Blog #1 - The Strategic Planning Process as a “Learning” Process

A lot of the characteristics of the strategic planning process resembles the machine learning theories widely practiced both academically and commercially.

 

On the surface, the strategic planning process requires the evaluation of the past experience and the analysis of the current situation, whereas machine learning relies on datasets collected from past and current events. The results come out of the planning process conclude past decisions and are intended to thrive in the future, whereas machine learning algorithms learn functions from data and are intended to provide useful results for future data. Furthermore, as suggested in the reading, most of the time, figuring out the right category of strategies is more important than crafting the exact strategy itself. Similarly, for most of the machine learning algorithms that work well, the feature engineering and choice of architecture is much more influential than the model training and calibration.

 

The parallel between strategic planning and machine learning goes on and on. It is fair to view the strategic planning process also as a "learning" process: learning from the pros and cons of past strategies and current situations, and then utilize that learned knowledge to craft a "solution" to deal with future challenges. Since we’ve established this series of similarities between strategic planning and machine learning, it’s also illuminating to examine what are the implications of certain phenomena in machine learning theories to the strategic planning process.

 

In machine learning, one of the most common evaluation of an algorithm is if it is overfitting or underfitting. More specifically, overfitting refers to the fact that the algorithm “learns” so much of the given training data that it loses the generality to be applied to wider range of data, whereas underfitting refers to the fact that the algorithm does not “learn” enough information from the data and the algorithm will mostly and dominatingly work on the assumptions programmers made on the metadata and architecture. Projecting the concepts of overfitting and underfitting to the strategic planning process, the most appropriate analogy would be how much we analyze and rely on past experience and current situation to determine our strategy. In the case of overfitting, the equivalent example will be that we rely too much on how our strategies work in the past and almost copy and paste successful strategies from the past for our new strategy. Unfortunately, such a choice loses the ability to evaluate new challenges and opportunities the new age has to offer and ultimately risking the potential to most properly adapt to current and future situations. On the other hand, underfitting is the most analogous to that we build the new strategy mostly on our perception of the situation without examining the real situation carefully, rigorously, and quantitatively. Such choice prevents us from really understand the real situation and falls gullible due to the limitation of our personal perceptions.