We’re in deep, friends, when the first thing we associate with the word model is math.
But cheers to being in deep, right? Here’s good company when all models are mathematical.
My tone when talking about models is different this year than last. A little less than a year ago, I tread on this same topic with much more reserve and doubt in my heart. Today, I visit the topic with less ego and more optimism. I don’t know if the reasons for my evolution in thinking are important, but I know the fact of it is: models have value. They are yet another tool for our toolbox, and one, in my mind at least, worth mastering.
Last year, I said that no model would ever replace an agronomist. I still think that’s true today, but I can edit some of the fear from that statement. No model will ever replace an agronomist, but they can make for a trusty sidekick.
In the hands of the right person, the information and predictive power of a model are incredibly useful to our pursuit of ever better efficiency and ever higher yields. The right person will do their due diligence with a given model—not use it to automate decisions. Therein is the danger ever present in a model—that it will be taken for truth instead of as the package of information it actually is.
When I wrote rather dismissively about models last year, my mode of thinking was very much either/or. A grower has a question and takes it to either a model or an agronomist. My bet was—and still is—that the agronomist would provide the better answer almost every time. But there’s no reason this has to be either/or. In fact, it shouldn’t be. It should be an and situation: a grower has a question and takes it to an agronomist who suspects an answer and seeks to sharpen and hone that answer with the mathematical data a model provides.
When we make it an and situation, we can triangulate. You know that GPS increases in accuracy as more data points are incorporated. The same goes here. The incorporation of models into our practice will increase our accurate diagnosis of a problem and pave the way to a more effective solution. This way, we end up at the front door of the right hotel, instead of the one down the street.
But like I said, due diligence, and here’s the due diligence part: we need some ground truth. I can’t make a case to y’all to embrace models without doing my part to check the mega data on which they’re built. So that’s the project. I’m thinking way ahead here, friends, but come next summer, we here at CVA are going to take a few promising models for a test drive. We’re going to field test them, check their accuracy, learn their personalities. We’ll put the poor or irrelevant models out to pasture. The good ones we’ll put into our toolbox so that when you come knocking, our answers are more accurate, more confident, and more tailored the details of you and your operation than ever before.
Because a model will never replace an agronomist, but when their powers combine…