Ad Placements -- Contextually Speaking

In the biz we call it "make goods:" when an error or omission with an ad calls for the publisher (or broadcaster) to make an adjustment. An example would be running a car ad next to editorial about a car recall. In fact, in print, there are proof readers who are responsible to catch just those types of errors.

The online world is a bit more dicey, as ads (or even editorial) are usually served dynamically, which is how the Drudge Report had this unfortunate combination of a gun promo above the tragic massacre in Binghamton, NY. So if proof-reading is not the answer, how can online publishers safeguard themselves from this type of truly disastrous positioning?

The key is to be able to contextually understand the nature and tone of a story. General Motors would no more want its advertisements next to a story of its current economic malaise than it would next to a recall article. One solution would be to have the editors tag the story as being negative or positive. It's a great idea but limited in practicality since stories may run that are from wire services, stringers or archives where tags would be limited. Further, it's debatable whether editors would take the time to tag (accurately) anyway.

A better solution would be to automate the process of tagging and create metadata by semantically analyzing the content; using a taxonomy for categorization and sub-classification, authority files for entities and analyze by tone and sentiment. This would prevent specific categories of ads from running on certain types of stories. True, the challenge is made more difficult if the ads are being served from a network. However, this should be an easy programming work around.

If I were a brand advertiser advertising online, as part of my risk management I would want assurances that my ad would not run under conditions that might damage my brand's reputation. Or else I would want a make good.