud software
Posted in Features | No Comments »

Fair Isaac Click Fraud Study

Tuesday, September 11th, 2007

I was very fortunate to have a discussion with Joseph Milana from the research department of Fair Isaac on a click fraud study they are currently running.

Fair Isaac is a company which supplies business intelligence product across a wide range of business sectors, the product they are most well know for is Falcon Fraud Manager which protects 20 of the worlds top 25 financial institutions against credit and debit card fraud.

Fair Isaac’s core competency is the analysis of large data sets, the data mining they do gives them the ability to do deep statistical analysis looking for patterns which appear pathological. Using their knowledge and techniques to spot fraud in the credit card arena, the are performing a study into click fraud.

Fair Isaac have issued a call to the broader ppc advertising community to supply data with which they can complete a very detailed analysis of click fraud in an attempt to ascertain if there is a real problem. I have agreed to help them collect this type of data through a call to action of this blogs readership to supply data for their study.

They require the following atomic level data and most saliently, after click information i.e. did the click convert The items marked in bold are required, the others are would like to have:

- Session ID

· Date-time of visit

· URL referrer (particularly relevant for tracking the traffic arising from affiliates)

· Client IP address

· URI Stem (the page accessed)

· User Agent (browser related data used by the client)

· Advertising Campaign (including Publisher)

· Keyword

· Daily spend limit

· Click cost

· Link ranking (i.e., w.r.t. the placement of the ad on whatever page by the Search Engines)

· Conversion Indicator

· Total time of the session

· Total number of pages migrated during the Session.

The output of the study will be a report to the general market place on their click fraud findings along with a private report to each company which participates in the study. It must be stressed that they cannot assist in any reparation of ppc fees.

If the findings are positive that click fraud is a serious issue, it is envisages that Fair Isaac will introduce a click fraud detection product. Their thinking is to place the product at a point where click fraud is detected and advertisers are not billed rather than relying on refunds from the ppc suppliers.

Any data supplied by advertisers will be protected via an NDA (non disclosure agreement) so you can be assured that any data you supply will be treated with the strictest of confidentiality.

Tags: click fraud software
Posted in Features | No Comments »