• by Tom Hespos , September 2, 2003

Online advertising is beginning to seriously deliver on one of its original promises - cutting down on advertising waste through the use of advanced targeting. As we learn to target and advertise to the people who are most likely to be interested in our products and services, we bring to the industry a philosophy that is not unlike that of traditional database marketing.

Smart marketers are segmenting their audiences by behavioral, geographic, lifestyle, and attitudinal factors into buckets, such that they can develop advanced ROI models based on which consumers are likely to turn into customers. Hand in hand with this comes the notion of targeting advertising based on any and all of these factors, often in combination.

Simultaneously, we've just started to get a handle on bringing reach and frequency to basic demographic audiences. As this begins to have an effect on how we plan and buy online advertising, we're faced with a conundrum: How do we determine reach and frequency against complex audiences?

Today, I can plug campaign information into any one of a handful of reach/frequency predictive planning tools and understand how each media placement contributes to my reach and frequency. But adding some of the Internet advertising industry's simple targeting filters throws a monkey wrench into the works.

Supposed I'm targeting people who own SUVs in the state of Illinois? If I want to evaluate how my campaign will reach these people, the first problem I'm faced with is determining how large the audience is. Most media planners would tackle this problem by having separate planning and buying targets. But that's a cheesy cop-out. If I really want to know how I'm doing against my core target, I need to know how many of them there are. A secondary problem is finding out how many of them are online, but I digress...

The online medium seems rather uniquely positioned to provide data on how large niche audiences might be. Research is conducted online cheaper, faster and more in-depth than survey-based research offline. Seems to me that audience size estimation is something that deserves a closer look.

Even the largest national survey bases cannot drill down to the level of granularity we need in order to estimate the size of many niche audiences with the confidence of statistical stability. There are simply too many questions to be asked.

However, if we use a mixture of declared and observed data to assist us in our research, we can start to build an audience tool that could not only give us robust audience size estimation data, but also the means to target advertising in the future.

We're just starting to leverage the targetability of Internet advertising. It does us no good if we can't understand with any degree of certainty how many people fit the targeting criteria we set. We need to know whether we're talking to an audience of 40,000 people or 40 million. And it needs to happen soon if traditional marketers, particularly the ones that are heavily dependent on database marketing, will trust the Internet with larger shares of their advertising budgets.