"accountability" is often as much about politics and perception as it is pure measurement. And, in many of today's firms, the analytics group within marketing is focused on the micro at the expense of the macro. And when they can't address the macro questions, then marketing's credibility is called into question. If analytics can address the macro, then sr. mgmt will have more confidence in it. And with more confidence, there's a better chance to get more funding to improve its ability to address the micro.
the challenge that marketing organizations face is demonstrating the ROI of the investments - which is a pre-requisite in most cases to justify the spend on incremental improvements at the micro level
I agree with Ron and Amaresh that it is not enough to use analytics to improve micro decisions, you must also be able to show the macro value of doing so. This is one of the drivers for decision optimization.
Essentially this process involves:
Defining a decision model that shows the mathematical relationships between the precictions, actions and reactions involved in a decision
Specifying overall constraints across the whole set of decisions
Running simulation/optimization routines to find the optimal set of decisions, within those constraints
Showing the (macro) impact of the incremental improvement in each (micro) decision
To show this in an example, here's a decision model for pharma marketing (grossly simplified). It has inputs (both explicit like census data and calculated like response score - a predictive model), a single action (give samples to doctors), various possible reactions and a single desired output (profit). Each link has a mathematical impact formula defined e.g. increasing the number of samples given to a doctor has what impact on their prescription volume? Building this is clearly non-trivial and requires both expertise and data analysis skills but it can be done.
Constraints in this case might include a total number of samples, a maximum number of calls to doctors and so on. Once the model is built you can run many simulations/optimizations (for each transaction in a large sample, for instance) and calculate the profit in each case. You can then show a macro view of the variation of profit across these different scenarios - the graph below shows various scenarios with varying numbers of credit offers made against total losses (from bad debt) and the resulting profit. These kinds of graphs show points of maximum value and allow you to compare scenarios (if I allow for more bad debt, can I make more profit).
This kind of analysis lends itself to supporting the macro decision making being discussed by allowing comparisons of proposed strategy to existing one in terms of how much more money might have been made. It also puts arguments about approach on a more rational and mathematical (as against emotional) level. To do it you need to build macro answers by aggregating micro decisions. What you cannot do is built macro answers by first aggregating data and then analyzing it - you need to be able to tie it back to the micro decisions.
Finally, Rolando over at Bizrules, had a couple of interesting posts also - micro or a million little decisions and macro or million-dollar decisions. I am mostly in agreement with him on these and posted comments so feel free to check them out.