Machine Learning Basics for Business Leaders

Make better business decisions by understanding machine learning—and the data that drives it.

Companies are in an ongoing state of digital transformation, and most are understandably keen to take advantage of the benefits of AI and machine learning. But it can be hard to separate the hype from the practical realities of the technology. So, if you're in a position to be making decisions for your business—whether you're a startup entrepreneur or an enterprise VP—it pays to understand the basics so that you can confidently decide when, how, and why to incorporate machine learning technology into your operations.

There are three types of machine learning—supervised, unsupervised, and reinforcement—and, if you're a decision-maker within your organization, it’s crucial to know the difference. Each type has a very different application, and the more you know, the more confident you can feel that you’re choosing the approach that meets your business goals.  

Supervised learning involves the use of labeled data, which is data that has been tagged or annotated in some way. This labeling is usually done by humans or at least with human oversight, to ensure accuracy. This labeled data is then fed into an algorithm, which uses that information to find similar characteristics in any new data it's met with. 

For example, in 2017, Stanford created a deep convolutional neural network that could diagnose skin cancer as well as human dermatologists could (when compared against 21 dermatologists). They did this by training the algorithm with 130,000 labeled images of known cancerous lesions. 

With unsupervised learning, the algorithm works with data that is unlabeled, meaning that a human has not explicitly tagged it. Instead, unsupervised learning works by identifying similarities, differences, and patterns in unlabeled data.  

A common form of unsupervised learning is clustering, which works by grouping together “like” categories of data to identify the relationship between them. The relationship is where the gold is, from a business perspective. Let’s say you operate an ecommerce website. If you were to run all of your customer data through a clustering algorithm, it could group similar data points based on purchasing behaviors, such as frequency, dollar value, last purchased, product types, and more.  

Once you have the data clusters, it’s up to you to make meaning of them. For example, you may see you have a segment of customers who purchase the same type of product, but only when those items are on sale. With that information, you might decide to send that segment a special coupon code for those products.  

On the other hand, unsupervised learning models can identify data points that deviate from the norm. This is useful for something like fraud detection. If a flurry of sales come in from a long-dormant customer account; or you get multiple orders in quick succession from the same IP address; or the billing and shipping address don’t match—something’s off. Those transactions could then be flagged for review by a human.  

Reinforcement learning is a type of machine learning that uses a simulator or training environment to teach a system (or its user). By trial and error, the system or person learns to determine the next best action within a given context. 

An obvious example of reinforcement learning is self-driving cars. They gather data from lots of different driving scenarios to prepare the system for as many as possible.  

Another application of reinforcement learning is employee training.  If your business uses heavy machinery or other complicated (and expensive) equipment, it benefits everyone to train new employees in a safe, simulated environment. An example of this is a 3D virtual reality program in which the employee practices operating the equipment and receives instant feedback on their performance.  

What all machine learning has in common is reliance on good data—and good data takes effort. You have to make sure you’ve gathered the right type of data and that it’s as accurate as possible. Data hygiene is so important that, when we embark on a machine learning project, we expect to spend up to 60 percent of our time gathering, preparing, and understanding the data. It’s that vital.  

Of course, in any data set, we expect some margin of error. For example, in Google’s emotion dataset, GoEmotions, 30 percent of the data was found to be mislabeled. The annotators analyzed 58,000 Reddit comments, labeled them according to a list of 27 emotions, and, well... humans make mistakes.  

Fortunately, we don’t need perfect data in order to make good use of it. But if you’re making technology decisions for your business, you need to weigh the stakes depending on your unique situation. For example, if you’re trying to teach a system to detect fraud, you need to have a data set that contains examples of fraud. And, since (fortunately) fraudulent transactions are a small number of the overall transactions, it would be a relatively small dataset. And so, if 30 percent of that small dataset was mislabeled, you might not have enough good data to train the system to detect fraud. Or, at least, it may not detect fraud with enough accuracy for you and your organization to feel confident relying on it.  

On the other hand, if you’re using a machine learning algorithm to perform sentiment analysis to understand how customers feel about your brand or products, you may wind up with a large data set. And in that case, even with some errors, you may see some prevailing attitudes that give you enough meaningful information to make an informed business decision.  

With all the artificial intelligence available to us, it’s still the human brain—and our ability to interpret data and make meaning of it within our specific context—that really powers these technologies. Ultimately, the success of machine learning initiatives depends on a combination of technical expertise, high-quality data, and strategic vision. Taken together, your business can harness the power of AI and machine learning to go confidently into the next phase of your digital transformation.  

Our path forward is not about perfection, but about persistent, thoughtful evolution. It's an invitation to view challenges not as insurmountable barriers, but as opportunities for remarkable achievement. 

By embracing a mindset of continuous learning, strategic flexibility, and new technologies such as AI and machine learning, we can not only improve efficiency and sustainability but also create long-term value. 

The journey of a thousand miles begins with a single, intentional step.

Are you ready to take that step? 

If you need assistance, please don't hesitate to reach out to us. We're eager to run this journey together, supporting you through a transparent, honest, and collaborative approach. We aim to work with your team to identify the most adaptable solutions perfectly aligned with your business needs and execute them efficiently, creating an immediate positive impact.

Broaden your perspective and take control of your success.