If I had access to my users' bank transaction records, how would you monetize that information?
For example: if I see that they make mortgage payments, i can obviously upsell on mort. refinance lead gen.
If you can't monetise using a dataset like that, I am at REALLY a loss for words ...
The whole point of marketing is that you are seeking to make inferences from normally very limited amount of data, usually with high levels of entropy. When you engage in different marketing "experiments" (e.g. different landers, ads, colours, etc) what you are actually doing is inductively identifying and selecting a combination of attributes to produce a model that creates a probability estimation for a certain behaviour you want to elicit within a target sample.
If you compare attributes as independent variables against the desired behaviour you are trying to elicit as your dependent variable, normally these will exhibit extremely weak r squares (really you are trying to infer whether a person is interested in eating out this evening for example, based on whether they are male or female, or between a certain age class), but despite these low correlations, you can typically compensate for this with massive volume. (This is the same reason why in regression equations a coefficient can exhibit extremely a slope but still possess extremely high t scores as n gets larger and larger).
What do you think the slope coefficient will be between someone who regularly eat out on Tuesday nights and the likeliness that he or she will be eating out tonight? I am willing to bet you a steak dinner that it will certainly be larger than the likelihood that a male between the age of 36-45 will eat out tonight. You combine such information rich actual (not inferred) behavioural and spending data, with other datapoints, run a few clustering, SVM, classifier, and Bayesian algorithms, and you can pretty much predict that person's next actions.