I have made a nice living building forecast models for companies, large and small. But people make big decisions all the time without a model. People make predictions about the future without a model. So it begs the question..."Why bother with a forecast model?!" As I sit and ponder this question over coffee, three answers come to mind and rescue me from my existential crisis.
It forces you to think about your market and the dynamics that play out in it in a structured way. Without exception, one of the most common comments I hear from clients after going through the forecasting process with me is: "It's great to have the numbers, but it's amazing how this process has made me think about my business in such a different way". The disciplined and detailed consideration of both controllable and uncontrollable influences on the relative success of your product(s) leads to actionable insights that help you create that very important ingredient in any business recipe: strategy.
It allows you to explicitly and transparently state assumptions. I am asked all the time about the accuracy of forecasts. We can talk all day long about cognitive bias and over-fitting of mathematical models, but one of the most influential factors on the accuracy of a forecast is the assumed context for the predictions. For example, first year sales may be very tied to sales force FTE's and marketing spend. If the level executed is 50% of that assumed when populating the model, it should be no surprise if the actual sales numbers are less than forecast. But, after the fact, the assumptions in your head at the time of the forecast cannot be retrospectively evaluated against the actual performance. A well documented forecast model can be the cure for the dreaded "what were you thinking when you came up with those numbers?"
It allows you to evaluate the impact of being wrong. Let's face it, we are forecasters, not fortune tellers. Expanding on number 2 above, documenting our assumptions allows them to be challenged. Challenges to assumptions, either by ourselves or by committee, leads to what-if analyses. What if we launch with 50% fewer sales force FTE's than we are assuming today? What if the product performs 10% better than we assume it will today? What will it cost to achieve that additional 10% performance, and does it result in increased NPV? Forecast models allow us to ask and answer those kind of questions in a systematic way.
So, as I approach the bottom of my coffee cup, I can now lean back in my chair with a sigh of relief; my life's work may have some value after all!