Q: Your perspective on how to use data beyond cookies. How might that benefit your marketer clients?
Dean: This is something that I think has been overlooked by agencies. We have many years of experience developing intelligence on our customer, and trying to identify propensity and buying signals from non-digital channels. When the digital channels started to become effective, connected and visual, performance data started to flow across digital channels. We are now starting to ramp up pretty quickly on connecting these direct, digital signals to a more complete profile from non-digital channels.
Why I think this is particularly relevant is that data for Omnicom Media Group and Annalect next year is not about number of cookies, but around the fidelity of data we have around those cookies. How much can we know about them, how much can we intuit about them? We have years of media research, behavioral research, years of attitudinal research, and we have some incredibly smart people who have always looked at audience groups and populations in terms of aggregates. We have some pretty detailed info on those aggregates, from their media viewership, to their propensity to share, their propensity to buy, their brand attitude and brand loyalty. When you think about all those data points that media agencies have been using for years to buy and plan media around, why wouldn’t you think about taking some of those things, and pushing them down into a digital data management platform? Basically adding what would be synthetic signals to groups of audiences and groups of cookies.
Q: What do you mean by synthetic signals?
Dean: Let’s imagine we have an audience segment that we’ve built for an advertiser called urban hipsters and we happen to know that that audience includes a like in fashion, technology, certain type of music, brands they align with, and that they live in a certain geo location. Let’s say that we know that this segment for the advertiser Pepsi, means that they have higher propensities to drink PepsiMax than anything else. What if we can match that definition of folks to people in our digital audience? Why wouldn’t we take some match keys like age and geo and use that to augment the data available about people in the digital channel (not with a direct signal, but with a synthetic signal) like you know what, this guy is going to want new sun glasses every month. As an additional example, we might also leverage what we know about TV viewership to augment our linkage to social activity, and generate synthetic signals we can test from this.
Q: I would visualize the synthetic approach as a way to gain well-informed reach, and as you are able to get people further down the funnel, you might be able to replace some of the synthetic signals with higher fidelity actions.
Dean: Absolutely, as you think about layering the synthetic attributes to the audience signals, you start to figure out which ones work and which ones don’t. If you see that person in a context where it’s possible to validate the signal, for example, you’ve intuit that a category has a high propensity for business travel and you happen to come across that they are booking travel to a business location, then you replace the synthetic signal with the concrete signal, i.e. planning a trip to London. That way you can make much more solid decisions on that audience, and how you buy media to reach that person.
Imagine the process being to take all this research, build a way for us to connect it to the digital cookie pool, overlay that cookie pool with synthetic signals, we can classify them separately, test and evaluate, replace with concrete signals if we see them, and then from there, validate if an approach like this really works.
But we know it works, because we’ve been planning media like this for a long time, and what this will give us is a great connection and testing environment, a lab if you will, to test those signals.