Advanced Ad-Serving Features, Part 2: Third-Party Ad Servers

(Originally published in ClickZ, November 2001) by Eric Picard

Last time, we discussed advanced features of site-side servers. Now let’s go deeper. This week, we’ll go into the even-more-advanced advanced features of third-party ad servers.

Third-party servers primarily serve the needs of advertisers and agencies. Sometimes they are called buy-side servers. They are part of the business infrastructure of these groups and must reliably and accurately deliver and report on ad serving and related user actions associated with the ads.

In addition to delivery and basic reporting, third-party servers provide unified comparative reporting for all publishers in a media buy, as well as many advanced features. From a feature standpoint, a third-party server is more complex than its site-side counterpart.

One thing to keep in mind: A third-party server is not able to “refuse” a call for an ad. If an ad tag is supplied from a third-party server to a site-side server and that ad is called, it must be served. Only a site-side server can schedule and deliver ad calls to users.

Beyond Banner Tracking

This is the big feature. Tracking beyond the banner enables the view of an ad session from impression to conversion (and beyond). This is a major reason a third-party server is a must for most advertisers. Some tracking types beyond the banner are:

  • Tracer tags. Tracer tags are single-pixel images placed on pages of the advertiser’s Web site so that activity on those pages can be correlated to the view or click of an ad. 
  • Post-click analysis. The user sees an ad and clicks on it. She arrives at a landing page on the advertiser’s Web site. She travels across three pages that have tracer tags on them. Each intersection of creative/tracer is credited to the advertiser’s reports. 
  • Post-impression (also called post-view) analysis. The user sees an ad but doesn’t click on it. That user (remembering the message) later travels to the advertiser’s Web site on his own. He moves across a number of pages with tracers on them. Each intersection of ad and tracer is correlated and credited to the advertiser’s reports. This analysis is a definitive branding measurement and is sometimes called a brand response report. Not all third-party servers collect post-impression data.


  • Cross-publisher reports. A major reason to use a third-party server is that reports are covered across all publishers within a campaign. 
  • Comprehensive data sets. Since both post-view and post-click data must be recounted, reports must be unified and comprehensive.


Some third-party servers offer advanced analytics capabilities. This is one of the fastest growing areas in the industry. Far more data is captured in an online ad campaign than in an offline one. Turning that data into actionable information isn’t simple. It takes days or weeks of human intervention and interpretation.

A powerful analytics package solves these problems by providing tools to get at actionable information more quickly. There are two basic types of tools to discuss:

  • Online analytical processing (OLAP) tool. This very powerful analytics tool enables the most control of data and reporting. Great power and flexibility comes at a great price, and few people are technical enough to use an OLAP tool to manipulate their data. In most agencies there are only a few, if any, people who can use these tools. It gets even sparser at the advertiser level. 
  • Wizard. To address problems with OLAP, some companies have started coming up with wizard-based interfaces for the most commonly asked questions. A good wizard-based interface can likely answer such questions as: Which publisher is the best media buy for my campaign goals based on the past six months of running ads across various publishers?


Analytics deals with historical analysis to improve ongoing and future campaigns. Optimization deals with live campaigns that must be improved while still running. When done by hand (as is most often the case), only so much can be changed. Humans can optimize to a level of detail only so deep. This is best handled by technology, which provides much deeper analysis of data. Two types of optimization are:

  • Real time. Real-time optimization is the most powerful. Changes are made automatically to creative in rotation across placements based upon actual results read by the optimization tool. Real-time optimization requires real-time data to make changes. Few ad servers use a real-time reporting architecture, relying instead on 24-to-48-hour-delayed data. Real-time benefits include microtrend discovery (intraday changes in behavior within placements) and greater lift based on feedback loops. Additionally — if the system doesn’t make changes automatically, relying instead upon human approval or intervention — the lift is going to be lower. 
  • Recommendation. For situations where real-time data isn’t available, recommendation-based systems are the alternative. These systems read data when available and provide a list of recommendations to enable the customer to make changes. This inherently is a poorer performing model as changes are not happening quickly. Therefore, additional learning for the optimization tool is lost. The faster changes are made, the better the system gets at predicting performance. Still, this is a better method than hand optimization.


  • Geographic targeting. Geotargeting is similar to site-side servers but somewhat less effective. You pay for the media regardless of whether you had an appropriate creative for the users an ad was served to. Wherever possible, try to geotarget at the publisher level. 
  • Profile-based targeting. As I detailed last time, ads can be targeted based on Web-surfing habits. Third-party ad servers have the same issues as site-side servers do. 
  • Session-specific targeting. Specifics include domain, browser type, and operating system. Again, this can be accomplished on the site side, usually to greater effect as the publisher only shows the ad (and bills you) when there is an appropriate fit. When served by a third party, you pay for the media even if it doesn’t fit your demographic.(Remember, there are plenty of other types of targeting I’m not covering here).

Trafficking Controls

Without a third-party server, trafficking ads to multiple publishers is a problem. It can be complex, with many points of failure. A good third-party server simplifies the process of trafficking campaigns and should provide valuable accounting methods for successful delivery and approval of your ads by the publisher.

Dynamic Ad Serving

Most publishers have a limit on the number of ads they will accept at one time. Usually this ranges from 5 to 10 creatives per week. Third-party servers use dynamic ad serving to rotate multiple creatives through one ad tag. This allows the advertiser/agency to traffic as many creatives associated with those tags as they want. This simplifies life for the advertiser and the publisher by cutting down significantly on the work done by both.


There are other ad server features not covered here. But this is a column, not a book! You should now be educated enough to talk to a salesperson without too much trepidation.

Next, I’ll write about a topic near and dear to my heart: how to work with tech companies for long-term success. It’s time to set a few things straight about this marketplace. Customers need to understand that while they are in a position to beat up their tech partners (notice I don’t call them vendors) on issues such as price, they should think twice. If there are any tech firms out there that would like to voice their thoughts on the topic, drop me a line.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: