The inherent problem with personalized ads (psst…its the process)
I have written a decent amount about the importance of audiences in the world of marketing. Over the course of the last couple of months, I have been engaged in some of the most advanced and in-depth audience work of my entire career.
My work with teams is spanning many categories and crossing all media channels. Over the course of this work, I have uncovered what I believe to be flawed assumption in the identification and curation of audiences. To unpack this issue, let me first provide a not-so perfect process of audience development.
If you cover the right hand side of the above visual and focus only on the lefthand side, you will likely nod along. The process and the logic embedded makes sense. So where is the flaw?
To put it simply, the flaw is this: The attributes used to define sub audiences cannot possibly be the attributes that drive response to the ads. Let’s fill this out with a hypothetical example using the baby diaper category:
While this is simply an example, I encourage you to think about the last campaign you were a part of developing. The structure and logic is likely close even if grossly over-simplified.
So where is the problem. The problem, as the previous tree starts in the second to last step, and is a function of what is uncovered when we “understand audiences.” In the diaper example we uncover that expecting parents feel out of control and they value brands and experiences that simplify their lives. New parents are just looking to get the job done; purchase and have a good stock of diapers on hand at the right price. These make sense, but the issue arises in the 4th step — particularly in the development of creative. Why? Because it assumes that the reason audiences respond to an ad is directly tied to their larger motivations. This is almost never the case. It falsely assumes that expecting parents are not interested in price and that parents of diaper-aged children don’t value quality. In short, when the defferentiating attributes of an audience do not dictate the different reasons one responds to an ad.
How then can we overcome this assumption? One way is large scale experimentation. This may sound like classic AB, but rather I propose something more extreme. Go-to market with one-homogenous audience > people who need/will need diapers. Run 2% of the budget where the single audience is fed both concepts.
The expected next step here is to select the creative with the higher overall performance and increases its rotation or kill the lesser performing concept all together.
Instead of doing th above, then optimal approach would be to analyze response into 4groups: Quality Responders, Price Responders, Responders of both, and lastly non-responders . “Responder” being defined as clicked-on, engaged, completed video, or other — doesn’t really matter. You can create a look alike for each to scale your audiences and later going to market with your remaining 98% of the budget.
In analyzing hundreds (if not thousands) of campaigns in the above manner, what is almost universally true is that the in-going assumption that quality vs price response is a function of expecting vs new parents is not the most discriminant variable that predicts responsiveness. The response generally shows massive cross over, but exposes the attributes that matter more (perhaps income, geography, purchase patterns, product availability, etc.)
By following the process I propose, you are first learning “who responds to what” then deciding how to create meaningful audiences resulting in better, higher quality audiences and campaign performance. By understanding the attributes that drive variation in response to different concepts, you actually have data & knowledge that has lasting value for future creative, audience optimization, and creative variation.
When you over prescribe the creative solution independent of understanding response, you are guaranteeing suboptimal performance.
Stop guessing and start knowing.