This is where I would begin to disagree

This is where I would begin to disagree with the touchy-feely path of customer experience and omni-channel experience design. Could you not create a big data model (perhaps a nominal logistic layered on another platform [forest trees method]) to understand the statistically relevant “volume” of behaviors? Then, based on that you could determine the statistical relevance of each component in the model, and understand how one part ‘interacts’ with an output vs the results of a nested or cross metrics to create the output. Then, when you know what volume of business generates the intended behavior could you not optimize and drive more efficiently?