To be sure, a sea change is under way in how we buy, execute, implement and optimize media buys. The focus today is on digital media buys, but non-digital media are quickly becoming part of that same process. We have witnessed the first programmatic buys for TV, radio and out-of-home advertising, and more will surely follow.
I’ve said before that media sales teams, as well as agency media buyers, had better learn how to code, as the majority of their current jobs will be replaced by algorithms in the next five to seven years. The whole media buying process will increasingly be managed in nanoseconds, in real time, all the time. It will be the (welcome) end of relics like upfronts or sweeps and other time-, calendar- or season-bound limitations.
To be honest, this does not really excite me. It is a logical evolution of the path first carved by the wizards of Wall Street. But more and more it has become an arms race between consumers — who are finding ways to avoid the onslaught of digital advertising — and the platform owners and advertisers (with their agencies) finding ways to create more exposures faster and (allegedly) more efficiently.
One of the things I always tell marketers is that smarter plans lead to better buys. Smarter plans will make your plans more effective, generating bigger impact -- and as a result, are more efficient.
Therefore, for me the real meat in marketing automation is to start thinking about the front end of the process. Perhaps I am blissfully unaware, but I don’t think anyone is really delivering radical new solutions to aid the creation of smarter plans.
In the olden days, this was the domain of marketing mix modeling. Reams of data were fed into an algorithm that could predict with pretty decent accuracy what would happen to your brand if the weather was hot or cold, or whether or not you invested an extra million dollars in TV.
Today, many marketers still rely on these same models, but the fact is that most of them simply can’t cope with the fragmentation and flux that is the modern media world. I know there are modeling buffs out there who will say that their approach has evolved and is capable of supporting the choice between Pinterest, TV or couponing (agent-based modeling, anyone?).
But to me, modeling is only part of what I think will ultimately be possible. I know that the military and the intelligence services already use complex self-learning risk-analysis and scenario-planning software. My guess is that with the increases of computation power at more and more affordable prices, the increasing availability of data and the speed at which artificial intelligence and self-learning software are evolving, it is only a matter of time before similar scenario- and risk-analysis software will determine if “now with 20% more” is a better strategy than a Snapchat brand activation.
If a self-driving car drives better than most motorists, perhaps a self-marketing algorithm can create smarter strategies than the average CMO. Perhaps marketing managers need to learn how to code, too.