Performics and the eTailing Group recently released a white paper studying the customer experience on a variety of types of affiliate web sites. The study can be downloaded from our web site. Lauren Freedman of the eTailing Group presented the research at our client summit last month and many affiliates in attendance wanted a copy of the paper.

Download the White Paper

This was our first attempt at research focused on how affiliates promote merchants on their sites. I think it has a lot of appeal to both affiliates looking to improve their sites as well as advertisers who need help understanding what criteria they should be looking for in an affiliate. We plan to enhance and extend the study in the future. If anyone in the peanut gallery has suggestions about how to improve please let us know.

The index is still pretty small. I hope they will expand it quickly. I hope researchers out there who have good data on bad software get in contact with Mr. Moore to figure out a way to share your data with this site. I think it has a lot of promise.

Here is the description provided from Microsoft:

A third-party cookie either originates on or is sent to a Web site different from the one you are currently viewing. [1].

The definition used by the Mozilla group is similar, “Cookies that are stored by a site other than the one you are visiting are called third-party cookies or foreign cookies.” [2]

Using these definitions, cookies set by a tracking system are not considered third-party cookies provided the link is clicked directly and is the primary link in the browser. There are some methods of linking that affiliates employ that do result in a network cookie being regarded as third party (e.g. HTML frame, IFrame, image reference, etc.). The following experiment will illustrate my point:

1. Delete any cookie on your machine associated with the clickserve.cc-dt.com domain (in IE, the cookie will be named “link”, in Firefox you should be able to find it by the domain).

2. Instruct your browser to block third party cookies

a. In IE, Tools -> Options -> Privacy, override default settings, block third party cookies
b. In Firefox, open preferences, view the privacy tab, check “for the originating site only”

3. Point your browser at this page: http://lea.freeshell.org/revenews-example.html – the page has an IFrame that is a reference to the Performics tracking server.

4. Check for the cc-dt.com cookie. It should not be there. The IFrame is third party content since the page is hosted at lea.freeshell.org.

5. Now, click this link: http://clickserve.cc-dt.com/link/click?lid=41000000009104360 (This is a live link from Performics’ network associated with a test account used for troubleshooting so I’m not making any money off this in case you were wondering).

6. Go back and look for a cookie from cc-dt.com (in IE, the cookie is named “link”).

I’m willing to bet that it’s there. The first example attempts to set a cookie from a third party context, the second sets the cookie from the first party context. The user is in fact requesting to visit Performics’ server directly (cc-dt.com) prior to being redirected to the merchant site.

If you still do not see a cookie there are other reasons why it may not have been set. Your browser may have been manually configured to block the domain, or a privacy/anti-spyware program may be targeting the domain, a text pattern in the url or other marker. This is a different, and much less pervasive problem (I’m working on more accurately measuring this now).

In looking through the numbers, I did see an almost unbelievable variance in cookie blocking rate depending on the source of the traffic. The numbers ranged from less than one percent to more than seventy-five percent in a few extreme cases. Most often, the distinguishing factor between a site with a high rate and one with a low rate was that the low rate site had only direct links while the high rate site used frames, iframes or other attempts at setting the cookie upon a page view. I’m going to follow up with a post on this, too as there are some interesting implications to this phenomenon of site-level variance.

Affiliates, this means that if you are framing tracking links, using IFrames to stuff cookies, or employing some other view-based cookie setting scheme, it is very likely that a large percentage of your cookies are being blocked. Try linking directly to see if your conversion rates improve.

I have already worked with one affiliate to modify their linking methods and have seen a two to six times increase in conversion since the change. Powerful stuff. (If you haven’t already, you may want to reset your cookie settings, but if you don’t, I’m not going to lose any sleep over it.)

Todd pointed out that numbers at CJ might differ from those at Performics for the same set of clicks and orders due to differences in accounting. While this is true, there are a couple points of clarification.

Performics counts a ‘visit’ for every click on a link in a thirty minute time period. If a user clicks the same link two times in two minutes, that’s one visit; if they click two different links (for the same merchant) in two minutes that’s two visits. The EPC calculations use our visit numbers rather than raw clicks.

