Thursday, February 22, 2018

The Best-Laid Predictions

Forecasting and future planning, in all their myriad forms forms, represents a struggle against an invariable truth: that the future is unknown.   At best, our most careful research and attentive modeling is based on the assumption that what is true in the recent past will continue to gold true in the future.  The most horrific blunders in history are laid upon that foundational assumption – but then, so have the greatest success.   To fail, to succeed, or to act at all requires some sense of what will happen in the future, particularly in reaction to something that does not exist at all in the present.

This can be difficult to observe in the aggregate: when financial plans fail, they do so with a shrug.  Something went awry “in the market” or “in the international economy” and there is not even the ghost of an attempt to recognize the flaw in the analysis – generally because in spite of the catastrophic failure, there are firm plans to apply the exact same analytical model to the very next decision.   The model was perfect – the world went wrong, and it’s assumed that the world will follow the rules next time and the model will work.

But where predictive models can be observed in a more granular manner, the inherent flaw of the present-future assumption becomes clear.    Ask an individual person what he plans to do in the future, and you can plainly see how he will speculate, confabulate, and make up plausible stories about a hypothetical situation in which he plainly has no clue how he will actually behave.    Even the most sincere respondent, reacting to the most unbiased instrument, doesn’t really know what he is going to do in a real situation that has not yet arisen.

When investigating behavior, you are implicitly making people think about things they do not normally think about – and often hectoring and harassing them until they lie in a manner that suits your preferences.    As a subject why he chooses to purchase a given brand of mustard, and he will shrug and say, perhaps, that “I like it.”   Press upon him, asking the reason he should like this brand over another that is virtually identical in all regards, and he will struggle to respond.   If you don’t like the lie he tells you, press him further until he tells a lie that you like – and then write that down and include it in a study where you declare, with pompous certainty, that you have identified “the real” reason consumers prefer one brand over another.       Use that as a basis of making your strategic decisions … and see what happens.

Consider that the very act of studious observation takes the customer out of the context in which a decision is made, if they ever are in such a context to begin with.   Consumer taste-tests of products rest on the assumption that customers consider the taste of a product to be the critical factor in making a buying decision – and do not consider for a moment that it will be any other factor.   In rare instances in which stores offer samples at the point of purchasing decision so that a customer can compare the flavor of two products they have not yet purchased, this may hold true – but this seldom ever happens.   And this is why Pepsi consistently beats Coke in blindfolded taste-tests, yet consistently sells less across virtually all markets where both are offered.


In one sense, asking the customers for their opinion seems a better course that relying on insider opinions – and it may be a more democratic method of deciding among options, but it is not necessarily accurate.   Customers do not know how they will decide in real-life purchasing situation and are speculating.   Observation studies can be very useful in exploring the present and recent past and, certain enough, the patterns of the past will tend to perpetuate if nothing else is changed – but they do so in a very superficial manner: one can easily observe and measure what is being done, but this does not accurately indicate the reason it is done – and it is the reason and the conditions, more so than the superficial act, that is most useful in the prediction of future behavior.

No comments:

Post a Comment