"Now wait," you say. "Isn't the whole point of targeting getting closer to the customers we want to reach?" Of course that's the goal -- and no one can say targeting isn't a good idea. However, how good is it in practice? In putting so much emphasis on what targeting data can tell us, we may collectively be ignoring what targeting cannot tell us and missing big opportunities in the bargain.
For all impressions, you can measure with certainty referring URLs, daypart, and other details like the geographic location of the IP address that caused the impression. That's what we know we know.
Then there's what we think we know (for only about 20% of the impressions!): Gender, age, household income, attitudes towards purchasing, and more. Then we use statistics to infer from that targeting data the "guesstimates" for these metrics based on analyzing thousands of impressions and identifying trends based on the 20% of impressions for which they are accurate. However, do you really know a given viewer is a 42-year old male MBA likely to spend on luxury travel goods and high-end electronics? Well... maybe. That's what we think we know, based on targeting data alone. It's impossible to be certain based on statistics alone. That's why over-targeting blinds us to new opportunities.
What about what you don't know you don't know? In the history of sales, explosive revenue growth often happens when a completely unpredictable new audience segment embraces a product and drives its popularity through the roof. For example, when Hush Puppy loafers became popular again in the late 1990s, it wasn't because the 35- to 70-year-olds who valued comfortable shoes started buying Hush Puppies again in droves. It was because hip 19-24 year olds discovered the brand, thought the shoes retro and cool, and bought them in droves. What about exploring new markets that might make use of a product or service in unexpected ways? This is the area of opportunity that requires creativity, insight, and experimentation. I'd call it good old human trial and error - but we're really talking about trial, error, and success.
Targeting can help you reach the audiences you suspect will respond well to a given campaign, but you need human ingenuity to take risks and reach out into the great unknown.
If only 20% of impressions can be combined accurately with other data sets to "know" who they are, we don't have precise knowledge of the other 80%. We allow algorithms to make educated guesses. That's why a truly great media plan will leverage those algorithms, but will also have major budget committed to exploring new prospects and reaching existing customers through appropriate media placement.
Traditional media planners -- especially when working in concert with subject matter experts who run vertical ad networks - offer a critical complement to targeting alone. To run an effective campaign, you need to balance automated targeting with the benefits of working with real people who select sites, know audiences, know design and content quality, and understand the context where that high-impact ad will appear and spark interest for both your existing customers and your future ones.