One of my favorite books in my library is a two-volume set called "The Complete Sherlock Holmes." It contains everything Sir Arthur Conan Doyle ever wrote about the super-sleuth, and it quickly introduces the reader to Holmes' style of solving crimes -- relying heavily on deduction.

Because Doyle was the one writing the story, Holmes was usually right when he came up with his conclusions. For instance, he once deduced by observing the condition of Dr. Watson's shoes that Watson had been out in bad weather recently, and that he employed a somewhat careless servant girl. He correctly deduced that since Watson's shoes were scuffed in a certain spot that someone had recently scraped mud off of them, and then logically assumed it was Watson's servant who did the work.

Logically, though, these feats of deduction don't make a heck of a lot of sense. If Holmes did anything to eliminate the countless other possibilities, it wasn't revealed to the reader. 

Deduction might work 100 percent of the time for a fictional detective, but it doesn't work as reliably for those of us in the digital advertising business who are addressing real-world challenges.

The challenges I speak of concern ad targeting. Unfortunately, as media buyers, we all need to be sleuths when it comes to figuring out exactly how online ads are targeted.

Over the years, I've seen a lot of smoke and mirrors when it comes to targeting methodology. What it comes down to is this: When a sales rep describes a targeting method, a buyer needs to drill down and find out if the targeting is based on facts, deductions, or a combination of the two.

By way of example, a proposal to deliver in-market car buyers can have a lot of behaviorally targeted options. An ad network or publisher can offer "in-market car buyers" but define them in radically different ways, often on the same buy proposal.

For example, a proposal might call someone who visits a page with pricing information on a minivan an "in-market buyer." Or it might call that slice of inventory "behavioral targeting -- visited pricing information." Obviously, if the description is more like the former than the latter, you'll have to do some digging to figure out how the salesperson deduced "in-market buyer" from the available data.

With every deduction comes risk that you're not actually talking to the audience you want to reach.

Here are some recent targeting options our agency might have missed critical details on if we hadn't been diligent about looking at the targeting methodology:

• A health network recently promised to deliver impressions against sufferers of a specific health condition. It had the data to do so, but it was married to ZIP codes, which was married to the IP address of the visitor. Think about that for a second –-- condition data is deduced from ZIP+4 data, which is deduced from IP addresses. That's two leaps of logic to overcome.
  
  
• A site promised to deliver competitive users within a specific product category. When we dug further, we found out that the site was projecting offline data to its user base by using lookalike targeting that lumped together competitive users based on their overall site visitation habits. The leap of logic here was that if someone used a specific product, other people who used that product would display identical or similar visitation behaviors on the publisher’s site -- and they were calling those people "competitive users." Seems like a big leap to me.
  
  
• A B2B site called site visitors "category users" based on their visitation to a specific content area of their website. The visit could have happened any time, at any frequency. I thought it funny that someone could be labeled a category user forever for accidentally clicking on a link two years ago.

These little details tend to be glazed over, especially when we're in a down economy and advertisers are looking for scalable yet targeted ad solutions. Always make sure you're asking everyone you take proposals from about how they're targeting their ads.

Tom Hespos is the president of Underscore Marketing and blogs at Hespos.com.