Artcyclopedia: AdSculptor Methodology

AdSculptor Methodology

A smaller site like ours cannot afford the tens of thousands of dollars required to use third-party auditing to verify the numbers we report. We feel a much better solution is to encourage advertisers to look at their log reports, or better yet their raw server logs, in order to confirm that our numbers are in general agreement with theirs.

While our numbers will never match yours exactly, this is a good exercise because it protects you against serious errors on our part, and it will give you some confidence that impressions and clickthroughs are genuine.

And while it's beyond the scope of our ability to help you with this, we think that if possible you should use cookies and tracking software to monitor the success of all your ads beyond the clickthrough. The AdSculptor system can sometimes generate very impressive clickthrough rates, but that's rarely going to be your ultimate goal: you also need the "right" audience to respond so that you can make a sale, register qualified customers, generate repeat traffic, or whatever.

What follows is a fairly technical discussion of how our ad tracking system works, and especially why our numbers should not be expected to match yours exactly.

Cacheing on the Internet
Cacheing is used by browsers to limit bandwidth and speed up a surfer's online session. Browsers "remember" HTML pages and images in their cache, so that if the same image shows up on page after page, it only has to be actually downloaded over the Internet once. Or if you return to a page you have visited recently, that page doesn't have to be retrieved again.

Cacheing is also used by the servers people use to get access to the Internet. AOL, for example, has a huge base of customers and is famous for cacheing millions of pages and images so that they don't have to be retrieved from remote servers over and over. Recent statistics indicate that between 15% and 20% of all web pages requested over the Internet are delivered from the cache of these "proxy servers," without any request to the web site's actual server.

This situation drives web publishers bananas, because they cannot determine exactly how often each page is visited and because it makes it impossible to know exactly how many views, or impressions, of an ad have been delivered.

Most banner ads today are delivered with aggressive "cache-busting" tricks to prevent cacheing. The Artcyclopedia does need to use such techniques to a minor degree in order to guarantee the accuracy of our reports, but in the end the number of impressions reported is likely to be less than those actually delivered. In other words, errors due to cacheing are likely to be in the customer's favor.

Counting Impressions
We use a tiny tracking image to count ad impressions. Every time it is retrieved from our server, the information is stored in our server log. We then run a program against our server log once a day to generate our advertiser reports.

The first thing this means is that if our site is "spidered" by a search engine robot or other kind of program, that will almost always not be counted as an impression because they typically retrieve the page's text but not the pictures. Also, some surfers set their browsers to not load images automatically, in order to speed up their surfing sessions. Text-only hits also will not be counted as ad impressions.

What if a surfer brings up a page which has your ad on it, clicks on an external link, returns to the same page, clicks on another link, returns to the same page, and so on? We currently do not charge for "304" requests (i.e. re-requests) of our tracking image. Based on our testing with recent versions of Netscape and Internet Explorer, you will typically not be charged for multiple impressions of a single ad in a single session.
Please note that it is impossible for us to test every browser, every browser version, every browser configuration, every operating system and every kind of proxy server. We believe our test results to be generally true, but they may not be universally true.

As of this writing, the address of our tracking image looks like this:
http://www.artcyclopedia.com/images/dot_10.gif?6165+10000032titian.html99999

Here is a brief explanation of the different elements of the URL:

http://www.artcyclopedia.com/images/dot_10.gif: The address of a 1 pixel by 1 pixel transparent image.

6165+10000032: A unique key identifying the specific ad placed by the advertiser.

titian.html: The page on which the ad appears.

99999: A randomly generated key generated by Javascript. This number is stored as a cookie on your browser for one hour, so re-loading the page will not cause a different number to be generated. The use of this key prevents AOL and other large ISP's from cacheing an ad's tracking image when one surfer sees the ad, and then serving the same image to other surfers without our knowledge.

Matching your server log to our reported number of impressions:

If you have chosen to display a logo or thumbnail image from your own server, then the number of times your it is loaded by a request from our site should be approximately equal to the number of ad impressions we report. Here are a few things to keep in mind:

Format: We append the Artcyclopedia page name to the image name, much as we do with our own tracking image. This limits (but does not completely prevent) cacheing, and in the case that the same image is displayed on multiple pages, it makes your server reports match our numbers much more closely.
Time period: We create our reports each morning, and our "day" runs from 6:05 A.M. Mountain Time to 6:05 A.M. the next day. (This is sometimes subject to change for technical reasons.) Your numbers will match ours most accurately if you synchronize the reporting period.
Server unavailability: Sometimes our server will deliver a page but your server will not deliver the corresponding image. This may be due to network traffic, lost packets, or your server temporarily not responding.
Cacheing: As stated above, we use a random key to restrict the cacheing of our tracking image. The image on your server is not displayed using a random key, and is subject to cacheing.

Counting Click-Throughs
We use a redirect script to track clickthroughs. That means that every time a surfer responds to an ad by clicking on it, they are in fact calling a program on our server, and that program redirects them to your site. Doing this gives us the ability to track essentially 100% of clickthroughs from ads delivered on our site.

Matching your server log to our reported number of clickthroughs:

There are two ways to accomplish this:

Use a special URL to track clickthroughs. This can be done as simply as adding "?source=Artcyclopedia" or something similar to the page name in the URL.
Look at the referrer information in your server logs.

The difficulties with auditing clickthroughs are somewhat similar to those with auditing total impressions:

Time period: As above, our 24-hour reporting period begins at 6:05 A.M. Mountain Time.
Server unavailability: As above, your server may not receive a clickthrough for various reasons.
Cacheing: Again with the cacheing. In some cases, the destination page for your site may be served up from the surfer's browser cache or from a proxy server cache, with no page request to your server.
Referrer info: If it's important to you to confirm our clickthrough numbers as accurately as possible, using option #1 is preferable. That's because browsers do not always report the referring page, and on occasion report it inaccurately.

Summary
No other advertising medium can report to you with the speed or the accuracy of the Internet. And the Artcyclopedia offers you a degree of targeting for your message that you're unlikely to find online or offline.

But the distributed structure of the Internet makes it impossible to come up with a bulletproof measurement, or even definition, of ad impressions.

We believe that the system we use is very accurate and very fair. If your reports show a substantial discrepancy from ours, or if you detect a pattern of activity that you feel you are being unfairly billed for, please let us know and we will do whatever we can to resolve the problem.

We also welcome comments on technical aspects of our tracking system. If you can see any notable source of over- or under-billing, please let us know.

John Malyon
President
Artcyclopedia, Inc.
(403) 547-9692

Back to the AdSculptor main page.