Focus on success, not perfection: Look at this data science algorithm for inspiration
Image: iStock
The quest for perfection often gets in the way of success. For example, Coca-Cola wasted millions on New Coke when the Old Coke was just fine. It might surprise you to learn that a frequently used data science technique can serve as a good reminder for leaders to strive for success, not perfection.
Let Naive Bayes be imperfect
Naive Bayes is a machine learning technique that's primarily used for classifying text. For instance, Naive Bayes may be used to identify spam in email. The algorithm takes your email text as an input, does its magic, and then determines -- or classifies -- your email into one of two categories: spam or no-spam.
This is unremarkable for a machine-learning algorithm; however, if you look under the hood, the algorithm makes bold assumptions that would ostensibly render it ineffective. For instance, word order or context is not considered, so the phrases "monkeys drive crazy cars" and "cars drive monkeys crazy" are treated the same. What is remarkable is that, notwithstanding these seemingly reckless assumptions, it works.
Naive Bayes illustrates that real life isn't always logical or formulaic. The fact that Naive Bayes works irritates many data scientists. A common slang used for Naive Bayes is idiot Bayes. Some data scientists feel Naive Bayes needs to be improved because, even though it works, they know there's something wrong with how it works.
The trouble with trying to perfect Naive Bayes is that you quickly run into a problem solving mess. Trying to account for all the semantics and contextual nuances of the English language is a fool's errand. And for what? The algorithm already works well, so what are they hoping to improve? This is the classic Achilles heel of our dear data scientists: Their incessant pursuit of perfection when success is already in-house. We'll give them allowances because that's their nature, but leaders and managers, you don't get the same pass.
When ignorance is bliss for leaders
Naive Bayes teaches us to focus on results, not methods. Methods do matter, but let's get straight on the means and the ends.
I once heard a comedian tell a joke about the game show, "Who Wants To Be A Millionaire?" The contestant is presented with four answers to choose from, and when they seem stuck, the host (originally Regis Philbin) encourages them to talk through the answer. The joke was that the contestant would talk it through with clearly fallacious and erroneous reasoning -- and then select the right answer. Something like, "Well...I know it was an African-American community leader around the turn of the 20th century, so I'll go with...Jimmy Carter." As funny as that sounds, as long as Jimmy Carter is the correct answer, the contestant wins the money.
We need to take this naive approach with leadership. Sometimes we get too hung up with our methodologies and forget about why we studied them in the first place. Kotter's eight-step process for leading change is great to know, but if you can bring about change in the first three steps, you don't need to finish the other five. If someone achieves self-actualization with low self-esteem, we don't need to investigate why Maslow's Hierarchy of Needs isn't working.
We can use Naive Bayes for practical, everyday purposes even though we know it shouldn't work as well as it does. Let the data scientists focus on the methods -- that's what they do. But as a leader, you'd be remiss if you joined them on the path to perfection. Stay on the road to success and keep your eyes on the outcome.
Final thoughts
Learning from your environment is the hallmark of a great leader. If your strategy incorporates data science, you probably have a great lesson that's right under your nose.
Your data scientists are using Naive Bayes for its pragmatics, not its elegance. As a leader, you should be naive in this respect as well. If you're lucky enough to stumble upon something that works, don't question it -- just appreciate what you have and focus on the next challenge.
There's nothing classic about Classic Coke -- that was just Coca-Cola's idea of undoing a bad mistake. They should've just called it Naive Coke and left it alone.
