Reinforcement Learning (RL) enables MYAA to learn in an interactive environment by trial and error using feedback from its own actions and experiences. Therefore, MYAA builds client specific results by testing continuously, at hyper-speed, constantly analyzing and re-analyzing the results, and then applying the optimizations to the campaign.

Combinatorial Optimization

MYAA relies heavily on machine learning to figure out the perfect combination of ways to engage with each customer. Some of the methods for achieving this are:

• Mixing and matching creative elements to find the right way to engage with different users at the right time and on the right channel
• Dynamic creative element combinations provide balance between performance and ad/audience fatigue
• Algorithmic research to test and optimize thousands of campaign variables in a fraction of the time it would take a room full of human marketers

Channel Contribution & Budget Optimization

MYAA wants to understand the most efficient and effective path to conversion. To achieve this she employs the following:

• She aligns to the attribution models/windows set in each of the channel providers
• Measures the impact and contribution that each channel has in differing customer journeys and touchpoints
• She continually tweaks the budget, dividing it across channels for different segments according to impact, finds optimal split and extrapolates across activity – dynamically running combinatorial optimization