We often hear of digital marketing nomads who claim to work so little thanks to automation. This is a myth. Marketing is never automatic.
Most written work on marketing offers a business-first perspective. For example: the popular 4Ps of marketing or the 80-20 rule. However, much of automation for data and marketing today is based on machine learning (ML) concepts. Thus, I offer a different perspective on marketing: the analytics-first perspective.
Learning the language of ML allows a framework for us to function in this age of marketing. Let's borrow some concepts from reinforcement learning; a subset of the wider ML world. Reinforcement learning has two key parts: exploitation and exploration. Much of the other parts of ML deals with exploitation: doing more of what works or doing it in the most optimal manner. Once resources are spent on exploitation, we are unable to spend it on exploration: trying out something new in hopes of finding a better performer.
The classic example used in reinforcement learning is the idea of playing slot machines in a casino. Suppose there are 3 machines: A, B, C. We could continuously play all 3 machines serially: A, followed by B and C until we find one that wins us some money. We are exploring the machines to find one that gives us returns. After an eternity, assume we do win some money from Machine A. Now we have a choice to make: do we keep playing on Machine A (exploit) or do we continue exploring?
Auto-optimization offered in most digital advertising platforms are focused on exploitation: find which copy or ad group works best, do more of that. Every time we choose to exploit, we miss out on exploring. A fine example is a particular copy could work very well in winter but if we continue exploiting that copy throughout the year, we miss finding out a new copy might have been better in summer. Beyond changes in weather, there could also be an entirely different marketing strategy that works better than the strategy you are currently exploiting. For example: when you are focused on growing awareness but you really would be getting more returns from growing advocacy. This supersedes exploiting current copies or ad groups and is not uncovered with ML.
At Amplified Analytica, I do not believe in the set-and-forget philosophy some digital advertising nomads advocate. With a deep understanding of the limitations of advertising algorithms in these platforms, I seek to apply the iterative cycle of:
Review current exploitations
Revise the trade-off between exploitation-exploration
Research opportunities to explore
Comentários