If Marshall McCluhan was right, that The Medium is The Message then all this fuss about one-to-one user targeting and segmentation may have been for naught.
For some time now the digital advertising industry has known that the foundations upon which adtech fortunes were built were going to erode and eventually disappear. The transition to a cookieless world has an entire generation of marketers confused and perplexed as to how they can ever figure out an equivalent method of identifying consumers as good as the old reliable cookie. But was it really ever that good?
Moreover, it’s not like generations of advertisers and marketers who came before and didn’t have these digital identifiers ever struggled for a strategy. In fact, they created many of the audience segments that the digital side inherited. As weird as it may sound, digital advertising owes a substantial amount of gratitude to the direct mail industry.
And with the announcement from Apple that the IDFA is going to become essentially useless with most users not providing affirmative consent to track across domains, the digital advertising golden goose that was only ever semi-transparent is not only dying, but it’s also being stuffed for Christmas dinner.
Which brings us back to the entire point of advertising: reaching people with messages that they will emotionally respond to, thereby increasing their awareness and favorable opinion of products and services that fulfill a need state. Advertising sets the table for consumer intent. It doesn’t create it alone, but it’s critical to creating it at all. More than influencers, more than product integrations, more than the correlation with social values and beliefs, advertising a product to a consumer adjacent to the content they have self-selected is the most effective way to begin a customer journey toward your brand.
For the past two decades adtech tools, programmatic buy-side, and sell-side platforms, data aggregators, and resellers have gone all-in on the precept that identifying the characteristics of a digital user represents all of the useful signals within the noise for ad decisioning. That’s like saying by isolating the second violin in an orchestra you can decide if the piece of music is any good. The reality is that consumers live in multiple modes, online and off, and that by convincing yourself that Dave is receptive to your gardening message because he visited a lawnmower site and then throwing that grass seed ad at him at 2 PM on a Tuesday is a great idea. Trust me, Dave is thinking about making it to Wednesday, not whether Bob next door has a better fertilizing plan.
Which brings us back to context. Recently YouTube announced 7 members of its Brand Suitability and Contextual Targeting roster within it’s YouTube Measurement Program or YTMP. Now, this is YouTube, the big Kahuna of digital video inventory and if they moved on contextual targeting with third parties we should all pay pretty close attention. IAS in on the list but for the most part these are smaller companies with specific competencies in understanding not only that a video is labeled “Cruise” but that it may be about the $8,000 Isaia cashmere suits he has custom made (true BTW) and not about the joys of being trapped on a vessel for 7-30 days with people you possibly wouldn’t want to share an elevator ride with. This is important. Not the elevator part, the “Cruise” part. If a travel advertiser went to YouTube armed only with keywords like “cruise”, “holiday”, and “Caribbean” the could reasonably end up adjacent to videos about the aforementioned movie star (Tom), greatest jazz singer of her generation (Billie), or an 80s music video by Billy Ocean.
By analyzing not only the expressed metadata about the video like the title and genre but also the non-expressed attributes that gather what is actually contained in the video and then ranked for suitability with your objectives, these companies are essentially segmenting the millions of available impressions available to you as an advertiser more closely using context.
The application of machine learning to video means that a typically laborious process of “logging” scenes and attributing metadata becomes a process that computational power and refinement of learning models can address. Cloud computing platforms can index entire catalogs of video available across ad-supported platforms to determine what contextual advertising programs are best suited to a campaign’s goals, and continuously refine its accuracy over time with closed-loop performance and attribution feedback.
But what about the user? How can I be as sure that Dave the lawnmower hunter is part of the audience I will reach? Here’s a tip: you can’t, and you weren’t sure about Dave in the first place. Here’s why: you made an inductive leap that Dave was in the market for grass seed because of signals that said he has an interest in gardening, lives in a place where lawns are common, and meets the age and income range for your product. You’ve biased yourself with these limited insights to believe that these data points are enough for you to understand the consumer wherever and whenever he is online. The reality is that Dave is a complex person (more than his neighbor Bob who is unnaturally focused on his lawn) who exists in multiple modes. Sometimes he’s thinking about his lawn, sometimes he’s finishing a report for the Wednesday staff meeting known as “Dante’s 10th Circle”.
By applying contextual targeting algorithms to the audience, that is to say, understanding that audiences have contextual attributes as much as the media they consume, as exhibited by the platform (e.g. YouTube) fidelity (e.g. HD), Timing (Day, daypart) you begin to get a substantially more dimensional view of this consumer aperture upon which your ad decisioning can depend. Again, Dave on the office browser at 2 PM on a Tuesday is different than Dave watching sports on his HD SmartTV at 11 AM on a Sunday.
So maybe all of the digital signals coming out of cookies and the adtech supply chain really won’t be missed so much once we’ve put some serious cloud computing muscle behind understanding context (content and user) and building predictive models that outperform digital identifiers.
All I know for sure is that Bob can’t stop buying gas-powered lawn implements and using them next to my home office window every time I’m on a Zoom call. I’m working on a predictive model based on the condition of his yard to rebalance my calendar.
