Setting Behavioral Targeting Standards, Part 1

With new advertising technology comes a need for standards among similar products and services. This generally occurs when the technology’s basic benefits are realized and accepted by market leaders and when focus shifts from testing limited campaigns to buying at significant scale.

The paid search market went through this almost two years ago. Behavioral targeting will go through it in a year or so, as it continues with its promise to deliver more relevant ads to interested users, resulting in more effective and efficient exposure for advertisers.

I know this because buyers are demanding behavioral targeting standards from publishers and their technology and services vendors. As buyers find success, they want a minimum amount of friction in buying, executing, and measuring campaigns across multiple sites and networks. That means standardized terminology, packing, pricing models, measurement tools, and data policies across all sellers, whether they’re portals, independent publishers, representative firms, desktop applications, or ad networks. It means replacing apples-to-oranges comparisons with apples-to-apples and oranges-to-oranges.

It won’t happen overnight. The market is too immature, and products are too dynamic to be easily tamed. Don’t forget, Web-based online advertising, over 10 years old, is just now nailing down a standardized definition of an impression.

Though it’ll take time to complete the job, prepare now for the development of basic standards for buying and selling behaviorally targeted audience segments.

We should start by establishing some best practices in developing behavioral segments. Below are my personal (and certainly biased) views on two behavioral segmentation practices.

Polluting Behaviors with Irrelevant Context

There’s lots of confusion about where contextual targeting ends and behavioral targeting begins. There shouldn’t be. Contextual targeting is about the page on which an ad is delivered. Behavioral targeting is about the known behaviors and characteristics of the person an ad is delivered to.

Why do people confuse the two? Some publishers use contextual robots or crawlers to define behavioral segments, rather than rely on standardized user navigation and editorial triggers to define the content a user requested.

For example, some publishers categorize any user who has read an article containing the word “car” in the past 30 days as an in-market auto buyer. Yet words such as “car” regularly appear in all types of stories, from accidents and police chases to reports on clear air standards.

A better approach is to only include audience members who read content specifically relevant to auto buying over that period, such as in auto classifieds and sections on new car features. To include every mention of “car” to define an in-market auto buyer pollutes otherwise valuable audience segments and certainly harms results. If early-stage users work from polluted segments, results will be unfavorable. They may think behavioral targeting doesn’t work.

Diluting Relevant Audiences With Irrelevant Ones

In an effort to better manage inventory, some publishers mix audiences relevant to their advertisers with unidentified, irrelevant, run-of-site audiences. The most valuable audience segments become diluted. Rather than deliver a car branding campaign in two, transparent components, say in-market auto buyers and run-of-site impressions, these publishers put both in the same campaign. They try to justify doing so by labeling the new segment with a “composition” score, sort of like the proof score in alcohol or the fat percentage score in milk. They hide the bad inventory in the good. This is wrong.

Dilution and artificial scoring aren’t the answers to inventory management. I advocate full transparency and online truth in labeling. Behavioral segments should be as defined and requested by the advertiser. Audiences outside those segments must be delivered separately.

Creating an artificial score for the segment’s audience composition probably only serves to commoditize the inventory. It hides the unique value and characteristics of the publisher’s brand and content.

Segmentation Practices to Adopt

Once again, my personal and biased views: Behavioral segmentation should be based on user navigation, not on hyper-contextualization. Segment definitions and descriptions should be simple and intuitive, not overly technical or complex. Finally, and most important, segmentation should be open and transparent. It’s the only way buyers and sellers can understand what’s sold and delivered, what works and what doesn’t. It’s the only way we can prepare for the development of robust standards.

Nominations are open for the 2004 ClickZ Marketing Excellence Awards.