I agree with Todd’s assessment that these numbers are not really directly comparable between networks. They are much more helpful when comparing two programs within the same network. One of the other benefits of EPC is that it provides an affiliate a benchmark for each advertiser. An affiliate who has run a program for a month or two with a relatively low EPC compared to the merchant’s average EPC can move on to focus on programs whose offers are better targeted toward the affiliate’s audience.

Michael also posted a suggestion on how to improve EPC. I like his line of thinking and this fits in well with some enhancements I already have planned for EPC…stay tuned. Todd remarked on the absence of publisher EPC. We do intend to display publisher EPC in a future release. Performics’ advertisers have access to this data through their program managers but it is not currently available in our interface.

The data includes key statistics on trends in search activity at shopping.com (not sure if it includes activity from syndication partners). You can also link to the top 100 searches for every shopping category on their site. This is very useful data; the commentary is good too although I don’t always agree with their conclusions. For instance, I’d argue that “camcorder” is moving up the list not because of spring break, but because it is almost graduation season and people need to buy them now so they can figure out how to use them.

You can view historical data, and should be a good planning tool for seasonal goods. Happy browsing.

The key to stemming this type of transactional fraud is to recognize it quickly and reverse it as soon as possible thereafter. These scammers are smart and they won’t waste their time if they see that you are on to them. If it takes 30, 60, 90 days to back out fraudulent activity the scammers will think you are asleep at the wheel and the message will get out. This site in particular shows the lengths to which people will go to find a “deal.”

Affiliates considering offering consumer incentives and merchants that work with these affiliates should ensure they establish and review processes to monitor transactional fraud regularly. Here are some of the key points to consider:

Merchants must have real time credit card authorization. If you are still authorizing in batch, table all your other development initiatives until this is fixed. Provide cancellation data to your network nightly (if you use a network provider, ensure they are in turn passing this on to your incentive based affliates). The faster you can get the data to your affiliates, the better.

Affiliates offering incentives to consumers must be able to automatically receive information on cancelled orders and have a clear process for dealing with problem members. Keep track of the names, Paypal accounts, email, physical and ip addresses of users you have flagged for fraud. Chances are they will pop up again; if you catch them on the way in you won’t have to explain to your merchants why you keep letting in the same scammers every other week.

Third party tracking servers (like those of almost any affiliate, search or web analytics provider) receive several pieces of information about a user’s computer when the consumer clicks on a tracking link. They also receive those same data points when a spyware application directs a user’s browser to a merchant through a tracking link. This information, provided to the tracking servers in the form of HTTP header data [w3c.org], can be used to help divine the source of an affiliate’s traffic.

Each time a user’s computer requests a tracking link, their web browser sends an HTTP header which provides some (potentially misleading) information about the web site they were previously viewing, the address from which they are connecting to the Internet, the type of browser they are using and some other technical details. The referrer component of the header is supposed to be used to tell the tracking server the web page being viewed immediately prior to the current request. This can be very helpful in identifying how an affiliate generates traffic [Chris Sanderson]. Sometimes this data gives us exactly what we’re looking for, but it can be obscured or disguised in a manner that tells us very little. I have seen this lead to misdiagnosis of fraud on more than one occasion. The remainder of this post will describe what referrer data is, how it can be used to help discover sources of affiliate referrals, and how it can be mangled for legitimate or illegitimate reasons.

I will try to explain in the context of that classic of American cinema, Ferris Bueller’s Day Off [IMDB]. There is a scene in which Ferris (a senior in high school) requests a table at a four star restaurant by giving a name he sneaked from the reservation book: Abe Frohman (who the host knows to be “THE Sausage King of Chicago”). The maitre ‘d was about to throw them out of the restaurant when he gets a phone call. On the other line was Ferris’ girlfriend, Sloane who asked for “Abe Frohman, The Sausage King of Chicago”, describing him as a spitting image of Ferris.*

Identifying someone you’ve never met is tricky. It usually helps to have a point of reference, a description of features of the someone (or something) you are trying to identify. One of the points of reference we use to identify affiliate traffic sources is referrer data. But this data can be misleading; in the scene from FBDO, Sloane is referrer data. The maitre ‘d, having no additional context, could use only this reference to authenticate Ferris and thus showed him to his table. References are only good if they are reliable, and there are a variety of cases where HTTP referrer data breaks down as a reliable source:

• Certain personal privacy software obscures or nullifies the header data provided to web servers
• Legitimate publishers may redirect traffic through their own tracking server or code their web sites such that the referrer is blank
• Illegitimate publishers utilizing software to direct traffic may programmatically alter the HTTP header information including the referrer to make the traffic look benign (at least from the sole context of referrer)

Despite this it is often fruitful to examine referrer data when looking into a suspected affiliate. I view it as a a mandatory step to perform when investigating a fraud claim. It can not only confirm fraud, but also exonerate a suspected affiliate and help us to build trust for the affiliate with the merchant.

Since this data may be misleading and inconclusive, it must not be considered in isolation. Regulating the integrity of an affiliate marketing program requires a host of other data points and testing methods. I will be following up to this post with additional entries describing other useful methods for testing and authenticating affiliate traffic.

* That scene and the following meal rivals The Blues Brothers [IMDB] for best inappropriate dining experience in a Chicago based film. If you’re interested here is the entire script [idiotsavant.com].

You might think it’s a stretch, but here goes…

When an individual is born into this world, the strongest predictor of their future prospects for health, education, economic status, etc. is their geography. I cannot think of a greater inequity, and I was born into a family that loathes inequity. A free and universally accessible Internet is a critical factor in eliminating that inequity. I believe that a free Internet will help to improve the standard of living for billions of people who have spent years, decades, centuries struggling for basic survival. A person anywhere in the world with a computer and a connection can now stand on the shoulders of countless giants for free, no strings attached.

The free Internet we have come to know does not exist without the work that you and I do every day. There are two critical elements for the Internet to be useful in underprivileged areas:

1. Free content
2. Access to hardware

I am not presently equipped to tackle the second problem. Thankfully others far more capable have already begun to address the issue (1, 2). But I do think that I can make a meaningful contribution to an economic environment that ensures free content and services will continue to thrive. The money that consumers spend directly with advertisers via the Internet, the investment in the infrastructure to support the necessary communication, and the subsequent dollars spent consuming and servicing the things they buy has provided the stimulus to build the Internet we know today. It will fuel pushing the Internet out to users who would otherwise never have a chance to engage in this global exchange. And, it is enriching people and companies with a penchant for change.

Companies that strive to protect the interests of their current and future customers are those that will win in the years to come. To wit, in the wake of Katrina, it was Wal-Mart, not FEMA that was able to mobilize efforts to help people in afflicted areas immediately following the disaster. Wal-Mart was heroically philanthropic, but I’m sure they secured untold future profits (and deservedly so) by being able to reopen stores and keep shelves stocked better than their competitors. Twenty years ago, Wal-Mart would never have been able to respond to the disaster as it did. Say what you will about Wal-Mart’s business practices, a thousand mom and pop hardware stores would never have been able to do what Wal-Mart did.

Technology has changed advertising just as profoundly as supply chain management. Digital advertising facilitates the most efficient and effective communication and commerce services between merchants and their consumers aside from face to face salesmanship. This places immense power in the hands of marketers; and with power, obligation. We have seen countless ways in which advertising has broken users’ trust. If trust is lost for good, the whole system will break down. The cost for currently free services would be forced directly upon individual users. The people who need the Internet the most would not be able to afford it.

On the other hand, if trust is cultivated Internet use will continue to grow. Just like Wal-Mart, it is in my (and your) company’s best interest to ensure that our customers – the individuals at the end of the wire – can reliably interact with us and have faith in our intentions. Doing so means sustainable business models and the continued evolution of the free services that make the Internet so valuable. I have the privilege of being in a position where I can work from within to help build the system the right way. I want to push the Internet’s reach as far as it can go while ensuring that it is done in a way that helps and not harms the people that we are reaching.

I am very excited to have an outlet to engage in a purposeful dialog with those who have the power to guide the future of an industry that cultivates billions of relationships across all barriers physical, economic and cultural. I’d like to thank Jim for inviting me to join the discussion here, and I hope that my contribution will meet the high standards set by my fellow authors.