Recently I read Matt Roche's third Online Publishing Insider on optimization approaches and thought "Wow, this guy is nailing it." Until I got to the end and realized that with his "drawbacks," he was confining his perspective to largely ecommerce-driven promotions that take place on a marketer's Web site, with audience information only the marketer can know.
There is an entirely separate dimension of behavioral targeting that marketers need to understand and embrace, as it will make not only their intra-site targeting better -- but their media planning and buying more efficient -- and by the way, make them look like rock stars to their bosses.
Let's address some of the specific drawbacks that Matt mentioned.
1) No such thing as a universal model. This is an interesting point Matt is trying to make. However, within a vertical industry a model (audience segment) can work for many clients. Take the specific marketing needs of packaged goods companies interested in reaching moms with kids. A segment created for this purpose is likely to be valid for most of the companies in that vertical -- and interestingly enough, will be valid for other industries as well, such as an auto company trying to build awareness for its new minivan line, or a health company promoting a new children's medicine.
2) Behavioral models require tuning and are opaque. A fine entrée purchased at a restaurant sometimes needs salt and pepper, but that doesn't mean you need to see the recipe to enjoy the meal. What is most important in this case for marketers is to get feedback on the audience segment targeted. This information should not only tell them that they are, in fact, reaching the right audience -- but that it is performing to their expectations. This information should include a laundry list of items, ranging from demographic and behavioral profiles to reach/frequency analysis, brand awareness metrics to ROI reporting. This feedback makes the marketer much smarter and has the opportunity to inform future media plans and product messaging strategies. Demand this type of information from your partner to ensure success and build expertise around the audiences you seek.
3) Behavioral can make mistakes. The goal of behavioral is to be more right that wrong. How many plans have you worked on where the market analysis was wrong and you had to scrap the messaging and start over? You have a much better chance of success when you're leveraging the information gleaned from your behavioral partner, rather than flying blind. Sometimes uncovering the unknown about your target audience reveals an entirely new aspect that can be leveraged for creative messaging, remarketing, cross-promotions or upselling.
4) Black box. How many media partners have you worked with that expose their optimization algorithms to you pre- or post-buy? None! That is because the how is much less important than who or what. As long as your partner is providing you with a target audience that is working, and relevant information about that audience that makes you smarter, there shouldn't be any issue. However, if you aren't getting the information mentioned earlier, then you are probably working with a company that isn't really doing behavioral.
5) Remove the voice. I completely agree here. You have to use the data, it doesn't lie. Marketing is part art and part science; with behavioral, the science can amplify the art, unearthing key insights that would never have been known with traditional planning means.
Behavioral will never replace the marketer, nor will it replace contextual buying. If used correctly, it should complement your existing programs, inform your messaging and make you smarter, improving all facets of the marketing program.