A Call for Counting Standards
Offline, advertisers are content to rely on sample-based audience statistics that are not completely accurate, but are at least reliable.
Online, there are census-based audience statistics. Online marketers can gather data each time an audience member looks at an ad. The problem is that due to network latency, network caching, counting architecture and other factors, different parties can count the same ad campaign in reliable but different ways. In particular, publishers tend to count higher, while advertisers tend to count lower.
Having multiple means of precisely counting advertising impressions online should be one of the benefits of online advertising. Each party has access to its own information sources and can conduct more informed negotiations.
The problem is that though free markets are efficient, they are also stressful. Many in the industry think it is too much work to keep track of different companies' ways of counting and how they differ from their own. Publishers need a counting method that all advertisers can respect. Advertisers need to be able to compare performance across a range of sites.
These counting methods are all over the board. The industry has no less than five common ad serving and counting methodologies in use by most ad management systems and ad servers. These include page view, ad insertion, ad request, monitoring downloaded ads and 302 redirects.
The argument over which methodology is better simply adds fuel to the fire for those who claim online advertising is in disarray. It is time, through the adoption of industry standards regarding impression and click counting, for the confusion to end.
The need for a set of standards is more than just a public relations problem; it is a business issue, as many companies are questioning whether they will ever accurately and consistently measure the success of their online advertising across the Internet.
Rectifying this is a sticky situation because of the nature of the Internet as well as the use of different tools by different service providers.
Here are some of the problems known to affect impression and click counts, as well as any known fixes.
Caching and proxy servers.
Caching refers to the local storage of files. It is designed to increase the speed with which computers access them. When an ad is retrieved from cache, the server has no way of knowing that the ad was requested. Content is neither requested nor delivered. This alters the count of how many impressions were actually seen.
On the Internet, caching occurs with browsers and proxy servers. Proxy servers are intermediary Web servers that lie between a browser and the content server and are meant to cache frequently requested content. They also are used by some Internet service providers, including AOL.
To maintain accurate counts, browser and proxy counting must be actively circumvented. One way to defeat caching is to make all content appear to be new, usually by appending a random number or time stamp to the URL. While most U.S. publishers and ad servers have a policy of implementing this technique, commonly called cache-confusing or cache-busting, it is often done incorrectly. Greater importance, therefore, needs to be placed on ensuring that standard cache-busting techniques are used.
Filters and robots.
Most publishers and servers use some form of log filters or cleansing algorithms. These are designed to ensure that only legitimate human activity is tallied in the final reports delivered to advertisers and marketers.
For example, many filters are designed to ignore known robots or spiders, which are programs designed to follow links or scour the Web in search of specific information. Others are based on user behavior, such as the request for several impressions within a very short time or excessive clicks on a banner.
Because there are so many variations for filters, the results often will differ significantly. The industry needs to adopt a standard practice to which all subscribe that will mitigate the effect of filtering on final impression and click counts.
Counting methodologies. As noted, there are several points in the process of receiving and fulfilling a request for a banner at which an impression can legitimately be counted. Both the Internet Advertising Bureau and Fast Search and Transfer ASA, an Internet company involved in search and multimedia technology, recommend 302 redirects as the preferred counting method.
This method consists of three steps:
• The browser makes a request to a site's Web server for a page that contains a banner; the Web server returns the HTML for the content page, including the HTML tags for the creative.
• The browser requests each element on the page, one of which is the banner. The request is sent to a "counting server" that returns a location redirect URL specifying the location of the banner; the impression is counted when the counting server receives this request from the browser.
• The browser then makes a final request to the "image server" for the actual image; alternatively, the browser can retrieve the image from its local cache or from an intermediate proxy server.
Though this is a more technically complex method, it can be universally adopted and circumvents many caching issues. It is therefore better for the entire Internet community, since caching can improve Internet performance.
There is no easy solution. No one company can implement a standard. It requires unification and agreement among companies accustomed to competing with each other. This has, historically, been difficult to achieve. The desire to find mutually acceptable and beneficial solutions needs to be shared by everyone in the industry.
This is a challenge our market faces, and it is a bigger obstacle than many would think. Many companies simply do not believe in standardization because they have grown too accustomed to promoting their own counting methodologies. While this might be a preference for these companies, it adds to the confusion over counting discrepancies.
• Maura Lewis is director of strategic services at Engage Inc. Reach her at email@example.com.